‘Best of’ the Data Athlete

Every once in a while, it’s important to look at where we’ve been in order to understand where we’re going to go next. That’s why in this month’s post, I’ve organized and summarized a “best of” digest for you.

Every once in a while, it’s important to look at where we’ve been in order to understand where we’re going to go next. That’s why in this month’s post, I’ve organized and summarized a “best of” digest for you.

Think of the attached as a simple, and valuable crash course for becoming a better data athlete. I selected these best of “Data Athlete” pieces based on your feedback. I would be more than happy to hear your feedback about this digest, as well.

best of

No. 1: ‘The Most Important CRM Metric You Might Be Missing.’ CRM is perhaps the most over-used term in marketing today. It means so many things to so many organizations, as well as the people in them. In this column, I focused on the value of a customer, both historical and actual, and potential or predicted.

No. 2: ‘Bigger Is Better — How to scale-up customer acquisition smarter.’ This article explores effective strategies for leveraging customer data in order to drive and scale up customer acquisition.

No. 3: ‘The Cost of Perfection.’ This column weighs the ideals of achieving perfection in raw data against the costs of inactivity and reactivity.

No. 4: ‘Better Beats Bigger Data, Every Time.’ This article focuses on common issues marketers face with marketing database development and provides insights on how to overcome these challenges to leverage data and create marketing and business value.

Drive Your Buyer’s Lifecycle, Increase Revenue and Retention

The process of acquiring and sifting traffic into engaged, and ultimately buying, prospects is critical to your customer acquisition efforts. Managing your audience is often referred to as the early stage of the “Customer Journey.” In this post, we’ll focus on the core and most pivotal part of your relationship

The process of acquiring and sifting traffic into engaged, and ultimately buying, prospects is critical to your customer acquisition efforts. Managing your audience is often referred to as the early stage of the “Customer Journey.” In this post, we’ll focus on the core and most pivotal part of your relationship with the consumer — purchasing from your brand.

Based on some years of experimentation and measurement, we can share a simplified and highly actionable approach that can make a difference in how you value and grow value among customers. This is the buyer lifecycle.

Mike Ferranti chartProspects: Before They Are Customers
Prospects, of course, come from many places: word of mouth and direct visits to your website and to your retail stores. Advertising and search drives them to on- and off-line points of sale. Prospects can be those who simply signed up on that ever-larger email signup popup on your homepage, or those who put items in a cart and “almost” purchased, but abandoned.

But prospects can also be those who we leverage statistical intelligence to hand-pick. Not just look-alikes but the “buy-alike” prospects with the highest potential value. See my prior column called “The Most Important CRM Metric You Might Be Missing.”

All of these prospects have the same thing in common, they have not purchased, and a level of investment and communications will be required to drive them to the next step. This cannot be overlooked without consequence. Prospects, regardless of the level of engagement or targeting, have a massive, and in some cases, a predictable difference from the buyers you seek to drive incremental sales from — they lack the most powerful signal of all behaviors — actually spending with your brand. Commonsensical enough, perhaps — but the prospect ‘batch and blast’ marketing that pervades retail emailers typically makes the challenge harder. Customer Intelligence is required to target, learn and test your way into viable prospect conversion strategies. We reiterate this point as it is often assumed that prospects, when contacted, will just buy — and they don’t. The bar is higher (see “Bigger is Better: How to Scale Up Customer Acquisition Smarter” for how to target the right customers, and the sophistication your competitors may be leveraging already).

To be sure, an analysis of your prospect base, which in a great many organizations is actually called the “email file” — another issue, in itself ― will help you determine who is likely to buy and who is not. This can be achieved by considering engagement measures, like opening and clicking your emails, visiting the website and micro-conversions. While these behaviors are correlated with the move from prospect to buying, it is not uncommon for the “average” prospect files to contain too many records of individuals who will never buy — they are lookers, not buyers. They may lack the means, intent or occasion to buy — or they may have experienced some change in their lifestage that moved them out of the market for your product. The opportunity is in identifying the highest value prospects and investing more thoughtfully in converting them.

Customer Value: Narrowcasting vs. Broadcasting

The traditional model for customer acquisition has essentially been a broadcast approach, reaching a large audience generally descriptive of the customer base. Contrast this with what is sometimes described as “narrowcasting.”

Virtually every brand we’ve met with in the last few months is hungry for new customers: The war for the customer is on. For more on growing your customer base, consider reading “Bigger is Better: How to Scale Up Customer Acquisition Smarter,” which is an article we published recently about how to grow your customer base.

Many organizations are hooked on customer acquisition. That is, in order to hit sales plans for the organization, new customers will be required in large numbers. It’s about as easy to kick the “acquisition addiction” as it is to kick any other for most brands. Try going without coffee suddenly, and see how your head feels. It’s not very different from reducing a business’s dependence on customer acquisition as a means to achieving revenue and profit targets.

Organizations that need ever larger numbers of new customers to achieve growth goals eventually will find the cost of acquiring incremental net new customers can become prohibitive.

Broadcast vs. Narrowcast
The traditional model for advertising and customer acquisition has essentially been a broadcast approach, reaching a large audience that is generally descriptive of the customer who a brand believes to be a fit. Contrast this with what is sometimes described as a “narrowcasting” strategy. Narrowcasting uses customer intelligence to understand a great number of discrete dimensions that a consumer possesses and can leverage statistical methods to validate the accuracy and predictiveness of targeting customers through these methods.

The chart below, depicting the value of customers acquired through traditional broadcast capabilities upfront and over time helps illustrate why “broadcast” strategies for customer acquisition alone aren’t enough.

Research for Mike Ferranti blog

Broadcast Acquisition Strategies Lack Focus on Customer Value
Large numbers of customers have been acquired in a trailing 13-month window – lots of them. The challenge is this cohort of customers has been acquired without adequate consideration of the right target.

Consider the fact that the target customer value of average or better customers is around $500. In the example above, the marketer has acquired a large number of customers who are lagging in their economic contribution to the business. While the customer acquisition metrics may look good, this was a large campaign and produced several hundreds of thousands of customers over its duration – the average value of those customers is quite low indeed.

Low Customer Value Manifests Itself, Even if Acquisition Volume Is High
When sales targets are rising, it becomes harder to justify the high cost of customer acquisition if the customers previously acquired are underperforming. This leads to a very common bind marketers are placed in. The only way to “make the number” is to acquire more and more.

The most competitive and high quality businesses steadily acquire and have a robust customer base whose economic contribution is materially higher. Consequently, profits are higher, and we have a fundamentally better business.

Oftentimes, “broadcast” advertising approaches define the target with a single criteria like age, income or geography. This can be effective, especially when the media is bought at a good value. However, “effective” is almost always defined as “number of customers acquired.” This of course is a reasonable way to judge the performance of the marketing – at least by traditional standards.

There is another way to measure the success of the campaign that is only just beginning to be understood by many traditional “broadcast” marketers: customer value. The chart above shows that this cohort of acquired customers had relatively low economic value.

Root Causes of Low Customer Value
What are the causes of low value? It would be fair to start with the ongoing marketing and relationship with the customer. Bad service could keep customers from returning. Poor quality could lead to excessive returns. Over-promotion could drive down value. Getting the message and frequency wrong could lead to underperformance of the cohort. These are all viable reasons for lower value that need to be rationally and methodically ruled out prior to looking elsewhere.

Therefore, if operational issues are not clear – either through organizational KPI tracking, or simply by monitoring Twitter — then a marketing professional needs to start looking at three things.

  1. The Target (and Media)
  2. The Offer (and Message)
  3. The Creative

Given the target is historically responsible for up to 70 percent of the success of advertising, this is the first place a professional data-driven marketer would look.

Target Definition Defines the Customer You Acquire, and It Drives Customer Value.
A fact that is often overlooked is that target definition means not just focusing efforts and advertising spent on consumers who are most likely to convert and become customers, but it also defines what kind of customers they have the potential to become.

In conversations with CMOs, we often discuss “the target customer” or the “ideal customer” they wish to introduce their brand to. The descriptions of course vary by the brand and the product. Those target definitions are often more qualitative in nature. In fact, only about 30 percent of CMO’s we engage with regularly are focused on using hard data to define their customer base. While these are helpful and create a vocabulary for discussing and defining who the customer is, those primarily qualitative descriptors are often sculpted to align with media descriptors that make targeting “big and simple.”

“While simplifying is good business, when simplicity masks underlying business model challenges, a deeper look will ultimately be required, if not forced on the organization.”

While we would not refute a place for those descriptors of a valued consumer, they do fall short of true target definition. Ideally, the process of defining the customer who a brand wishes to pursue must begin with a thorough inventory of the customers it already has, and a substantial enhancement of those customer records which provides vibrant metrics on affluence, age, ethnographic, urbanicity, purchasing behaviors, credit history, geo- and demo-graphics, net worth, income, online purchasing, offline purchasing and potentially a great deal more.

The Most Important CRM Metric You Might Be Missing

Virtually every organization we have worked with in the past year is working on managing, improving or optimizing their relationships with customers. This work falls under the umbrella term “Customer Relationship Management” or “CRM.” It is, of course, the oldest “new thing” that marketers have focused on, en masse, for a long while.

CRM keyVirtually every organization we have worked with in the past year is working on managing, improving or optimizing their relationships with customers. This work falls under the umbrella term “Customer Relationship Management” or “CRM.” It is, of course, the oldest “new thing” that marketers have focused on, en masse, for a long while.

“CRM,” as it is, is a term that means many things to many different organizations and to different individuals in those organizations. This has created some confusion and leads to missed expectations in organizations.

Through meetings and executive interviews with brands, we have found that the majority of marketers will eventually describe the primary purpose of “CRM” initiatives as growing the value of customers who do business with them. We say “eventually,” because the initial responses to the question “what is the objective of your CRM initiative” gets quite a few answers including:

  • Know our customer better
  • Improve communications with our customer
  • Grow customer relationships (the most common response, and also the least actionable)
  • Decrease the usage of promotion
  • Reduce the volume of emails sent

These are just a few of the ways the organizations we work with begin to define their CRM initiatives; but to really make a difference in the business, CRM needs a clearly defined vision:

Intelligently managed customer relationships grow customer value. It drives incremental profit by either reducing the cost of promotion or driving incremental profitable revenue. CRM requires ongoing testing and learning, which can strategically inform customer acquisition and, in turn, increases the quality of the business.

“Intelligently managed customer relationships grow customer value. It drives incremental profit by either reducing the cost of promotion, or driving incremental profitable revenue.”

Can You Really Grow the Value of Your Customers?
Given the continuing trend of technology and data-driven CRM, it often comes as a surprise that few organizations have a heavy concentration of high-value customers. In fact, it’s the norm.

In a study Kaplan and Anderson published in the Harvard Business Review, the following was found across all industries:

  • 10 to 25 percent of customers drive 100 percent of profits
  • 50 to 60 percent deliver no profit at all
  • 10 to 25 percent deliver negative profitability

Some may find the magnitude of these facts surprising, perhaps even alarming. Not surprisingly, these profitability metrics correlate entirely to our experience across many dozens of organizations in working with their customer databases. What is sometimes an “uncomfortably large” percentage of revenue and profit is driven by a small group of the most valuable customers. In the luxury segment, where some brands have created an “accessible luxury” segment, the results grow even more staggering.

One example where we’ve seen this is among premium luxury brands that have grown “more inclusive” in their customer base. The concentration of customer value in the organizations is often almost exclusively in the top 10 to 15 percent of customers. When we have revealed this insight and evidence, the very business model may need to be rethought. To be sure, across all segments, customer value is a very big deal to all organizations — and, therefore, CRM.

Do Brands Have ‘Bad’ Customers?
This is a topic that is also hard to engage on. Often, marketers dedicate many hours and PowerPoint slides to focusing on the successes, and how good our customers are for our business. That’s an entirely intuitive point. These great customers also have an inverse; that is, customers whose value isn’t quite so great.

Some of these organizations have a material number of what might be called “bad customers,” altogether. But given that customers are the key element to realizing value in every business, how then can they be “bad”?

Let’s be clear, “bad” may carry a visceral sense of judgment. That’s not the point here, at all. The point is to meaningfully differentiate between customer groups or segments that naturally exist today in your database. “Good” or “bad” for the data-driven marketer really means how profitable the group is, or if it’s profitable at all. Simply put, a “bad” customer” must exist if a “good” or “great” customer does. Perhaps more “PC” — all customers have value, yet the value they hold for an organization is very, very different.

“All Customers have value, yet the value they hold for an organization is very, very different.”

You may even have a term for a segment of your customer base that you can’t afford to service well as “cost-control” customers. This happens in financial services, for example, where cost control may mean higher fees and online self service only. While that specific model does not necessarily apply to every business, all businesses have various segments of customers by value — both realized, and potential.

An Example: The Luxury and Accessible Luxury Categories
In the luxury category, brands sometimes become “more inclusive” (for example, in 2008 and at the depths of the Great Recession), which often means either markdowns or a product line for the “accessible luxury” category. As a result of this, customer value inevitably declines. In our experience, that decline was driven by decisions years earlier to scale at the cost of customer quality.

In these scenarios, if you were managing a CRM initiative, you’d have what’s known as a “dual-universe” problem — you can’t manage the value of these very different customers the same way. They may require a different P&L to account for them, and understand their value to the business.

A simple starting point in understanding a “dual universe” goes like this: Segment out your customers into the two groups — those who buy your true premium product, and those who have bought everything else. Analytics can then be leveraged independently across those groups.

The key to understanding if you have good and bad customers is, of course, the speed and dexterity you have to analyze customer data and your ability to measure and monitor changes in customer value by cohort. That’s a tall order for a lot of organizations today. Most are still focused on revenue through acquisition, rather than a strategic view where customer value is crafted first through the unique kind of customer acquired.

Good Customers — The Heart of Your Business
Good customers typically have longevity. Good customers purchase frequently, they have higher order sizes, or monetary value to your organization, they tell their friends about you, and while they appreciate a product they like on sale, they can also pay full price to get what they want.

Most importantly, while great customers generally cost more to acquire, and are harder to come by — good customers are quite profitable.

When a customer is considered good in most situations, they sometimes have the potential to become great ones. And therefore the mission of the CRM practitioner becomes, in simple terms, to ID the similarities and differences between them, make communication more relevant, and shape the value of each sale systematically. Growing customer value for your “good customers” can fill several of these columns, and we’ll put a series on migrating the good customers to great ones. (leave me a comment, or email me if you’d like to see those in the next couple of months).

Great Customers, or ‘Gold Customers’ — The Backbone of Your Business
The challenge for these “great customers” is they are often few and far between. If you’re in a business, where you have many great customers, you are either very, very fortunate, or you have not created a meaningful stratification of customers by value! This is one of the reasons that an intelligent segmentation of customers by value is an eye-opening engagement for most marketers and CRM practitioners.

Great customers, in most cases, are not only few in number but — counter to what may be one’s “gut feeling” — they quite literally carry the business. If you were to assume the contribution of customers to your bottom line followed a normal distribution, (think the bell curve, with a big fat middle), you would be quite surprised by what it most likely looks like. That contribution is stacked heavily to the top standard deviation, or way to the right side of the curve.

The insights we glean over time and across industries on organizations’ “Gold Customers” is the genesis and the reason CRM as a practice exists today.

“The Insights we have gleaned over time and across industries on organizations’ ‘Gold Customers’ is the genesis for and the reason that CRM as a practice exists today.”

The Best Way To Influence your CRM and Customer Value — Smarter Acquisition
This comes as a curveball to many CRM practitioners, especially those early in their CRM careers and experience. Nothing but nothing will change the performance of your database more meaningfully than adding more customers with higher potential value.

Put another way, great — or “Gold Customers” — are the backbone of a business, in that they are primary drivers of profitability, and they are the reason we’re engaging in CRM. So it’s imperative that we not only treat them differently and market to them wisely — but very simple math suggests we must also be acquiring more of them to increase the value of our database, our customer base and our business.

The Most Important Metric of Your CRM Strategy: Potential Value
There are many ways to measure your customers, their behaviors and their value. Concurrently, the most strategic way to grow your business and the value of your CRM initiatives is to collaborate with and inform your customer acquisition; that is to say, you can sculpt potential value through who you market to in the first place.

Customers who can’t afford you, don’t have the habits, beliefs, credit or lifestyles that your great (most valuable) customers do simply won’t or can’t buy like those who do. Those who do are your MVCs (Most Valuable Customers) and those who are ever further from this ideal are your least valuable.

Therefore, there is nothing we can do as marketers and as CRM practitioners that will improve the value of customers now and over time more so than acquiring more of the right ones. The strategy to how we do that is covered in another important article I’ve published as part of the body of work in this column on, “How to Scale-up Customer Acquisition Smarter.”

When you take a holistic view of your marketing, and place the appropriate value on the role of customer intelligence from CRM into your customer acquisition approaches, you can have an ever greater impact on the No. 1 metric we discuss herein — the potential value.

A high-potential value in the customer database then can be translated into ever-greater revenue and profitability, in a scalable and methodical fashion. While potential value is unlocked through all of the strategies and tactics we engage with through CRM — it all starts the most important “inputs” to your CRM — the customers themselves; moreover, acquiring the right ones.

5 Data-Driven Marketing Catalysts for 2016 Growth

The new year tends to bring renewal, the promise of doing something new, better and smarter. I get a lot of calls looking for ideas and strategies to help improve the focus and performance of marketers’ plans and businesses. What most organizations are looking for is one or more actionable catalysts in their business.

The new year tends to bring renewal and the promise of doing something new, better and smarter. I get a lot of calls looking for ideas and strategies to help improve the focus and performance of marketers’ plans and businesses. What most organizations are looking for is one or more actionable marketing catalysts in their business.

To help you accelerate your thinking, here is a list of those catalysts that have something for everyone, some of which can be great food for thought as you tighten up plans. This year, you will do well if you resolve to do the following five things:

  • Build a Scalable Prospect Database Program. Achieving scale in your business is perhaps the greatest challenge we face as marketers. Those who achieve scale on their watch are the most sought-after marketing pros in their industries — because customer acquisition is far from cheap and competition grows more fiercely as the customer grows more demanding and promiscuous. A scientifically designed “Prospect Database Program” is one of the most effective ways great direct marketers can achieve scale — though not all prospecting databases and solutions are created equally.

A great prospecting database program requires creating a statistical advantage in targeting individuals who don’t already know your brand, or don’t already buy your brand. That advantage is critical if the program is to become cost-effective. Marketers who have engaged in structured prospecting know how challenging it is.

A prospect database program uses data about your very best existing customers: What they bought, when, how much and at what frequency. And it connects that transaction data to oceans of other data about those individuals. That data is then used to test which variables are, in fact, more predictive. They will come back in three categories: Those you might have “guessed” or “known,” those you guessed but proved less predictive than you might have thought, and those that are simply not predictive for your customer.

Repeated culling of that target is done through various statistical methods. What we’re left with is a target where we can begin to predict what the range of response looks like before we start. As the marketer, you can be more aggressive or conservative in the final target definition and have a good sense as to how well it will convert prospects in the target to new customers. This has a powerful effect on your ability to intelligently invest in customer acquisition, and is very effective — when done well — at achieving scale.

  • Methodically ID Your VIPs — and VVIPs to Distinguish Your ‘Gold’ Customers. It doesn’t matter what business you are in. Every business has “Gold” Customers — a surprisingly small percentage of customers that generate up to 80 percent of your revenue and profit.

With a smarter marketing database, you can easily identify these customers who are so crucial to your business. Once you have them, you can develop programs to retain and delight them. Here’s the “trick” though — don’t just personalize the website and emails to them. Don’t give them a nominally better offer. Instead, invest resources that you simply cannot afford to spend on all of your customers. When the level of investment in this special group begins to raise an eyebrow, you know for certain you are distinguishing that group, and wedding them to your brand.

Higher profits come from leveraging this target to retain the best customers, and motivating higher potential customers who aren’t “Gold” Customers yet to move up to higher “status” levels. A smart marketing database can make this actionable. One strategy we use is not only IDing the VIPs, but the VVIP’s (very, very important customers). Think about it, how would you feel being told you’re a “VVIP” by a brand that matters to you? You are now special to the brand — and customers who feel special tend not to shop with many other brands — a phenomenon also known as loyalty. So if you’d like more revenues from more loyal customers, resolve to use your data to ID which customers are worth investing in a more loyal relationship.

  • Target Customers Based on Their Next Most Likely Purchase. What if you knew when your customer was most likely to buy again? To determine the next most likely purchase, an analytics-optimized database is used to determine when customers in each segment usually buy and how often.

Once we have that purchase pattern calculated, we can ID customers who are not buying when the others who have acted (bought) similarly are buying. It is worth noting, there is a more strategic opportunity here to focus on these customers; as when they “miss” a purchase, this is usually because they are spending with a competitor. “Next Most Likely Purchase” models help you to target that spending before it’s “too late.”

The approach requires building a model that is statistically validated and then tested. Once that’s done, we have a capability that is consistently very powerful.

  • Target Customers Based on Their Next Most Likely Product or Category. We can determine the product a customer is most likely to buy “next.” An analytics-ready marketing database (not the same as a CRM or IT warehouse/database) is used to zero-in on the customers who bought a specific product or, more often, in a specific category or subcategory, by segment.

Similar to the “Next Most Likely Purchase” models, these models are used to find “gaps” in what was bought, as like-consumers tend to behave similarly when viewed in large enough numbers. When there is one of these gaps, it’s often because they bought the product from a competitor, or found an acceptable substitute — trading either up or down. When you target based upon what they are likely to buy at the right time, you can materially increase conversion across all consumers in your database.

  • Develop or Improve Your Customer Segmentation. Smart direct marketing database software is required to store all of the information and be able to support queries and actions that it will take to improve segmentation.

This is an important point, as databases tend to be purpose-specific. That is, a CRM database might be well-suited for individual communications and maintaining notes and histories about individual customers, but it’s probably not designed to perform the kind of queries required, or structure your data to do statistical target definition that is needed in effectively acquiring large numbers of new customers.

Successful segmentation must be done in a manner that helps you both understand your existing customers and their behaviors, lifestyles and most basic make up — and be able to help you acquire net-new customers, at scale. Success, of course, comes from creating useful segments, and developing customer marketing strategies for each segment.

Bigger Is Better: How to Scale Up Customer Acquisition Smarter

While we’re all focused on delivering our holiday plan, every CMO I’ve spoken with in the last three months is focused on the same things. In 2016 “we need to achieve greater scale” — in other words, to get bigger.

While we’re all focused on delivering our holiday plan, every CMO I’ve spoken with in the last three months is focused on the same things. In 2016 “we need to achieve greater scale” — in other words, to get bigger. It would seem for sure, bigger is better. Bigger is the American way. Bigger sales, bigger profits, bigger staff and teams, bigger assortments, and bigger margins. Yes — bigger it seems is much better to CEO’s, CMO’s and board members everywhere.

“The Onceler” from Dr. Seuss’s “The Lorax” succinctly said:  “… I’m figgering on biggering.”

For marketers, new customer acquisition is the most effective way of to grow the organization “bigger.”  It’s the lifeblood of growing organizations, and is generally considered a sign of a business’s overall health.

Yet customer acquisition can be resource and budget intensive, even if an solid value, and marketers require effective more intelligent approaches to achieve greater scale while maintaining budget guidance.

So as you lay out plans for 2016 and how you’ll scale your business “bigger” — and hopefully better as well — it likely makes sense to think through the most effective ways to drive scale. Leveraging customer data with ever increasing intelligence is the common thread from some of the best strategies I’ve worked with brands on successfully over time.

Programmatic Advertising
Advertising is being increasingly automated, and over time, it’s expanding across web, mobile, and now television and “over the top” television (think web-based TV — where a programmatic buy may land your ads on Netflix one day soon). Programmatic display advertising is largely however a web based phenomenon. It adds data and certain controls to your ability to target ads that are very different than the traditional “insertion order” contracts of yesteryear where you were locked into a certain number of ads for a certain amount of time at a fixed cost. Now you can choose your targeting criteria using oceans of third party data sets that are overlaid through cookie exchange.

Some of the downsides of “PA” — well it’s still advertising. That is, we’re displaying ads that are increasingly more relevant, but are they really seen? Do they have the palpability or engagement of other forms of advertising? Generally not.

Viewability is another problem, as is fraud. Often times PA’s don’t provide an accurate identification of where your ads are even being served, much less if they are served well below the fold, where basically you’re paying for an ad that no one’s ever seen, or is going to see.

Also, for the highest value ads, more advertisers are likely to bid on them. Like any bid based vehicle, programmatic advertising models ultimately bid up the cost of reaching the consumer that more marketers want, effectively making for maximum competition (and again, higher ad prices) even if excess inventory may drive some ad inventory costs down today. Remember, this competition amongst advertisers bidding up programmatic advertising is in large part designed to best serve the media outlet, even while delivering better ad products to the market.

List Rental and Univariate List Selection
With the explosion of data access and availability, it’s easier than ever to rent a list and prospect to individuals who visit Morningstar mutual funds, live in New York City or are active runners, and virtually anything else you can dream up. List quality remains the challenge, and testing your way into a process is critical.

Univariate list selection is typically a select by an interest, though it can be on many other singularly focused targeting criteria. It’s a list where members share one criteria, such as an interest in sailing or presence of children.

The downside of traditional list rental is price and performance. You really have to be buttoned up to make traditional list rental on a CPM basis work well, and you tend to get it performing best when you or your agency have developed a relationship with the data source, and earn preferential pricing and terms. There are instances where the target and the list line up very well, and thoughtful univariate list rental can work well — typically these are substantially niche or “edge” cases or, very much ‘mass market” circumstances. The huge “middle” is high risk for a univariate list, in my experience.

Multivariate List Selection
This is list rental as well, but instead of working off a single common attribute, a data broker takes your set of multiple “selects” and produces a “count” of the number of matches. Examples can include age (usually banded, not to the year) and other demographic attributes like income, presence of children, and interests. It is relatively painless to create a target that sounds a lot like your target customer, if you work with reputable data vendors.

Combining multiple data points zeroes in on your target, and is usually expected to improve response. Since each select or targeting criteria adds cost to the list rental, your conversion requirements must go up to achieve your economic goals for the program. The same can be said for programmatic display buying. You pick the criteria for the target you want to advertise to, at a cost.

Most “look alike” marketing methods are little more than a multivariate selection.

The question it will help to answer beforehand, if you can, is are you choosing the right variables or criteria? If it sounds right — it may be, but that’s not a guarantee. Because data vendors don’t have a stake in the outcome and are typically responsible only for providing the data you ask for, list rental of this nature is often maligned for poor performance and exorbitant cost. This may not always be the case however — in many cases in my experience, the marketer didn’t define the target particularly well.

Target definition is no small matter, especially when your mandate is to efficiently become “bigger.”

“This underscores the real challenge — when picking the criteria for prospecting, you are, either consciously or unconsciously, presuming that the criteria you are choosing, is actually predictive of buying behavior.”

Multivariate Modeled Selection, The Next Level Up in Sophistication — and Efficacy
In many cases, it may not be enough to choose variables and simply pick criteria to define your prospect. So some proven math and statistics can become the marketer’s best friend. Models are statistically derived targets. When we build a model we start with data, not opinion. We then use a marketing database like BuyerGenomics, to provide customer intelligence and understand the varying segments of buyers that exist — even if we didn’t have resolution to see them yet.

With these insights we can now develop a model that shows us the criteria or attributes of our customers that are most impactful. They often are clear enough and may sound like your expectations, though we do find nuances, which can improve the selection of prospects we will market to.

Our models can combine data points to infer new ways of looking at who our target customer is, like ethno-demographics, the relationship between ethnic attributes and income/age, or geo-density, the concentration of a customer in a specific location, for example.

This is surely a step up from “look alike models” (different from user selected “look alike criteria” in multivariate marketing).

Multivariate modeling can also require that we enhance customer data with many, many variables and then determine the variables that are most predictive of who our ideal prospect really is. As a result, we have a fairly reliable approach for testing our target definition and validating that we’re zeroed in on our target.

True Response Modeling: The Gold Standard
For marketers seeking to scale up and grow substantially, for example thousands or tens of thousands of net new customers, a true “Response Model” is the most sophisticated, and highest value in targeting a universe of most likely responders.

This should not be confused with any of the increasingly common “look alike” approaches, which are common to programmatic display advertising, univariate or multivariate targeting. It is also substantially more effective than the multivariate modeled approach we’ve already described above.

A true response model has a number of characteristics that can dramatically distinguish it, and improve your prospect marketing performance. Firstly, it’s a custom model, you can’t get a custom response model built by someone who knows nothing about your business — and it’s a different solution altogether than buying a target based on one or more variables or targeting criteria.

Response models actually use the target from, for example, a multivariate model, like we described above. This “effective target” is then marketed to, and the responders from that target are then further analyzed and refined into one of the most exacting targets, one which we can project or simulate the lift we expect over the multivariate model.

What does this all accomplish?

The RM solution gets us to a place where we can now identify the statistically validated “ideal” target customer and then back out ever larger populations to target to maximize the number of new customers acquired.

Putting it All Together
The prospect marketing approaches herein often build on one another, and the most effective growth for performance marketers often comes from building on prior learnings and successes. It all begins, of course, with having your data in good order, and having governance and hygiene process that are working.

But to change the acquisition game, getting started is often the hardest part. That generally means having a plan to move from investment, to learnings, to optimization and then to achieving the scale desired. While it’s usually not an instant gratification effort, the approaches described herein can be done expeditiously and with ever-increasing intelligence and efficacy.

So if you’re organization is “figgering on biggering” in 2016, there’s no better time to assemble the Data Athlete’s on your team, and focus on proven prospecting technologies we’ve discussed herein. When you do, your customer base will grow “bigger,” and done well, your growth will have increasing mathematical certainty.

The Cost of Perfection

Data has become the most strategic asset in modern businesses. It is now a “raw material” that any business requires to create and keep a competitive posture in its category. In order to convert this plentiful resource into business value, the data has to be refined, made easily accessible and deployed into the hands of marketers.

Data has become the most strategic asset in modern businesses. It is now a “raw material” that any business requires to create and keep a competitive posture in its category. In order to convert this plentiful resource into business value, the data has to be refined, made easily accessible and deployed into the hands of marketers.

From the perspective of working with dozens of marketing and IT organizations, it’s all too common to see this process grinding slowly along — even as opportunity costs rise for the organization.

Those opportunity costs are real — as the organization doesn’t have the visibility into its customers base and the wider market to make new strategic investments that competitors can’t even consider. Put another way, the opportunity cost is the foregone competitive advantage.

A Case Study of a Strategic Growth Opportunity Cost
A large “brick-and-click” retail brand asked us to look at its business, as it lacked a clear plan on how to achieve ever-increasing growth goals pushed down from the C-suite.

When the conversation started, it was a very tactical discussion. Lacking a strategic dimension or business context — the outcome or definition of success ultimately was to do something new and fresh.

The C-suite wanted outsized growth. So while the market they served was growing modestly, growing only with “the pie” or the market itself would not meet expectations.

Taking share was going to require a more advanced strategy.

With a large retail store footprint in North America and a robust online business, the company struggled to find credible new ways to scale sales and share further.

Shifting the dialog to the evidence of a specific business issue, and the impact that addressing that issue could or would be expected to have, helped focus the organization on the more concrete opportunities that existed.

The Symptoms
Some fundamental business and marketing “health” metrics were either inaccessible or non-existent.

Two prime examples were: the growth rate of net-new customers, and the value of existing customers.

Beyond those, the value of a customer over time, and the cost of acquiring customers from various sources relative to their upfront contribution to the top and bottom line, to the business and its value over time, were elusive.

When we asked to look at source data to get the answers for the organization, we learned that an 18-month project had been under way to clean and organize “the data” and measurement of these KPIs would have to wait.

With a rush to activity without some important benchmarks to define what must change, we agreed there were elements of a “ready, fire, aim” approach. However, the data was “inaccessible” or otherwise “unavailable.” IT owned the data.

Another case of the new “Data ‘Mine’-ing” (not to be confused with productive “Data Mining”). In the interest of “controls, clean data and process” — data is held hostage and doesn’t create business value. As marketing waits, the value of the data (recency, for one dimension) may decay — and competitive advantage diminishes.

These sounded like pragmatic purposes. After all, one would have to suppose that if it really needed to be cleaned, then it had to have been quite “dirty”  in the first place.

What possibly would the value be in working with such “dirty” data?  It makes sense on the surface to me — plus, who likes “dirty,” in general?

Emotionally, I feel and experience the same concern. This dislike of “dirtiness” has made robotic vacuums a mass-market product — iRobot alone has sold more than 14,000,000 of them — and counting!

Yet this same data that was being “organized and cleaned” for the prior year was remarkable for what it did not do … create value.

This example seems paradoxical for many. Yet there are highly rational reasons for this behavior.

Why Does This Happen? The Wrong Conceptual Model
The organization had a conceptual model that did not serve the business. Those most reasonable-sounding priorities were misplaced.

Why? because “clean” or “not clean” data is actually highly dependent on the specific purposes for that data. The same applies for almost any other description of data (and many other things) made without meaningful context.

One example is sentiment analysis. In this case, limited or even no cleaning at all could work with Bayesian methods on unstructured data, such as reviews on Yelp.

Granted, every review would not make an equal contribution to the final determination of the consumer sentiment about that brand — but does it have to? Of course not.

Your House is Dirty.
How did that statement make you feel? I wrote it, and for me it is really uncomfortable. I’m not a total neat freak, but it gives me an emotional, visceral reaction.

Why?  Because I like “clean.”

But if someone wanted to come in with a white glove, I guess that person would be able to get a smudge on it somewhere. Now maybe I could clean up the place to pass even the most stringent white-glove test …

But what if we came in with a microscope … regrettably, I’d find microscopic organisms in every home — mine, included.

Yuck.

But this being the case, we also know that our homes aren’t any less livable or enjoyable.

Let’s say we irradiated it. Like the perfect “cleaning” you may be envisioning for your data. Let’s just say we applied ultraviolet radiation to every surface. Now it’s as clean as we can get it …

Is home that is anything short of irradiated better than being homeless?

The corollary … is good even if ‘not perfect’ data any better than being data-less?

The reality — “clean” in data and elsewhere really is in the eye of the beholder.

Ever been in a college dorm or a fraternity house on a Sunday morning where no one’s complaining about how it looks? However scary it might be to you and I … it’s clean enough for them.

Yes, the same can be said for your data. It depends on how you wish to use it, and what the outcome you’re looking for is. Even the fraternity house looks perfect and smells like fresh lemons the day that the parents (and their checkbooks) come to visit.

Accuracy vs. Precision
Instead of clean vs. dirty data, marketers do well when they consider how accurate the data is for a specific purpose, vs. how much precision it could produce.

Big Data being “big,” we simply don’t need to hit the bulls-eye every single time; which is critical, because that’s not likely.

If the collection methods are logical and reasonable, even if only 90 percent right … that’s 10 fails out of 100 tries … we can still have precision for a given purpose.

This example from Jim De Novo’s “Drilling Down” makes a great example of why accuracy (AKA, “clean” data ) isn’t the only thing that matters — precision is what matters.

Mike Ferranti blog artWhen our data is inaccurate, but precise, we can use it to predict what will happen next.

In the bulls-eye on the left, we keep aiming for the perfect bulls-eye. We keep missing, however, and how much we miss by, or where the next shot will land is hard to say.

In the bulls-eye on the right, the attempts are precise. That is, they do not hit the bulls-eye consistently — but they are also consistently near the bulls-eye. We can realistically expect to know where the next dart will land.

Perfect Is the Enemy of Good
Similarly with data, we don’t need some theoretical “perfection” to be practical. When we have a large data-set (and in the digital age, they are usually sufficiently large) with some level of random error in it, we have precision, and we can predict the customer will buy more bath soap than perfume.

Better yet, one of the beautiful things about statistics (and computers) is the ability to assess, measure and account for error, outliers and still produce predictable outcomes.

In marketing, especially at scale, we’re looking to optimize performance. Rarely do we get it truly and totally perfect — not just because we’re not building medical devices to implant in a person or bridges that millions will walk across … but because a few percentage points of improvement in profit can redefine the leader in a category.

The Bottom Line
In the example we began with, we used a fairly weak proxy for the “ideal” data we couldn’t get our hands on for our analysis. With all of its limitations, we were able to discover an opportunity to grow customer value and take share from competitors with an eight-figure return … and if we had used only the data the organization already had on Day One, 18 months prior, that rate of return could be double.

Marketers need to have a bias to action, and start using the data they have today. It is far too easy to succumb to a narrative that leads us down the path of inactivity and reactivity.

Clean and perfect may sound or feel good — but the corner office and a big promotion requires action and results.

Don’t delay in the hopes of theoretical perfection that really never happens — take a shot and see what is actually feasible.

If someone, however well-intended, “scares you into inaction” over visions of some perfection, cleanliness or readiness of your raw data, perhaps progressive marketers have to start asking what — or whom it is — who’s “just not ready.”

3 Database Marketing Strategy Takeaways

An old friend in business called me to share her “crisis of confidence” in using her pricey new database/CRM/analytics system. She had led the organization through a major investment in “cleaning, compiling and organizing” the data to make it more “usable.” It was a herculean task,

An old friend in business called me to share her “crisis of confidence” in using her pricey new database/CRM/analytics system. She had led the organization through a major investment in “cleaning, compiling and organizing” the data to make it more “usable.” It was a herculean task, and she was proud of her accomplishment – but she was struggling to produce a material outcome beyond project completion.

The Business Problem
After wrestling with their data and building reports for another six months, there was a sinking feeling, one you may have even experienced yourself … for all the effort – where was this going? How is it driving the business? Are we making better decisions for it – or are our decisions just different? How will we justify the investment and produce returns? The catch all “infrastructure spending” was tossed around briefly.

After looking at a handful of reports and documents, I had many questions for her. Not surprising for an exceptionally successful executive like her, she gave me some fair and honest answers. One response she used more than once was, “We don’t know.”  That’s not trivial. In many organizations, it’s risky to think it, much less say it. Sometimes the best answer is “I don’t know” – but surrounded by data and smart people, it’s not entirely unreasonable for some folks to feel uncomfortable with a candid “I don’t know.”

Get Comfortable with ‘Not Knowing’
Just saying “I don’t know” can be the first step in solving the problem – so long as you’re wed to the fact that you “just know” you can’t even take the next step, which is problem definition, because you don’t have a problem if you “just know” something is good, important or even working – data and evidence aside. Take the challenge – I guarantee you this small act will spark more ideas, action and solutions for the de minimis time it takes than anything else you can do.

Take Away No. 1:
If you don’t know something, say so. Say it out loud, even. It will help in emotionally and logically moving on to defining the problem.

By now, I’m probably close to losing a few folks who are reading this. “I don’t know” is not something they’re comfortable with. If you are one of them remember, “I don’t know” is not where the process ends – it’s very often where the solution we’re hungry for begins to reveal itself.

Mike Ferranti blog pullquote

Problem Definition: It’s 90 Percent of the Problem
In discussing what the problem really was, we found another common issue. The problem wasn’t well-defined in the first place. The problem was essentially to “clean up the data” and to “have organization.” While that was a good thing to do, it didn’t solve any business problem. The manifestation was they now had a big (expensive) bucket of data that was judged to be better, or more valuable, than it was beforehand.

How was it better? The answer was it was more organized and more clean. Did it answer any specific questions? After looking closer, we saw it did. But did it actually begin to solve any specific problem? This was less obvious.

Here is where problem definition is so important. The problems that were defined as “organization” and “cleaning” weren’t business problems. They were symptoms of a data capture process that didn’t work, and that process came from a lack of a clear strategy.

Boiling the Ocean: Solutions That Are All Things
The specific problems being experienced were many and diverse. Focus was low. This was, in large part, a solution that was intended to do all things for all people.

I’m a direct marketer. I started my career in software development. I have a great appreciation for large systems and for what is commonly known as the “data warehouse” – a large database system that often starts with financial system data. Warehouse solutions often contain every cost in the enterprise, every operations metric, inventory, logistics, marketing, human resources and more.

But surely these are not “function-specific” solutions. In the vast majority of cases they are starting points, and they are not solutions to the problem the marketer has in selling one more widget. Those solutions need to be borne of a very specific set of marketing problems, and utilize a specific set of data – and in a specific format and data model – to actually solve them. That marketing-specific solution would likely need substantial transformation if taken from that “warehouse” solution. And when complete, it would virtually be a whole new dataset, altogether.

Take Away No. 2:
Ask yourself, are we taking a “Boil The Ocean” approach?

Boil the oceanAfter some discussion, we aligned that she had surely accomplished a lot, and that we could now access and view data about many things in the organization, including in marketing. But there were no specific capabilities that would speed the time-to-value present, and it was hard to make progress. Also, the data my friend’s organization quite reasonably thought was its most important failed to highlight the huge differences between the value of customers over the longer term. That created a strategic problem. The organization was trying to fix the long-term and its strategic business problems by looking at the wrong data and taking the wrong actions. These are very bright people with a compelling rationale for their course of action.

In the end, “Boil The Ocean” approaches are short on strategy, or are built on a strategy so grandiose, they become difficult or impossible to execute.

The Root Causes of These Strategic Challenges
So ultimately, how do such quality organizations go down an inefficient path like this? It ultimately comes down to a skills gap. What must change? It’s the skill set in marketing.

In the digital age, there are two major skill sets that we must buy, hire or develop in our organizations. Neither is trivial in marketing, and neither is possible without patience and focus.

Skill Set No. 1: Technology, Logic, Data
Marketers have traditionally come from a promotional and creative background. The big idea was always the highest-valued commodity. Today, things are changing faster – and permanently.

Marketers today are consistently spending more of their time with technologists, developers and data designers. The logical problem-solving skills by these folks are very different from those proposed by professionals with a creative or project management background. They need to solve problems that are not even being discussed on the way to solving the problems that are.

Because most organizations have some expertise with technology, and work with technology providers, the key takeaway here is that marketing data-specific applications require a different set of tech experience. Working with marketing data for marketing outcomes is unlike working with other types of data – the experience your IT department has working with finance or logistics data isn’t as useful as marketing “purpose-specific” data and technology experience.

Skill Set No. 2: Math
Barring some advanced direct marketers, marketers don’t always come from a math background. Only now are VPs beginning to have development, math and statistics experience. In an age of analytics, and now with the advent of tools and technologies to leverage large data sets, a solid understanding of math and basic statistics is becoming increasingly important.

Here’s an example to help make the point about the comfort level of using math and basic statistics to think about data:

You’re looking at the incomes and affluence of a customer base. With 1,000 members in the group, we have an average income of $100,000. That’s pretty telling, you might say.

We get more data, and the 1,001st customer is added to the sample – it’s Warren Buffett (net worth $67 billion). How useful is that average now?

There are many expressions for this common scenario – where outliers in your data can skew your numbers. From this come the expressions “The average lies” or “the tyranny of the averages.” Surely, the average isn’t the best number – though it’s a shortcut and a starting point. But it’s best to compare it to the median before taking too much faith in it – and a distribution histogram might tell an even better story about the composition of your customers’ incomes.

histogramDoing so will show if they are truly random … and follow a Gaussian distribution (AKA, normal distribution) and you can imagine this histogram with Warren Buffett added to the sample.

The takeaway here is the concepts required to evaluate and think about data require experienced and trained analysts. And those trained and experienced in evaluating marketing data are also required.

In the end in the scenario I described with my friend and client at the beginning of this column, her organization built a very large system, finished it on time and about at budget. But what the company invested in and created had some fundamental shortcomings. It was not a “purpose-specific” marketing solution – and it was conceived by a competent IT organization that was tactically adept – and strategically adrift.

Takeaway No. 3: Marketing Must Drive Marketing Outcomes
Marketing must drive marketing outcomes. Due to the discomfort that marketing often has with technology, math and statistics, key strategic decisions are quietly left to IT, a vendor or to chance.

This, of course, jeopardizes the results early. Marketing leadership can always ask good business and marketing questions and hold IT and technical resources accountable to prescribe only solutions that have a clear and simple strategy for achieving those goals.

The Bottom Line
Begin your database marketing endeavor with the “end in mind” by describing what success would look like in business terms. The decisions you’ll need to make will be easier.

If the discussions turn technical early, and they often do, ask business- and outcome-oriented questions to steer the conversation back on track. Maintain a focused strategy from the start and don’t let inexperience with math, programming or database science derail you. Get the help you need – and that means individuals and teams who know not only their discipline, but can answer the questions that matter most to you.

When you follow these strategic guideposts, your next big data-driven initiative will pay dividends now and into the future.

Marketing Machines — Possible or Pipedream?

True data-driven marketing is still “just a dream” for many marketers, rather than a reality. Under this vision, systems data-mine autonomously, and present fresh actionable insights at your desktop in the morning.

True data-driven marketing is still “just a dream” for many marketers, rather than a reality. Under this vision, systems data-mine autonomously, and present fresh actionable insights at your desktop in the morning.

For about 99 percent of marketers, this may sound too good to be true — and in all candor, it usually is.

But it is important to know and recognize that the intelligent application of mathematics and statistics, and the creation of purpose-specific algorithms, have been quietly creating value for years now. Yet the typical marketer still struggles to find enough time to get the mail out, or execute well-thought-out website marketing experiments against a control. (see “Analytics Isn’t Reporting”)

So there have never been more skeptics of the legitimate power of the intelligent application of data, even as the C-suite expectations of a data strategy that creates competitive advantage grows. Sound like your experience, industry or career? Sure it does.

But as investment continues to grind higher and competition grows, progress continues to be made.

The Amazon of Data, Is of Course, Amazon.
You may know that Amazon.com elected to release to the public some technology it uses internally in making recommendations and determining what you’d be likely to buy and when. Amazon took the same tool-set it uses and published it on Amazon Web Services. “Pretty neat” you might say …

Mike Ferranti infographic

Because we get so many questions about how Amazon does it, and how all of this actually works, we’ll break down the AWS Machine Learning and Prediction tool-set so that qualified organizations have an idea of what’s possible.

For the purposes of this article, a “qualified organization” is one that has development talent, experience with data and at least a basic working knowledge of statistical methods. Of course, experience developing models is very helpful, as well.

We call these “requirements,” because Amazon’s tools, and every tool like it (Google has a similar tool-set for the Google Cloud Platform) requires significant programming to use. They also have a learning curve for inexperienced developers and organizations that haven’t developed competencies in structuring and transforming their data to a treatment that is readily ingested and workable with these tools.

What AWS Tools Do
AWS offers a “Machine Learning” and “Prediction” tool-set. These are two related components. Machine Learning is used to ingest large amounts of data and identify patterns in that data. A typical example is extracting promotional history and responses, and utilizing it to identify what customers are most likely to respond to a marketing promotion or offer.

When Should You Use Machine Learning and Prediction?
Generally speaking, machine learning works best when a simple “logic-based” algorithm doesn’t work, or doesn’t work consistently. Simple (or even complex) logic defines a set of rules or requirements for a decision the algorithm makes to be determined. This is also called a deterministic or rule-based approach.

If there are a lot of variables, say hundreds or more — you can’t realistically develop “brute force” rules that cover every scenario that you’d need to create value. You may determine a favorite color of a buyer with a simple rule that says if the majority of their purchases are in red, then they like red. But each purchase is influenced by more than just color… there is style, season, price and category of product, material, size and discount, to name a few. As the permutations of these combinations of variables grow more complex, a simple deterministic rule-based approach can break down, and make a prediction that doesn’t work more and more of the time.

If and when business rules begin to collide with one another and discrepancies require more rules to manage these logical collisions, Machine Learning can help sort through your data in ways rule-based algorithms cannot.

“In short, you can’t realistically create or code all the permutations and business logic costeffectively.”

If your data set is very large and the diversity of variables you have is high, any “brute force” approach is destined to fail. Running through a set of rules on a sample of a few thousand cases may still work. Now what if you have millions of raw records? This can be possible even without a multi-million record customer file, given we may be looking at the colors and other attributes of items purchased during a period of years. Machine Learning can help make the task scaleable, and when you’re using Amazon’s computing power to do it, scale becomes the easy part.

Here’s An Overview of How the Prediction Process Works
So here’s an executive-level overview of how we use Machine Learning, and how it works if you build your solution on top of AWS, or Google’s developer APIs.

1. Problem Definition — Begin with The End in Mind: Here’s the step too many really don’t get right. If you’re going to venture into Machine Learning with AWS, or anywhere else, first you must define the core problems or opportunities you wish to pursue. You’ll have to do so describing that which you can observe (through your data) and an “answer” a model is expected to predict.

2. Data Preparation: Your data is going to go into a “training algorithm” where the tools will identify patterns in the data that will ultimately be used to predict the answers you’re looking for on a like dataset. Look at your data before it goes in. Be curious. Do some logical testing on it. If it is not adding up to the common sense “sniff test,” odds are very good it won’t add up later, either.

3. Transformation: Input variables and the answers you seek from models, also called the “target,” are not tidy such that they can be used to train an effective, predictive model. So you have some heavy lifting to do to get the data into new variables, “transforming” it to a more prediction-friendly input. For example, you may have a set of transactions that a customer had with your brand, but you need to summarize that into a count of transactions for that customer, and an average time between purchases. These two new fields will be more predictive and useful. A command of logic and statistics helps make these calls, as does experience.

4. Implement a Learning Algorithm: Your input variables have to be fed into an algorithm that can sort and find patterns in your data — also called a “learning algorithm.” These algorithms are specialized to help establish models (statistical relationships) and evaluate the quality of the models on data that was held out from model building.

5. Run The Model: We generate predictions against a new or holdout sample of the same format of the same source of data. You can’t run this predictive model on the same sample you used to build the model. This begins the iterative process

6. Iterate … Then Do It Again: As is any process where you’re engineering new outcomes for the first time, this process is generally iterative. It’s usually not realistic to expect a killer result on the first pass. You’ll likely massage inputs and training methods a number of times before the output starts looking good. More on what a good output looks like in a future column, though. For now, you need to know that the first product won’t likely be the final product.

The Bottom Line — Easier Still Isn’t Quite Easy for the Average Marketing Organization
While Amazon and Google may be among the easiest websites to use, and have made tremendous contributions to the proliferation of data science by providing structure and programming tools with which organizations can develop new capabilities, using Amazon AWS for Machine Language and Prediction is not for the creative marketer or even the “traditional” Web marketer.

There is also a rising category of upstarts in data-driven and database marketing apps that add intelligence to the process and can provide marketers with a significant head-start in advancing their marketing intelligence.

Data Science requires a combination of technical, mathematics/statistics and marketing/business skills. This combination is in great demand the world over, and so it’s not easy to hire top contributors to implement all of this. But for organizations with the programming bench, or external experienced business partners, tools like AWS and Google Cloud Platform can provide a substantial leap forward in using data to make superior decisions.

Remember, the outputs of the predictive process don’t have to be “right” 100 percent of the time — and they won’t be. They only need to make the numbers break in your favor enough to have a material impact on your revenue and profit now — and over time.

After all, that’s really what the data science discipline is really all about.

Getting Started With Data-Driven and Database Marketing

By now, you’re likely to be tiring of all “the talk” about data-driven database marketing media outlets and pundits who go on and on. It’s the “new, new thing” ― for most. This month’s “Data Athlete” column is intended to be the antidote to some of that. Let’s start with where I got the idea for this … from practicing marketing professionals, who were anxious to leverage data to improve performance but needed to understand it more before they could act. So I’ll get right into some basic examples, and we’ll build on it. Feel free to find them basic, boring or ― better yet ― question-enducing.

In prior columns, we’ve covered the purpose and value of the marketing database in growing profit and business performance. With that as a foundation, let’s talk about some examples. Let’s be clear; this isn’t as strategic as it is tactical. It’s for the marketer who’s considering approaches and is looking for examples so as to better understand how mountains of data become actionable and usable.

A Good Place to Start: How Many?
The first question you might be surprised to know how few marketers have an accurate answer to is “how many customers do I have?” Many marketers, especially multi-channel retailers, are focused on units sold, revenue and store attribution ― and for good reason. These are important metrics. For a brand looking to build a relationship with a customer, and maximize the Customer’s relationship with the brand –you’ll need to do more. Getting an accurate customer count is a challenge for brands that do not have a significant history of marketing directly to the end customer. If you’ve sold through distribution/wholesale or you sell through retail stores that do not have personally identifiable information (PII), it may be a challenge for your organization.

If you’ve covered this base, then you’d be surprised to know how few brands have the data in place and organized. The biggest reason for this is the data is often tied up in an “IT” database, where it’s logged by the POS, website CMS or another system.

Moreover if you examined a transaction file, you would likely see many tens of thousands of rows of transactions, many of which are anonymous given the process in which they were captured. This creates some challenges. Without PII, we can’t associate the transactions to the customer. Even when we have the customer PII, the data is virtually never “clean” enough out of the box to just match on the name ― “John,” “Jon,” “Jack,” and “J.” are not the same to a database or POS system ― but they have to be in a comprehensive marketing database.

Data processing matching programs repair, cleanse and transform transaction data from a set of “raw transaction data” into a comprehensive buying history by customer.

For now, let’s assume we’ve completed this data transformation phase and have repaired various missing fields, and solved for the typical data capture problems that we should expect. We’re on to doing something with our newly minted customer data file. First, we can get a clean customer count! We now know we have 100,000 or 10 million customers.

Now we can begin to leverage the timing of those transactions. We’ll know not just how many units we sold, but who we’ve sold them to, and what else and when those customers bought.  We can derive the timing of those purchases and begin to mine for statistical significance and opportunity. Before we go too far with leveraging timing data, let’s take a step back and first understand the breakout of new, repeat and loyal customers, too. Our first big question of course was “How many?”

To do this, we need the accurate customer count we started with, and now can use timestamps on transaction history to organize purchases by frequency.

We can see the difference between someone who purchases two times in three months and two times in a lifetime. Getting interesting yet?

But what else can we and should we know? Another imperative we’d urge you to “start with” is what percentage of the marketing database are new buyers vs. repeat and how many were new in a given period ― this helps you see if you are growing, and how fast or slow. Then move on to “how many” transactions, how many sales per customer? Per period? Calculate the average? How about a distribution of purchase frequency … oops … we can now see you have a one-time buyer problem … (or hopefully not).

Hopefully, I’ve illustrated in just a few paragraphs how much can be gleaned from transforming and organizing your raw transactions into a rather basic customer database. Let’s move on from “How many?” to … “How much?”

The Logical Next Step: How Much?
If you’re a mature and sophisticated database marketer, again, most likely, “you’ve got this.” But still, today the majority of midsize retailers and on down do not. So how much are your customers worth to your brand? (Yes, they are “priceless,” but really now, how much have they bought from your brand?).

Let’s move on to “How much do new customers spend, on average?” Two-time buyers? All buyers? How much by store, by geography? By salesperson? You can answer all these questions once you’ve worked through the data transformation we’ve discussed herein. How much were they worth by quarter (ie, how does seasonality impact customer value?)

How Much Promotion?
If your POS or e-commerce raw transaction files contain “discount applied” (most do) then we can go on to ascertain “How much promotional value did we trade for incremental sales?” How much did we use promotion to acquire new buyers? This of course requires that the raw transaction file contains promotion credits on the receipt record. Like the other data quality issues we spoke about, returns and credits can create serious challenges to making your numbers line up, but those can be handled through more advanced matching logic, as well. The trick is to take them into consideration from the start ― don’t “worry about them later.”

Put another way, we’re quantifying customer value. This is strategically very important and for good reason, because:

“That which can be measured, can be maximized.”

This brings us two the last big focuses that your database marketing solution can help you answer, “Who is the Customer?”

Completing The Picture: Who? (Is The Customer?)
Continuing to build from the bedrock we started with in a raw transaction file, we can move on to learning about who the customer is who bought, repeat purchased, established loyalty and possibly went dormant. Now we match those customer records to data about the customer and the lifestyle.

This requires more matching logic, spinning through your database and matching individual customers based on unique identifiers or combinations of fields you may already have. Even then, you or a service provider will then need to perform iterative matching to maximize your coverage of these data fields.

Some of the more valuable and important categories to focus on of data enhancement include:

  • Demographic
  • Ethnographic
  • Lifestyle
  • Financial/economic

When this data is identified and completed, we “extend” the customer record so we can now answer questions that inform messaging, creative and even product selection/assortment.

Bonus Question: How Did They Do That? (Math and Models)
The element of database or data-driven marketing that generates the most interest is utilizing statistical methods to forecast or predict behavior and customer value. These are the least accessible methods but may offer the most sustainable competitive advantage.

With a well-organized database, we can now begin working with statistical methods and running calculations called models or model scoring.

These methods answer questions like: Who is likely to try an aggressive new design on the product? Who is most likely to drink brown liquor at dinner? What customers have the highest probability of attrition?

While these methods may be considered the most “glamorous” of the data-driven marketing sciences, it is perhaps most important to realize that they are only possible or cost-effective when they are built on a robust foundation. That is, a modeler will spend 90 percent of the time and effort on getting data into a format to be able to develop sensible and useful statistical outcomes ― which drives up cost and time to value if the fundamental database design doesn’t support this use case for the data.

Fig. 1 is an example of algorithm-derived segments in a BuyerGenomics database.
Fig. 1 is an example of algorithm-derived segments in a BuyerGenomics database.

In sum, consider how many customers you have, how much they bought, who they are and how they did what they did when you execute your data-driven marketing program. Answers to these questions will enable you to communicate better with your customers and unlock the value from your existing customer database.