Home

An Utterly Fictional Case Study in How to Assess Customer Purchase Behavior

Cohort analysis: important, trendy (to the extent data analytics can be trendy), and fun to say aloud. Done right, cohort analysis provides actionable, constructive insight. Done not so right, cohort analysis leads to well-intentioned, misguided, and ultimately wasteful marketing spend.

There are many helpful primers on cohort analysis, so I won’t rehash the basics. But for a straightforward Stats 101 definition, consider cohort analysis the work you do after segmenting a larger population into chunks…chunks with some behavioral commonality. (“These are the customers who happen to purchase during the month of April.”)

Cohort analysis is a particularly powerful (and essential) tool for SaaS marketers. It’s practical, constructive, and dare I say easy to analyze segments when you have just a single channel for customer engagement…like say, a website. No slight intended: Cohort analysis lets SaaS marketers derive insight directly from web-activity user behavioral analytics. (For an overview of website traffic behavioral analysis, read Avanish Kaushik’s excellent data blog “Occam’s Razor,” in particular this post on hit- vs. session-metric analytics.)

But what do you do when web is one of multiple sales channels? Cohort analysis can help, but only if it’s applied to the wider context of a customer journey. Cohort analysis without context is, well, guessing. You have to review a lot of data to get to insight. The real questions to ask: What is the right data? How should I segment cohorts? What can be learned from each chunk? And perhaps most importantly, now that I know this, what do I do?

Meet Marla

Let’s look at an example case study. An anonymized case study of a company to which I once consulted. Company B has a new marketing lead. Let’s call her “Marla.” Company B sells three products: Widget Junior, priced at $100; Widget Max, priced at $500; and Widget Suite, priced at $5,000. The products are sold with renewable three-year licenses.

Traditionally, Company B marketing has promoted individual products with individual campaigns. Sales are flat: Widget Junior and Widget Max sales are growing slightly, Widget Suite sales are tapering.

Marla inherited a healthy budget, as well as execution responsibility for several previously-planned marketing campaigns. But was it time to rethink product strategy? Where should Company B prioritize its marketing spend?  Fortunately for the sake of Marla’s data analytics, Company B tracked customer behavior with a marketing-automation solution, one that was fully integrated with the firm’s CRM, e-commerce, and financial-reporting systems. (Okay, so in real life it wasn’t, but suspend a little disbelief for this blog post.)

Where’s the money coming from now? (Hint: Start with the triangle.)

Marla sought to better understand Company B’s customer purchase behavior. Using CRM and accounting data, she segmented Company B customers by revenue over three years. She classified small, medium, and big spenders into cohorts as Minnows, Trout, and Whales, and stacked the cohorts in a traditional customer pyramid:

Toph Whitmore Customer Pyramid - Whales, Trout, Minnows

She treemapped the cohorts to gauge relative number and relative revenue as a percentage of total over that three-year span:

Toph Whitmore Tree Maps - Whales, Trout, Minnows

Minnows dwarfed Trout and Whales in number, but the Whales cohort kept Company B financially afloat.

Where’s the money coming from tomorrow?

Industry analyst estimates for the size of the future Widget market were all over the place, but in general, the research community was bullish on the space. Using CRM data, Marla pie-charted revenue by industry vertical by segment:

Toph Whitmore Pie Sectors - Whales, Trout, Minnows

Then she measured three-year-historical relative revenue by product by segment:

Toph Whitmore Pie Products - Whales, Trout, Minnows

Marla looked at her cohorts by “first touch” engagement:

Toph Whitmore Pie First Touch - Whales, Trout, Minnows

The Whales were a high-touch cohort, and a diversified group. Trout and Minnows on the other hand, came from just a few industry sectors. Minnows found Company B via online search, while Whales tended to engage with Company B after an in-person meeting. But what did that all mean?

Deconstructing the sales cycle (and then putting it back together)

Next, Marla evaluated the customer purchase journeys, effectively turned her funnel sideways, and came up with a timeline to measure relative “time-to-close” for each cohort:

Toph Whitmore Engagement Timelines - Whales, Trout, Minnows

Unsurprisingly, Whale deals took the longest to close. But most Company B sales work was ad hoc, with no formal engagement model, no cadenced touch strategy, only individual, product-specific sales reps selling the way to which they were accustomed…Each sales win was a one-off deal.

Historical sales data illuminated some commonalities—Closed deals (those that moved customers through the funnel from target to suspect to prospect to purchaser) tended to share common touchpoints, whether in-person, phone, or email pings. Marla mapped what she considered to be idealized Company B touch cadences by sales cycle by cohort:

Toph Whitmore Engagement Timeline Minnows - Whales, Trout, Minnows

The short Minnow sales cycle typically starts with a web inquiry. That’s followed by two-to-four email follow-ups sent personally from the Widget Junior sales rep. If the prospect doesn’t purchase the product after a couple of emails (or after a free trial), the sales rep (or trial-eval customer support) follows up with one or more emails or a call.

Toph Whitmore Engagement Timeline Trout - Whales, Trout, Minnows

Trout–buyers almost exclusively of the Widget Max–purchase after multiple-touch engagement with a sales rep and/or the trial-eval customer support team. That engagement includes three-plus personal emails, and five-to-ten phone/conference calls.

Toph Whitmore Engagement Timeline Whales - Whales, Trout, Minnows

The Whale sales cycle–longest of the three–is more difficult to generalize, with each individual deal progressing at its own pace. The Whale deals tend to require extensive sales engagement, with sales rep making in-person on-site visits, and frequent conference calls.

So what has Marla learned? Customer purchase behavior differs significantly by amount purchased. Though Minnows and Trout businesses are growing, that growth is fueled only by a few specific verticals. With those verticals now saturated with Company B product, the potential for future growth seem limited. Whales generate revenue, but are sales-to-Whales margins sufficient to carry the company? Each customer cohort favors one product over others—Only the Whales purchase all types of product in significant amounts (but even the Whales leaned heavily toward the high-priced Widget Suite).

CAC and LTV: Time for the Heavy Lifting

What was the net present value of each type of customer relationship, and what did it cost to establish each customer relationship? Marla sought to determine the average long-term value (LTV) of a customer within each cohort, and then determine the cost to acquire each of those customers (Customer-acquisition Cost, or CAC).

Marla started with CAC. Company B had recently employed activity-based costing models, and used them to measure product-development efficiency. To estimate CAC, Marla pulled historical CRM data, then attached cost estimates to each touchpoint. After plenty of back-and-forth with finance, she added in amortized labor, COGS estimates, and marketing costs.

LTV was easier—Marla assigned a simple discount rate, NPVed the lifetime revenue estimates, averaged by number of customers in each cohort, and backed everything into today’s dollars. She found herself staring at a chart like this:

Toph Whitmore LTV CAC table - Whales, Trout, Minnows

Marla knew that channel-sales LTV should ideally exceed CAC by at least a factor of three (and preferably more like five). Only the Whales came close.

“Now that I know this, what do I do?”

Marla’s LTV analysis delivered good news and bad news. The bad news: Margins—especially at the low end—weren’t great. The good news is that all three sales cohorts were technically profitable, and Marla could see opportunities to grow revenue and reduce costs.

Whales were the most diversified business, with only small numbers of customers in multiple market segments. The potential market for Whales was big. Trout and Minnows customers came from  a narrow breadth of verticals. Unless she could broaden sales into more industries, Trout and Minnows revenues were vulnerable to sector-specific downturns.

Based on what she’d learned, Marla initiated four new programs to achieve four new objectives:

  1. Reduce cost of sales to Minnows. Minnows occasionally grew to Whales, so it didn’t make sense to eliminate the low end of the market. But pursuing Minnows manually made CAC nearly exceed LTV. So Marla automated all responses and communication with Minnows to lower CAC considerably. The Widget Junior sales rep was retrained, and began selling both Widget Junior and Widget Max to Trout. Impact: Minnows revenue dipped, but Trout sales grew.
  2. Go to where the Whales are. There is great opportunity to capture more Whales, and given the revenue dependence upon the Whale cohort, it’s all about the Whales. With C-level buy-in, Marla hired multiple Widget evangelists with sector-specific knowledge, pairing each with a senior business development exec. And then she put them on the road, sending them to speak at industry-specific tradeshows, and to network at partner events. The face-to-face outreach strategy brought results: more-robust channel leadgen, and 30% growth in Whale revenue year one.
  3. Shift marketing campaigns from product- to solution-based messages. Marla initiated marketing and event-based campaigns to target other verticals. (She actually reduced campaign spend, and saw mixed results: Trout began to arrive from new sectors, but Minnows sales stagnated, and didn’t diversify to the same extent.)
  4. Introduce disciplined, cadenced customer-journey engagement maps. This was more a sales initiative than marketing, but Marla worked with her sales colleagues to define—and then implement—best-practice customer engagement journeys, with cadenced touches (some email, some phone, some promotion) targeting specific customer cohorts.

What next?

So where should Marla go from here? Here’s what I proposed:

  • Evaluate buying frequency and churn. Reducing customer churn can preserve and lift revenue. Tiny improvements can make a big impact on the bottom line.
  • Cohort the cohorts. Company B is targeting verticals with its Whale business development. The marketing team will need to better understand each specific vertical—both demographically and behaviorally—to win each industry segment.
  • Improve engagement. Again, more a sales effort, but Company B must implement best practices to ensure each of its cadenced touches (be they phone, email, face to face, or promotional outbound) is achieving its objectives.
  • Measure, then iterate. Then lather, rinse, and repeat. Marla’s initiatives will only work to the extent that she can continue to make them better.

As a marketer, can you ever know your customers “well enough?” A sales rep would say yes, you know, since the sale just closed. A behavioral marketing scientist would beg to differ: There’s always the possibility to sell more.

Effective cohort analysis paints a vivid customer picture, but perhaps its most important impact is the promise of continuous improvement. Companies that employ measurable cohort analysis will understand their customers. And as those companies collect more data, their analytics will get more accurate. And that’s a marketing operational model that will grow with the company.

Advertisements