![](https://thescienceofppc.com/wp-content/uploads/2020/08/soppc-23-wp.jpg)
There’s something magical, mystical, and entirely unrealistic about forecasting on limited data. In Michael Lewis’ classic Moneyball, Billy Beane is the data-centric hero – a new breed of baseball manager for data-intensive times.
But Billy knows when data has its limits: he regards pitchers to be like writers, as opposed to thoroughbred racehorses. Strong, hard-throwing, etc.? Nice on the surface, but not predictive of major-league success. Getting outs is a special talent; some pitchers come into their own later when they discover a new pitch, or discover the secrets of their own talent. Never say never.
What does that have to do with marketing?
I believe there are parallels. In marketing circles, you sometimes run across the beguiling notion that we might be able to predict high lifetime value customers from the very beginning, if we just had great data.
Crystal balls are in short supply, unfortunately. It would also be nice to predict which early-stage companies would become the next Uber or Amazon, too. But events unfold. There are twists and turns. Uber wasn’t even technically possible until a certain generation of iPhone made it so; similar startups had come and gone in the past, with barely a whimper.
If you’re having a great week in a retail business, sometimes you ring up the hoped-for lifetime value (LTV) from a customer – and then some! – on the first purchase. Recently, I looked at a purchase record for a consumer goods purchase in excess of $3,000 – multiple items, representative of a customer who knew what they wanted and moved quickly to achieve their goal. No dithering or worrying about the budget. The purchase was 7x the average order size for this retailer. The perfect customer! Thanks for stopping by. Y’all come round again real soon, k?
You could tell that the customer was a professional with a high income, with not a lot of time on their hands. You could also glean that their home was likely a fancier place in a rural area (perhaps an acreage) with average to below-average home values and average to below-average incomes, especially when measured by a national average. Their neighbor down the road might have been quite a different story. That neighbor might have purchased much less, interacted more on Facebook with “likes” (without ever purchasing anything), etc. Using the levers available to us as marketers, we could easily have missed the big spender if we’d been attempting too much “precision” in bidding.
In some regions, there are plenty of Millionaire Next Door types who seem unassuming until they swoop in and quietly make a big purchase. They’re just not the norm. You have to wait until they come out of the woodwork. As a Fredericton, NB cab driver once told me, after noting that his “mum back home in the UK” also had a German car like mine: “back home, these cars are everywhere.” By contrast, “people here like to hide their money.”
A little like the way Barry Zito hid his major league pitching potential behind an 87-mph fastball.
Stranger danger: divining the unpredictable user
As sorting mechanisms, our channels’ most obvious targeting methods themselves do very well. Myriad additional attributes of users and user behavior can help, but they’re unpredictable. Keywords, merchant feeds, photos of individual products, positioning via ad copy, etc. – these core drivers, which we control and to which prospects respond – really are powerful means of connecting and targeting. While we add as much granularity to bid strategy as makes sense, predictions of lifetime value (not averages, or typical patterns, but predictions based on complete strangers who aren’t yet customers) get us into complicated territory.
In retail, some of those who buy more frequently in higher quantities are what we might call semi-pros. In confections, some of you might buy a small jar of silver sprinkles for a cake. The owner of several ice cream trucks, by contrast, is buying “near wholesale” from a popular candy retailer all the time, because his customers love the sprinkles.
But back to the home furnishings & accessories realm.
The more-avid-than-average homeowner, the semi-pro, and the pro might all camouflage their value to you by spreading purchases out over many months or years. And some more complex purchases or relationship-driven bulk buys involve phone calls, chats, networking, supplier relationships, and other interactions the Smart Bidding bots have no way to take account of.
Realistically, it’s difficult for many businesses to predict “super-lifetime-value” prospects with any consistent accuracy.
With that being said, sometimes patterns hit you over the head. People buying certain luxury goods today are putting those luxury goods into expensive homes. You can literally visualize – using Google Maps – the brand new modern detached home that will house the new items. The Maps picture shows various trucks parked on site, a stunning new home built after the previous older home was torn down, and an obviously good neighborhood (if you’re plugged in, you know all about the schools, the shops, the commute times, and the home values on that street) that is becoming even more exclusive. You also know that furniture and décor styles one day look hopelessly outdated as people move in and move up in these new contexts. (That happens across the economic spectrum, although I’m giving an example of a pricier home). So this vivid scenario of “what type of person lives here, and what will they need” lends confidence to the campaigns as a whole. You can be patient, expecting that every so often someone will order a more luxurious piece without batting an eye.
I can also tell you a few things about “stage of life” as it influences customer value (from observing our known demographic data). With very rare exceptions, the ROAS for a 20-year-old looking at an expensive home item more appropriate to a 40-year old buyer will be quite low. Large, “sophisticated” orders may be more associated with people 35-64. And purchase behavior over 65 may be spiky. Older people can afford nice things, in many cases. But they may also be downsizing. They might take longer to decide. Or (until everyone over 75 is extremely comfortable buying home furnishings from their phone) they may simply prefer to transact over the phone or in person. (Note: Google Ads gives some demographic data such as probable age, but the oldest age bracket is 65+. 65-74 would probably be a useful breakdown these days, as people live longer, healthier, more adventurous lives, but it isn’t available in the interface.)
Another complication is that people buy things for their spouse, mother, daughter, boyfriend, etc., without necessarily signaling that this is the case. They may be gifts, or it might be that one person does a lot of the research for purchases, but the purchase process as a whole is more of a family or group effort. Or family members take turns researching a purchase. Toss in the fact that some people own rental properties, or Airbnb properties! And are chairs and framed art for a home, home office, or small professional office? Beware of the “fully autonomous individual” myth in purchase behavior.
At some point, then, we can outsmart ourselves by expecting absurd levels of precision in predicting who will be our best long-term customers. Sometimes, big wins and big growth surges are the opposite of subtle. I’m convinced that in many cases, human entrepreneurs have better instincts than the bots might. Sure, we might “overnarrate” (telling ourselves sweeping stories to simplify a complex reality, or picture a persona rather than allowing data to be data) and thus leap to faulty conclusions, but at other times we “know” what’s going on because we have spent our whole lives mixed up in certain economic and cultural narratives. We look at the size of that house, understand the context (the street), and get a sense that there may be more rooms to fill. 😊 Jumping on opportunities is a uniquely human knack.
Acquire customers first, build loyalty later
To some degree, future purchase behavior is contingent on how events unfold, as opposed to being predictable from Day Zero. (Remember, major league pitchers as “writers” of the story of the game?)
This unpredictability might dictate a less subtle approach to customer acquisition strategy than you might think. For the sake of simplicity, let’s give two disparate approaches nicknames: The Soothsayer Approach and The Modified Land Grab.
- The Soothsayer Approach is the more twee strategy of the two. It’s premised on being surgical with bidding predictions, based on supposedly advanced data about who is likely to buy more in the long run. We aggressively chase after those highly valuable prospects, and do whatever we can to avoid low-value buyers.
- The Modified Land Grab takes a humbler view of our powers of prediction, aiming to acquire valuable customers before the competition knows what hit them. The word “modified” here simply acknowledges that we might not be looking to acquire every prospect with a pulse and a mouse; we still go at the task with strong positioning and targeting in place to ensure reasonable and sustainable profit margins.
(Although you could undertake either strategy using manual approaches or automated approaches, for the present purposes, this distinction might persuade you to consider Target CPA as a dark horse bidding strategy when you were all but convinced that Target ROAS was your ticket to printing more money. Or you might get tired of waiting for the bots to “pounce on the opportunity,” and revert to manual approaches, particularly if you have a conviction about certain keywords, geographic areas, or demographics you want to go hard after.)
There is some sound marketing logic behind this. Unless you have a lot more data about people than you probably should have (privacy is a moving target in our industry and especially across different jurisdictions), you could make interacting and upselling existing customers a high priority. You don’t have great visibility at the beginning as to which ones will become loyal (or have more budget).
Land Grab is a bit of a misnomer, since many businesses will be looking to acquire market share while retaining a certain amount of positioning rather than indiscriminately marketing to everyone. But once that positioning is intact, acquiring a customer – any customer – may be a sounder take on one’s marketing push than to tie yourself into knots with subtleties.
IKEA probably doesn’t know which of those new customers buying a few potted plants and a single pack of silverware (total order value: $135) will turn out to buy $1,500 worth of furniture over a lifetime. They do know that they’re pretty good at converting existing customers to strong lifetime relationships, and have even had good success at convincing home renovators to use IKEA vanities and kitchen cupboards. A lot of that success is based on progressions through life stages and life cycles. Like you, I’ve been seeing IKEA ads on TV and in print publications, and hearing the radio ads, for many years. While often quite clever, the effort hasn’t been particularly targeted.
Even with all the data these organizations have at their disposal, many consumer attitudes still seem to be shaped by word of mouth. How many consumer conversations, over the years, have been about whether IKEA mattresses are good quality and good value, or just the opposite? And don’t get me started about online reviews 😊. (For the record, I’ve purchased three IKEA mattresses in the past couple of years. They’re good. One we bought over 20 years ago was just about average. But it was good value.)
Use your instincts—and your influence
The takeaway for your business might be that the most important determinant of long-term success is whether a customer was new or not, and whether or not you acquired them (hence, acquire more, aggressively, while you can – The Land Grab), as opposed to being laser-focused only discovering what secret method will unlock those high lifetime value purchasers from the get-go. Assuming solid positioning weeds out inappropriate interactions, acquire a wide variety of customers, build trust with that customer base over time, and use segmentation to drive more revenue as the relationship grows.
As an entrepreneur or experienced marketer, you’ve got rare instincts. You’ve got data and almost a sixth sense for where markets are going and what your target consumer is all about. So I want you in there – via the Campaign Experiments mechanism – seeing if your version of events can beat the bot’s. Beyond that, if you’re meticulous with targeting, address multiple channels appropriately, and are willing to be patient and “let the chips fall where they may” as to whether a given customer turns out to be a motherlode of revenue, then you might be doing about as well as you can possibly do. No crystal ball required.
There is plenty you can do to influence the course of events (email, content, social, remarketing, etc.) after folks either buy or become part of an audience that has agreed to receive your messages. Here, you can even do better than a great major-league pitcher! I have news for you: simple observation, backed by statistics, proves that the pitcher cannot steer the path of the baseball once it’s flown off the hitter’s bat. By contrast, you have a reasonable degree of influence over your existing customers’ interactions with your business over time.
Is it really realistic to predict with pinpoint accuracy every prospective customer’s story? It is there to be written – by them, with your participation.
Some others agree with the take I’ve presented here.
Kevin Ryan of Motivity Marketing started down this “follow common sense, or Big Data if it’s really, really big and you’re really sure it will help” path some time ago. Essentially, if you are a gigantic company with a team of data analysts and incredibly fecund datasets that allow you to plumb the depths of omnichannel attribution, great. If, on the other hand, you’re a small to midsize business, a fast-growing startup, or someone trying to help them outdo the competition in the here and now, quit pretending to be something you’re not; quit trying to round up degrees of data precision that are beyond your current capacities. Be granular, be efficient, but recognize that you can’t know everything there is to know with the data you’ve got at your disposal. Perhaps it’s heartening to know that our customers are too cagey to let us into their lives far enough to predict everything about them.
In a similar vein, I recently talked with a researcher inside a strategy unit at Google. Publishers like Google are aware that more stringent government privacy policies may roll back the amount of information we can use to target consumers with advertising in future. Insofar as that’s the case, we’ll still be optimizing many facets of campaigns, but the lifetime value story is one that will be written by companies in their interactions with customers: not predicted by omniscient math Ph.D’s.
Great entrepreneurs (and those around them) have always followed the scent of growth opportunities. I don’t expect that to change.
Read Part 24: If One Part of Your Ad Group is Broken, Your Tests Could Be Misguided