How public SaaS companies report churn, and what you can learn from them

While doing some research for another post I just stumbled on this excellent overview from Pacific Crest on the churn rates of publicly listed SaaS companies. I’ve seen posts with churn benchmarks of public SaaS companies before, but this one is by far the most comprehensive collection I’ve seen and I think it’s very useful.

What’s maybe even more interesting than taking a look at the numbers themselves is to see how different companies define churn (or the inverse, retention). Since there is no official US-GAAP definition of churn or retention, different companies use different ways to measure and report these metrics. And because public companies are under the scrutiny by the SEC, any non-GAAP metric they report must be accompanied by a razor-sharp definition.

Most public SaaS companies report churn in the form of their dollar-based net retention rate, i.e. the inverse of net MRR/ARR churn (as opposed to account/logo churn), which compares the recurring revenue from a set of customers across comparable periods. Here’s a particularly nice description of this metric, coming from AppDynamics:

“To calculate our dollar-based net retention rate for a particular trailing 12-month period, we first establish the recurring contract value for the previous trailing 12-month period. This effectively represents recurring dollars that we should expect in the current trailing 12-month period from the cohort of customers from the previous trailing 12-month period without any expansion or contraction. We subsequently measure the recurring contract value in the current trailing 12-month period from the cohort of customers from the previous trailing 12-month period. Dollar-based net retention rate is then calculated by dividing the aggregate recurring contract value in the current trailing 12-month period by the previous trailing 12-month period.”

If you take a look at the data assembled by Pacific Crest you’ll see that many companies use the same logic with minor variations. For example, some companies look at the trailing 12 month period, while others look at calendar years, quarters, or months.

Some companies exclude customers that do not meet certain criteria, for example:

• Box includes only customers with $5k+ ACV and annual contracts
• Alteryx considers only customers which have been paying customers for at least one quarter.
• AppDynamics includes only customers who have been paying customers for at least one year.
• Zendesk excludes customers on the starter plan.

This makes perfect sense: It tells you what type of customer the company is focused on, and you can see the retention metrics in regards to this type of customer.

Other companies use variations that I think are questionable. Some companies report customer count-based retention, which I think is much less interesting than dollar-based retention. Some report renewal based on the number of seats; one company, Fleetmatics, reports churn based on the number of vehicles under subscription. But the majority of companies does report dollar-based net retention rate in a way that allows for an apples-to-apples comparison across companies.

What can you learn from this?

(1) There is not one perfect definition of churn that is right for every SaaS company. Depending on the specifics of your business you might want to:

• focus on monthly, quarterly or annual retention
• exclude customers that churned within the first, say, two months
• include only customers that represent the core of your business, e.g. customers above a certain ACV

(2) Having said that, dollar-based net retention is the way to go. You should stay close to the definition above and tweak it with care.

(3) There may not be one perfect way to define and measure churn, but there sure are lots of ways to get it wrong. :) One classic example is to calculate a monthly churn rate and to mix in annual plans with monthly plans. By including customers on annual plans who aren’t up for renewal in the period you’re measuring you’re underestimating your true churn rate.

(4) Whatever metric you choose, make sure that you use it consistently and that you have a razor-sharp definition.

Bonus tip: Whenever you report numbers, be it in monthly updates or in a Board deck, include footnotes or an appendix with definitions of every metric that you’re reporting. I can almost guarantee you that this will save you ten minutes of discussion with your VC Board member(s) who (understandably) want to make sure that they understand the numbers you’re showing them. :)

Update / September 17, 2019: Another bonus tip, we recently invested in a company called Brightback that helps you reduce churn by making it easy to implement sophisticated, personalized "churn deflection" pages and workflows. Have a look! :)

Why (most) SaaS startups should aim for negative MRR churn

If you've followed my blog for a while, you know that I have a bit of an obsession with churn. Having significant account churn doesn't necessarily have to be a big problem and can't be avoided completely anyway. MRR churn sucks the blood out of your business though. That's why I think that SaaS companies should work very hard to get MRR churn down, as close to zero as possible, or even better achieve negative MRR churn.

Before I continue, here's a quick refresher on the terms that I'm using. If you're a SaaS metrics pro you can skip the next two paragraphs.

Your account churn rate, also called "customer churn rate" or "logo churn rate", measures the rate at which your customers are canceling their subscriptions. If you have, say, 1,000 customers on February 1st and by the end of the month 30 of them have canceled, your account churn rate is 3% p.m. in Feburary. Note that this assumes that all 1,000 customers are on monthly plans and can cancel that month – if some of your customers are on annual plans, you need to calculate the churn rate of that customer segment separately.

Your MRR churn rate, sometimes also referred to as "dollar churn rate", is the rate at which you are losing MRR through downgrades and cancelations. If you have, for example, $100,000 in MRR on February 1st, and by February 28 you've lost $4,000 of these $100,000 due to downgrades and cancelations, your gross MRR churn rate is 4% in February. Assuming you have $6,000 in expansion MRR in the same month – i.e. an increase in MRR of existing customers, e.g. due to upgrades to more expensive plans or additions of seats – your net MRR churn is minus $2,000 and your net MRR churn rate is minus 2% in that month. For more details on these and other SaaS metrics, check out ChartMogul's SaaS Metrics Cheat Sheet.

Thanks for your attention, SaaS metrics newbies, and welcome back pros. The following two charts show the disastrous effect of MRR churn, using an imaginary SaaS startup (let's call it Zombie.com) with $100,000 in MRR that has a net MRR churn rate of 3% p.m. and is adding $10,000 in MRR from new customers each month:

MRR development of Zombie.com - click for a larger version

MRR development of Zombie.com - click for a larger version

The first chart shows how much new MRR from new customers Zombie.com is adding (light green), how much MRR it's losing due to churn (red) and what the net change is (dark green). The second chart shows the resulting MRR (blue) and the ratio between new and lost MRR (orange), inspired by Mamoon Hamid's great "Quick ratio" of (Added MRR / Lost MRR), which I recently learned about.

As you can see in these two charts, not only does the net new MRR of Zombie.com go down every month. It actually asymptotes to zero, which means that the company is hitting a wall at around $350,000, at which it stops growing.

The math behind this is of course trivial, since the assumption was that the company is adding a constant dollar amount of MRR every month, while churn MRR, being a constant percentage of total MRR, is growing. So what happens if instead of acquiring new customers linearly, you manage to add new MRR from new customers at an ever increasing rate?

Here's another imaginary SaaS startup, let's call this one Treadmill.io. Like Zombie.com, Treadmill.io has $100,000 in MRR in the beginning of the timeframe that I'm looking at and has a net MRR churn rate of 3% p.m. Unlike Zombie.com, Treadmill.io is adding new MRR from customers at an accelerating rate, though: In the first year it's adding $10,000 per month, in the second year $15,000 per month, then $20,000 per month, and so on. Let's look at the charts for Treadmill.io:

MRR development of Treadmill.io - click for a larger version

MRR development of Treadmill.io - click for a larger version

The MRR development of this company looks much less depressing, and after ten years it reaches close to $1.5M in MRR. However, as you can see in the first chart, as well as in the declining orange line in the second chart, churn is eating up an ever increasing part of the new MRR coming in from new customers. If Treadmill.io doesn't manage to decrease churn, it will have to acquire more and more new customers just to offset churn, and keeping net new MRR growth up might become increasingly difficult.

OK, but what if you're acquiring new customers at an exponential growth rate? Let's look at a third imaginary company called Weed, Inc. Like Zombie.com and Treadmill.io, Weed starts with $100,000 in MRR and has a net MRR churn rate of 3% p.m. The big difference is that Weed is adding new MRR from new customers at an exponential rate. Starting with $10,000 in the first month, the company is growing new MRR from new customers 10% m/m in the first year; 8% m/m in the second year; 6% m/m in year three; 4%, 3% and 2% in year four, five and six, respectively; and 1.5% from year seven onwards. 

Here are the charts for Weed, Inc:

MRR development of Weed, Inc. – click for a larger version

MRR development of Weed, Inc. - click for a larger version

Not much to complain about: After ten years, Weed, Inc. has more than $19M in MRR. The big question, though, is if a development like this is realistic. In order to offset ever increasing churn amounts, Weed needs to acquire new MRR from new customers at an extremely ambitious pace. In the last month of the ten year model that I'm looking at, Weed adds about $870,000 in new MRR from new customers, almost 5% of the company's total MRR at the beginning of that month. To acquire so many new customers, Weed needs either a viral product (very rare in B2B SaaS) or extremely scalable lead acquisition channels.

I'm not saying that it's impossible, but I believe the much more likely path to a SaaS unicorn is by getting MRR churn to zero or below – which means you have to make your product more and more valuable for your customers and acquire larger and larger customers over time.

Update / September 17, 2019: Bonus tip, we recently invested in a company called Brightback that helps you reduce churn by making it easy to implement sophisticated, personalized "churn deflection" pages and workflows. Have a look! :)

When deers morph into elephants, SaaS nirvana is nigh

By now you’re probably sick of my infamous animal analogies. Sorry. But I just love them and want to resort to them one more time. :) Namely, what I want to talk about are deers that can morph into elephants, or more generally, smaller animals that can morph into bigger animals. ⁽¹⁾ In other words, I want to talk about account expansions, which are the result of a successful “land and expand” strategy.

The premise of this strategy is that it’s usually easier to get a minor commitment from a customer first and then work your way up towards a larger ACV, rather than trying to get a large deal from the get-go. There are different ways how SaaS companies have successfully employed land-and-expand strategies:
 

• Yammer is a classic example. Typically a small team in a company starts to use Yammer for internal communication. Then they add more and more people, usage might spills over to other teams or departments, and eventually Yammer’s sales team can come in and upsell the customer to an enterprise account. It’s hard to imagine a hotter, more qualified lead than a company where dozens or hundreds of people are using your product already!
• Dropbox is similar, but the difference is that you can start using Dropbox even as single user. Plus, they have another great growth vector, since people keep adding more and more files to their file storage.
• EchoSign: In this Quora post, EchoSign founder Jason M. Lemkin (one of the top SaaS experts and our co-investor in Algolia and Front) describes how EchoSign grew many departmental deployments into large, six-figure accounts over time (he also gives you the caveats).

Another way to get bigger and bigger accounts over time is of course to target startups and grow with your customers. Zendesk is extremely successful at employing land-and-expand strategies, but the company has also been fortunate enough to acquire customers such as Twitter, Uber and many others when they were still pretty small.

If your land-and-expand strategy works so well that your account expansions offset churn, then your MRR churn rate becomes negative – a state which I’ve previously described as the holy grail of SaaS. It’s hard to overstate how transformative this can be to a SaaS company. Think about it: Negative MRR churn means that even if you’re not growing, you’re still growing. More precisely, even if you stopped acquiring new customers tomorrow your recurring revenue would still continue to grow.

It’s no surprise that SaaS investors start to salivate when they see SaaS companies with negative MRR churn. Just a few days ago, Tomasz Tunguz of Redpoint highlighted that New Relic, which has filed to go public, has a negative MRR churn rate of about 14% per year. Especially for later-stage public SaaS companies, revenue churn is one of the most important metrics to look at. You cannot understand a company like Box, which is spending seemingly crazy amounts of money on customer acquisition, without understanding this metric. ⁽²⁾

---

⁽¹⁾ If you have no idea what I'm talking about, please read this post.
⁽²⁾ And yet, I have the impression that this metric hasn’t fully arrived in the world of financial analysts and accountants yet. There doesn’t yet seem to be a standard way of reporting it – every company defines the metric a little different, and some aren’t reporting it at all.

Excel template for cohort analyses in SaaS

[Note: This post first appeared as a guest post on Andrew Chen's blog. Andrew is a writer and entrepreneur and has written a large number of must-read essays on topics such as viral marketing, growth hacking and monetization. He was kind enough to publish my post on his blog, and I am republishing it here.]

If you’re a long-time reader of my blog (or if you know me personally) you’ll know that cohort analyses are one of my favorite tools for getting a deeper understanding of a product’s usage. Cohort analyses are also essential if you operate a SaaS business and want to know how you’re doing in terms of churn, customer lifetime and customer lifetime value. I’ve blogged about it before and have included “Ignore your cohorts” in my “9 Worst Practices in SaaS Metrics” slides.

My feeling is that over the last 12 months the awareness for the importance of cohort analyses has grown among startup founders. One reason may be that thought leaders like David Skok have been writing about the topic, another reason are web analytic tools like MixPanel and KissMetrics that make it simple to create cohort analyses.

And yet, many founders are still having difficulties with cohort analyses, be it with the collection of the data or the interpretation of the results. With that in mind I wanted to create a simple cohort analysis template for early-stage SaaS startups.

You can download the Excel file here.

The idea is that you have to enter only a small amount of data and everything else is calculated automatically. Specifically, what you’ll have to type in (or import from a data source) is the basic cohort data: How many customers did you acquire in each month and how many of them were retained in each subsequent month. If you also want to see your churn on an MRR basis and get a sense for your CLTV, you’ll also have to enter the corresponding revenue numbers.

If you’re not sure how to read a cohort analysis, here’s a quick explanation:

Here are some brief notes on each of the arrays in the sheet:

A1: This is where you enter the raw data. Start with January 2013 and enter the number of new customers that you’ve acquired in that month. Then move to the right and enter how many of those January 2013 customers were still customers in February, March, April and so on. Then move on to the next row. If your data goes further back than January 2013, extend the table accordingly.

A2 and A3: A2 takes the data from A1 and shows it in “left-aligned mode”, making it easier to compare different cohorts. As you can see the columns have changed from specific months to “lifetime months”. A3 shows the number of churned customers as opposed to the number of retained customers. Both A2 and A3 aren’t particularly insightful to look at per se, but the data is necessary for the calculations in B1, B2 and B3.

B1: Shows the percentage of retained customers, making it easy to see how retention develops over time as well as to compare different cohorts with each other. What you’ll want to see is that younger cohorts are getting better than older cohorts.

B2. This is kind of like the “inverse” of B1, showing the percentage of churned customers as opposed to the percentage of retained customers. In any given row, the sum of the percentages of churned customers plus the percentage of retained customers equals 100%.

B3: B3 is similar to B2, but the difference is that churn isn’t calculated relative to the original number of customers of the cohort but relative to the number of the cohort’s customers in the previous month. Let’s say you have a cohort with 100 customers and after 6 months the cohort has been reduced to 50 customers. If you lose 5 customers in month 7, this represents 5/100=5% churn in B2 but 5/50=10% churn in B3.

So what’s the correct number? There’s no right or wrong here, it depends on the question that you want to ask. If you want to know e.g. “How many customers do I lose within the first six months?”, B2 (in conjunction with B1) gives you the right answer. But if you want to know what percentage of customers you’re losing per month (important when you look at data across multiple cohorts and for lifetime estimates), take a look at B3.

What you’ll want to see in this table is that after a usually relatively high churn rate in the first lifetime months churn starts to stabilize (because the people who never really adopted the product in the first place are now gone).



C1-C3: Same as A1-A3, just for MRR instead of customer numbers.

D1-D3: Same as B1-B3, just for MRR instead of customer numbers. What you’ll want to see is that your MRR churn is lower than your customer churn due to account expansions.

E1 and E2: If you enter the CACs for each cohort, these tables show you when each cohort breaks even.

Also take a look at the second tab in the Excel sheet, which calculates/estimates customer lifetime and customer lifetime value on a cohort basis. Note that the data is highly speculative for younger cohorts for which there isn’t much data yet.

Further notes are included in the Excel sheets.

If you have any questions or comments, please feel free to reach out!