Latest / SaaS Business with Fexingo: Software-as-a-Service Companies, ARR, and Recurring Revenue / How SaaS Firms Use Cohort Analysis to Predict Churn
Transcript
- Lucas: If these conversations are useful for what you're building or running, consider supporting the show directly. A handful of listeners chip in monthly through buy me a coffee dot com slash fexingo, and that's literally what funds making this many of these. Luna: Yeah, it really helps keep the lights on and the research sharp. Thanks for those who already contribute. Lucas: Let's get into it. We've spent the last twenty episodes dissecting pricing models, renewal strategies, and health scores. But there is one analytical lens that I see founders consistently underutilize, even though it is arguably the most powerful tool for predicting future revenue stability. It is called cohort analysis. Most SaaS leaders look at their aggregate net dollar retention and feel safe because the number is holding steady above one hundred percent. But aggregates lie. They smooth out the cracks until the foundation gives way. Luna: Right, so if you have a massive new wave of high-value signups, it can mask the fact that your older users are quietly leaving. You only see the average, not the bleed. Lucas: Exactly. Think about it like a deck of cards. If you shuffle them all together and weigh the pile, you might think the deck is balanced. But if you pull out every card from the first suit and lay them out, you might find half of them are bent or torn. In SaaS terms, a cohort is a group of customers who started their journey during the same time period—usually the same month or quarter. When you track them horizontally over time, you stop asking 'Are we retaining customers?' and start asking 'What happens to this specific group after six months?' Luna: It shifts the focus from general satisfaction to specific lifecycle stages. Why do you think companies resist breaking their data down this way? Lucas: Because it requires admitting failure in specific buckets. It is psychologically easier to report that overall churn is down five percent year-over-year than to present a board meeting with a heatmap showing that the March cohort has lost thirty percent of its potential value by month eight. Most CFOs and CEOs prefer the former because it feels like good news. But the latter is actually useful news. It points to a root cause. Maybe the March cohort had a different onboarding experience. Maybe they were sold a feature set that wasn't ready yet. Or maybe the market shifted and their use case became obsolete. Luna: So the danger isn't just losing revenue; it's losing the diagnostic clarity needed to fix the product or sales process. Lucas: Precisely. Let me give you a concrete example from a mid-market analytics firm I spoke with recently. On paper, their annual recurring revenue was growing steadily. Their gross retention looked fine. But when they sliced their data by acquisition channel and signup date, they found something alarming. The cohort that came through their partnership channel in the second quarter of last year was churning at double the rate of direct sales leads. And it wasn't happening immediately. It was creeping in around month four. Luna: Month four is telling. That is usually past the initial honeymoon phase but before long-term loyalty sets in. What was happening at that specific point in their journey? Lucas: It turned out to be an integration issue. The partners were selling the platform as a seamless connector to legacy databases. But the actual API documentation was outdated, and the implementation team didn't have the bandwidth to support those specific legacy systems. By month four, when the partners moved on to new deals, the support burden fell entirely on the internal team. The customers felt abandoned. They tried to troubleshoot, failed, and eventually canceled. The aggregate numbers hid this because the direct sales cohort was performing well enough to offset the loss. Luna: That is a classic case where cohort analysis saves you from making bad strategic decisions. If they hadn't seen that slice, they would have kept investing heavily in partnerships. Lucas: They would have doubled down on a leaking bucket. Instead, once they saw the cohort divergence, they paused new partner agreements, fixed the documentation, and retrained the implementation squad. Within two quarters, the March cohort's retention curve flattened out and stabilized near the company average. That is the power of looking at the horizontal slice. It turns vague complaints into specific engineering tickets. Luna: But how do you actually build this without hiring a army of data scientists? A lot of smaller SaaS teams think they need a complex data warehouse to do this. Lucas: You absolutely do not. You can build a basic cohort analysis in a spreadsheet today if you have clean subscription data. The key is consistency. You need three columns: Customer ID, Signup Date, and Monthly Recurring Revenue. Once you have that, you create a pivot table where the rows are signup months and the columns are months since signup. Then you sum the revenue for each cell. It is that simple. The magic isn't in the tool; it is in the discipline of updating it monthly and challenging the assumptions behind the numbers. Luna: So it is less about advanced math and more about consistent hygiene. What is the most common mistake people make when they first try this? Lucas: They confuse cohort size with cohort significance. If you look at the last sixty days of signups, your cohort is tiny. The volatility will be insane. One big enterprise deal closing in a small cohort can skew the entire percentage. I always tell founders to wait until they have at least fifty active seats in a cohort before drawing hard conclusions. Otherwise, you are reacting to noise, not signal. Patience is a metric in itself here. Luna: That makes sense. Small sample sizes are dangerous. So once you have a stable cohort, what should you be looking for specifically? Lucas: Look for the inflection point. Identify the exact month where the curve starts to dip below the baseline. For a B2B SaaS product, that is usually between month three and month nine. If the dip happens at month three, it is almost always an onboarding or activation problem. If it happens at month seven, it is likely a value realization or competitive displacement issue. The timing tells you which department owns the problem. Product, Sales, or Customer Success. Luna: It effectively assigns accountability based on data rather than politics. That is a huge shift in culture for many organizations. Lucas: It removes the guesswork. When Customer Success says 'our clients love us,' you can ask 'what does the month-five retention look like for the Q1 cohort?' If the data says otherwise, you have a constructive conversation instead of a defensive argument. It aligns the whole organization around a shared reality. Luna: I imagine this also changes how you evaluate new features. Do you use cohort data to decide what to build next? Lucas: Constantly. We often assume that the features driving expansion revenue are the ones to prioritize. But sometimes, the features that prevent churn are more valuable to the bottom line in the short term. If you see that customers who adopt Feature X have forty percent lower churn in their first year, you don't just celebrate that number. You make Feature X mandatory in the onboarding flow. You drive adoption aggressively. You protect the base before you chase the new logo. Luna: Protecting the base is cheaper than acquiring new business, obviously. But linking specific product actions to retention outcomes is still rare. Lucas: It is rare because it requires cross-functional collaboration. Product needs to talk to Sales, and Sales needs to talk to Finance. Cohort analysis forces those silos to break down because the data doesn't care about your org chart. It just shows you what happened to Group A versus Group B. It is brutally objective. Luna: Do you see any risks in becoming too obsessed with the cohorts? Like, could it paralyze decision-making? Lucas: Only if you treat every fluctuation as a crisis. You have to remember that markets move. Seasonality exists. A dip in one cohort might just mean that segment buys software in Q3, not Q4. You need context. But generally, the risk of ignoring cohorts is far greater than the risk of over-analyzing them. You cannot fix what you cannot see. And aggregates are blinders. Luna: So for the listener who wants to start this week, what is the very first step? Just exporting their billing data? Lucas: Start with your CRM and your billing system. Export the list of all active customers, their signup dates, and their current monthly spend. Clean it up. Remove test accounts. Then build that pivot table. Spend ten minutes looking at the first twelve months of your oldest cohort. Compare it to the newest cohort that has twelve months of data. Look for the delta. That difference is your growth engine—or your leak. Luna: Simple, actionable, and immediate. It turns abstract fear into concrete data. I think that is the best advice we could give. Lucas: Agreed. It is not about having the perfect dashboard. It is about having the right question. Who are my customers, and what happens to them over time? Answer that, and you answer almost everything else in the business. Luna: We will leave it there for now. Keep slicing that data, and let us know what trends you find. Lucas: Thanks for listening. See you next time.