Let me start this post with a question. If you look at the chart below, what do you see?

What you see, I assume, is the 12-month chart of a SaaS company that isn’t doing too well. Our fictional SaaS company, let’s call it Acme, Inc., hasn’t grown its customer count, and while MRR has increased, the growth rate of ~40% over a one year period isn’t particularly exciting. What you’ll also notice is that the 40% MRR growth must be entirely owed to a (remarkable) increase in the company’s ARPA.

What I bet you can’t see in this chart is that Acme is serving two different segments, let’s call them “rabbits” and “deer” ⁽¹⁾. The customer and MRR numbers shown in the chart above are the aggregates of the numbers of these two segments. Let’s look at these two segments side by side:

What becomes visible now is that the company has historically been selling mostly to rabbits, but that part of the business is on the decline. The company’s emerging deer segment has been growing very fast, though. In the last 12 months, Acme almost tripled its deer customer base, increasing deer MRR from $30,000 to $89,000⁽²⁾. If the company continues to grow its deer segment at this pace, it will soon generate the vast majority of its revenue from deer customers, and the rabbit segment will soon become an almost irrelevant legacy business.

If you look at the aggregate numbers only, the very promising development of the deer segment and the potential that it holds for the future of the entire company get almost completely masked by the underperforming rabbit segment. If you fail to segment your data you may, therefore, end up underselling yourself to investors. Even more importantly, looking at aggregate numbers or averages across your entire customer base may lead to poor business decisions, e.g. when it comes to how you allocate marketing dollars or the precious time of your AEs. As you might know, I firmly believe that in order to build a $100M ARR SaaS company, you need to know which animals you’re hunting. Without proper segmentation, you’re most likely hunting in the dark.

Tap water, one of those things that don’t get better with segmentation.

If you’re not convinced yet, let’s turn to some real animals and imagine a zoo director who tells you that his three rabbits and two elephants eat a total of 500 kilograms of vegetables per day. 100 kilograms per animal per day! 😉 If you need even more proof that averages can be misleading, let me just tell you that on average, Pawel and I have a normal height.

The math behind these little phenomenons is, of course, trivial. And yet, I often talk to companies that I think could do a better job of breaking down their data into segments. Why is that? There are (at least) three reasons:

It’s early, your customer count is small, and maybe your price points are all over the place because you’re experimenting with pricing all the time. You might have a couple of customers paying you $100 a month, a few at $250, one at $500, a couple at $800, some at $1000 and one paying you $1100 a month. In a situation like this, it’s not clear how you should segment your customer base.
You know how you’d like to segment your customer base in theory, but you’re struggling to get clean metrics broken down by segment. In this case, you should immediately check out ChartMogul’s segmentation features.⁽³⁾
Sometimes founders aren’t sure if it’s OK to show segmented data e.g. in a fundraising deck or an investor update, i.e. they’re not sure if it’s OK to cherry-pick. My answer is always: It’s completely fine to show partial data, as long as you are 100% transparent about what you included and what you excluded and as long as you can explain why you chose to show a particular segment.

To drill down on the last point a little, consider, for example, an early-stage company that has 50 highly active trial users and a much larger number of trial users who signed up but never used the product for real. There can be various reasons for the high initial drop-off, and the company should try to understand and address them. However, by looking only at the average activity of all users, which is very low in this scenario, you might not notice that there is a significant number of people who love the product, which at the company’s stage might be more important than a well-oiled conversion funnel.

Let’s turn to a chart that shows the development of Acme’s churn rate and growth rate:

As you can see, Acme’s annual growth rate (or growth factor, I should say) is hovering at around 1.4x and churn is at or a little higher than 2%.

But if we look at these numbers separately for deer and rabbits …

…. we notice that the rabbit business is deteriorating, as evidenced by a high (~3%) and steadily rising churn rate and a slowdown in growth, which wasn’t visible in the aggregate chart. The deer segment doesn’t look great on this chart either but it has a significantly lower churn rate (1–1.5%).

Here’s another one, showing Acme’s sales & marketing spend and CAC payback time:

You can see that Acme has been increasing its sales and marketing spend month after month and has managed to keep its CAC payback time in a narrow corridor of 10–12 months. At this point, you probably know what comes next, so here goes:

Breaking down the sales and marketing KPIs by segment reveals that CAC payback time has more than doubled in the rabbit segment, while it keeps decreasing for deer! If you were the CEO of Acme and you didn’t have this information, you would keep wasting money on acquiring rabbits — money that would most likely be much better spent acquiring more deer.

Let’s turn to another fictional SaaS company, BCME, Inc:

The take-away from this chart is that BCME has more than tripled its customer base over the last 12 months, but the company’s ARPA has plummeted from more than $100 to less than $70.

You guessed right, it looks different if you peel back the onion:

BCME has almost quadrupled its rabbit customer base, while the ARPA in this segment went down from $50 to $40. In its deer customer segment, on the other hand, BCME has increased its customer count from 40 to 71 while at the same time increasing ARPA from $1000 to about $1130.

If we zero in on the development of the company’s ASP (Average Selling Price, ARPA for new customers), the different trajectories of the two segments are even more pronounced. You can’t see this in the charts, but if you look at the source data you’ll see that rabbit ASP fell from $50 to $28 while deer ASP increased sharply from $1000 to over $1500. The declining rabbit ASP and the growing deer ASP could be caused by a variety of factors such as changes in pricing, marketing mix, or competitive landscape, to name just a few. What’s clear is that a significant change is taking place in each of these segments, a development that deserves attention and that would be overlooked by measuring ARPA/ASP across the entire customer base.

Before I’ll let you go I’d like to show you this chart of yet another fictional SaaS company, CCME, Inc:

It looks like in the last 12 months, CCME’s churn rate dropped from about 4% to a bit more than 2.5% per month, an impressive achievement. (If you, too, want to decrease your churn rate by almost 50%, check out Brightback.⁽³⁾)

But lo and behold, a few months later it appeared that the company’s churn rate suddenly jumped to more than 3.5% and never returned to the ~2.5% low:

What happened? In month 3, CCME introduced an annual plan which was selected by 50% of the company’s new customers. Those customers obviously couldn’t cancel in month 4 (nor in the following 11 months), so the number of cancelations went down a little bit in month 4. Because the company kept the annual plan offering (which continued to be picked by 50% of new customers), the percentage of customers who couldn’t cancel increased with each month. As a result, CCME’s churn rate appeared to decrease month after month — until month 15, when the first annual plan cohort reached the end of its contract duration.

Once again, looking at the segmented data makes things much clearer:

As you can see now, the churn rate of the monthly plan segment didn’t change. For the segment of annual plan customers, churn cannot be measured until 15 months (which makes it all the more important to monitor customer health, which avoids bad surprises later on and gives you a chance to address churn before it happens). And even then, the simple (number of cancelations in month n) / (number of customers at the beginning of month n) formula (which is used for the chart above) can be highly misleading if the number of customers in the annual plan segment is growing fast. That’s why the best way to understand churn is a cohort analysis.

PS: There’s an expression in German that says “Man kann nicht alles über einen Kamm scheren”, which means that you can’t lump everything together. The word-by-word translation of this saying would be “you can’t shave everything over one comb”. I know that doesn’t make any sense for English ears, but when it comes to segments in SaaS, you really can’t shave everything over one comb.

⁽¹⁾ If you’re not familiar with this nomenclature, check out this post.

⁽²⁾ If you’d like to verify that I didn’t cheat and that all three charts are indeed from one and the same dataset, you can check out the data behind the charts here.

⁽³⁾ You probably know this already, but here’s the official disclosure: ChartMogul and Brightback are Point Nine portfolio companies. (Acme, BCME and CCME are not.)

The importance of customer segmentation in…