Insight is never in sight
Behavioral consumer data is a hugely valuable asset for any brand. It’s why every brand invests heavily in DMPs, CDPs, CRM solutions and what not. Processing behavioral data enables analysts, engineers, designers and marketers to see what consumers are doing in and around their branded experience. This knowledge often informs plans and actions to improve the brand’s consumer journey, from marketing to UEX and customer service. Unfortunately, behavioral data is absolutely insufficient for that purpose. Because an observation is not an insight.
Big data isn’t just called data for a reason
Marketers and analysts have learned to deal with seemingly insurmountable heaps of data. The data is generated by billions of micro-events (like a visit, an impression, a click, scrolling etc.) from millions of users. All of these events are recorded, sorted, tagged and stored in petabytes worth of data warehousing capacity, farmed out to stacks and rows of servers in air-conditioned facilities the size of entire football fields. That is called “big data”.
Every brand that generates big data, loves to talk big data. But anybody with an email account can easily tell how hard it is, to meaningfully process big data. A quick glance at the last few “targeted” marketing emails often illustrates how brands are awkwardly inept at using (big) data to make their experience more relevant, let alone personalized. As a passionate hobby photographer, I buy almost all my gear on Amazon, including bodies, lenses, tripods, memory cards, backpacks, gloves (with the “trigger finger”), and more. Yet, judging by their (re-)marketing execution, Amazon somehow still doesn’t seem to ‘know’ about my passion for photography.
One would expect the Amazons and Walmarts of this world would have to at least be able to make a few educated guesses about a customer’s identity after a number of transactions (e.g. sufficient to distinguish gifts from regular purchases). But we’re obviously not there, yet. Big data is hard. Heck, even small data seems hard!
Make it “as simple as possible, but no simpler.”
It’s hard, because behavioral data usually comes without explanation. A dumb algorithm might categorize my behavior as ‘Photography Enthusiast’, only if 80% of my transactions happen within the respective sub-category. The ‘DSLR Photography’ sub-category is usually categorized under ‘Audio & Video’, which in turn sits within the ‘Electronics’ vertical. But if memory cards are listed in a different sub-category, gloves in the ‘Fashion’ vertical, and the back-pack in the ‘Sports’ vertical, a dumb algorithm will fail to connect the dots between these purchases.
Large data sets (like engagement- and transaction data) still have to be simplified, even when machine learning algorithms are deployed to make the data more digestible and useful. The carbon units which design and program the ML and later analyze the output are still human, after all. And the ones making decisions based on the data analysis, are arguably even “more human”, requiring ever simpler perspectives.
But data aggregation is — by design — always obscuring aspects of consumer behavior. Various informative patterns, only visible at more granular levels, may cancel each other out in aggregation or disappear in normalization. Segmentation (which is just a more refined type of data aggregation, based on demographics, behavior, psychographics etc.) can add even more risk and ambiguity to data interpretation, depending on how the segments are defined.
Too much simplification can render data useless (hence the Einstein quote in the sub-header), sometimes without anyone noticing. While seeing WHAT people are doing at scale can be instructive, it is merely an observation. Observations are typically obvious, easy to acquire, cheap, and usually not enabling any competitive advantage. Especially not when extracted from aggregated data.
Always go from the WHAT to the WHY
I’ve once asked a (wicked smart) data scientist to help me understand what type of creative execution drives the best CTR for a critical, high-traffic onsite placement. Their answer was to do some A/B testing to figure things out. I’ve re-phrased my question to ask which specific components of a marketing execution (product offering, messaging, incentive, art direction, CTA etc.) would motivate users to engage with the asset, and to which degree. Put simply: WHY do people (not) click on banners? Their answer was to conduct a more complex, multi-pronged set of sequential of A/B tests. ¯\_(ツ)_/¯
In reality, even the most sophisticated A/B testing regime remains limited to making behavioral observations. Whichever type of statistical regression is applied to split testing results, it will not amount to producing any motivational insight. And it never actually crossed the data scientist’s mind to ask consumers directly about their (lack of) motivation to click on the banner.
Understanding WHY people do WHAT they do is somewhat laborious, because most often it requires some form of consumer research. But it helps brands avoid the capital error of masticating behavioral data through a conscious or unconscious (but mostly arbitrary) interpretation ‘process’ to arrive at a conclusion. In other words: it is invariably more legitimate to utilize the customer’s first-hand understanding of their own behavior, than to rely on a Kool-Aid-drinking marketer’s insight-free (and likely implicitly biased) hypotheses.
Contrary to observations, insights are neither obvious, nor easy or cheap to develop. But precisely because of that, they can be competitively advantageous. They’re also more fun to identify. This sounds basic, but directly connecting with consumers to get from the WHAT to the WHY is not just insightful, sometimes daunting, and generally good business practice. It’s actually a lot of fun.
Take a walk in their shoes
But what does a true insight look like? An insight illuminates a previously hidden or implicit but universal truth about consumers, a transaction occasion, or a market entry point. It elicits a cerebral ‘a-ha’ reaction, or an emotional moment of enlightenment. And in the best case, it identifies a currently unaddressed tension for the brand to resolve. Also: You’ll know one, when you see one. Here’s a hypothetical example of a fashion retailer going from the WHAT to the WHY:
The WHAT — the unenlightened observation, based on behavioral data: The hypothetical fashion retailer (“Hy-Fare”) just pulled transaction data from the past 12 months. The data is aggregated through demographic and RFM segmentations, which helps to read segment performance, and isolate target group behaviors. One of the target groups, the female suburban Gen-Z (“FSGz”), is of particular interest for Hy-Fare. While it is currently converting at low frequency and low spend, the company has high hopes to grow revenue with the target. The Hy-Fare CRM department launches a marketing campaign with frequent, targeted incentives, aimed at boosting both purchase frequency and basket size. However, while purchase frequency slightly increased, it came at the expense of basket size. What’s worse: profitability had eroded due to incentives primarily subsidizing existing behavior, and structural economics tightening under higher frequency and stable fixed costs.
The HOW — the faceted understanding, based on attitudinal information: Hy-Fare has commissioned a quantitative consumer research study to collect information on how the FSGz target sees the world, and how they think and how they feel about Hy-Fare and its branded experience. They’ve learned that FSGz are indeed enthusiastic about fashion. They constantly look for the latest trends to express themselves, and importantly: they love the thrill of hunting for the best deals. Based on this additional information, the CRM department pivoted to add inspirational content around the latest fashion industry trends, as well as offer more browse-based experiences (on top of the cash incentives). Hy-Fare saw their engagement rate tick up, but both purchase frequency and basket size remained flat. Neither customer research nor revised marketing efforts seemed to have payed out as expected.
The WHY — immersive understanding, based on true human insight: In order to more deeply understand the FSGz target group’s motivations, Hy-Fare is now investing in an ongoing consumer research panel. Direct interaction with FSGz from the panel allows Hy-Fare marketers and analysts to ‘walk in their consumers’ shoes’ (i.e. pretend to be them for a few hours) and gain a more qualitative understanding of them. The insights gleaned from immersive techniques such as ethnographies, journaling, shop-alongs, IDI’s and more, can be immediately and quantitatively validated through the broader panel. It turns out FSGzs are on a budget, but still can’t be seen wearing the same outfit too often in public. As a result, they may shop for cheap items and accessories to keep their wardrobe fresh, but also focus on selling or swapping the items they’ve worn too often. For the environmentally conscious FSGzs, the sustainability aspect of exploiting fashion items in this way is an appealing, secondary benefit. And that is WHY they currently convert at low frequency and low spend with Hy-Fare.
Based on this true human insight, Hy-Fare leadership decided to expand their offering with a used fashion category, and is now investing in buy-back programs for their customers. The Hy-Fare CRM department launched a ‘sell-to-buy’ marketing campaign that taps into fashion micro-trends, and they are now using predictive modeling to optimize messaging, incentives and timed customer contact strategies. Both transaction frequency and spend have increased as a consequence, as Hy-Fare acquired a significant share of selling- and swapping volume from the previously untapped market. Profitability remained stable, despite logistical investments into circular commerce, making the overall investment ROI positive.
…but how about the real world?
Hy-Fare isn’t real. But the customer behavior is. And a bunch of start-ups like Thred-Up, Poshmark, Farfetch, Vestiaire Collective, eBay and The Real Real are now battling for their share of a ~$24bn fashion re-sale market. The retailers who haven’t reacted meaningfully, are seeing ‘used return’ rates soaring, and without a doubt their margins suffering. But big-name brick-and-mortar retailers like Nordstrom, Gap, REI (and more) are embracing the customer insight as an opportunity — many in partnership with Thred-Up.
And that’s just one example of one insight in one industry. The fact that it remained invisible to so many incumbent players, for so long, is a bit of a shocker. Go ahead and keep making your observations on behavioral data, but don’t expect to unlock any significant growth opportunities, without putting in the work to unearth a true human insight, first.