COC: A new metric for thinking about cancellations in SaaS business models

This is the second article in a series on novel ideas for SaaS metrics, which started with The unprofitable SaaS business model trap.

You’ve probably know that cancellations kill growth in scaling SaaS companies. You probably know this is because as you get large it’s hard to continue growing fast enough to cover the amount of customers you’re losing by cancellation.

I have a better, clearer way to think about this financial effect, one that makes it more obvious why this “headwind to growth” is larger than you think, and an stronger way to compute exactly what the effect is.


The cost to earn a dollar of MRR

What does it cost a SaaS company to add $1 of new monthly recurring revenue?

Using the typical acronyms:

CAC (Cost to Acquire a Customer) is the total, all-in cost to get a full customer in the door.  Marketing and Sales costs, plus the fully-loaded salaries of the folks in those divisions, including commissions and everything.

MRR (Monthly Recurring Revenue) is the revenue you get every month from a customer, averaged across all customers. Thus, MRR = R / N, where R is your total revenue, and N is your total number of customers.

Since it costs CAC dollars to get MRR dollars of revenue,

p = “cost to earn a dollar of MRR” = CAC / MRR

p is also “pay-back period”

Students of SaaS business model metrics will recognize p as the “pay-back period,” meaning “the number of months before a customer ‘pays back’ the cost to acquire that customer.” That’s why I picked the variable p – as in “payback period.”

It’s fun that these two interpretations of p are both accurate.

COC — the Cost of Cancellations

Suppose your monthly cancellation rate, in terms of MRR, is c.  For example, if 2% of your monthly revenue cancels in a single month, c = 0.02.

Now, just to stay even in revenue (forget about growing!), you to have to replace c dollars of MRR.  And each dollar of MRR costs you p dollars to acquire.

Hence the formula for what I call COC (the Cost Of Cancellations) — the fraction of revenue you must spend merely to keep your revenue flat:

COC = pc

Some examples make the utility of this metric clear:

If your cancellation rate is 2%/mo (c=0.02), and your marketing pay-back period is 6 months (p=6) — typical for a healthy SaaS business — then 12% of your revenue every month will be spent just keeping revenues even.

That’s a tremendous percentage of revenue just to keep from shrinking!  That doesn’t include cost to serve (customer support and servers), that doesn’t include overhead costs, that’s don’t include spending to actually grow revenues, that’s merely to stop shrinkage.

Enterprise SaaS businesses often have lower monthly cancellations but much longer pay-back periods.  1.5% cancellation and 18 month pay-back period means a whopping 27% of revenue is spent replacing cancellations.

A no-touch SaaS business driven by word-of-mouth marketing might have lower pay-back periods due to efficient acquisition costs, but have higher cancellation rates due to the lack of human touch.  I know a prominent SaaS business with a pay-back period of 2 months but a cancellation rate of 5% — that’s 10% of revenue to stay even.

That’s actually pretty good if you think about it.  5%/mo cancellation means 50% of their revenue cancels each year — crazy high! — but a very low cost-to-acquire means the company is still spending only 10% of revenue to stay even, and of course an additional 10% of revenue spent on marketing causes them to grow at a reasonable clip.

Important COC Lesson #1: Three metrics matter, not one

The massive size of COC for most SaaS businesses should be a wake-up call.  COC expense can often be as high as R&D or G&A — that’s an insane, unprofitable, and therefore untenable expense-line.

A SaaS business must work constantly to reduce COC.  Because the definition is so simple, it’s obvious that reducing COC means decreasing cancellations and decreasing p, and decreasing p means decreasing CAC and increasing MRR.

This is a key insight to many SaaS operators, because typical metrics literature focuses only on reducing cancellation rate, which is only a third of the story.  Furthermore, unless you haven’t yet reached product/market fit, cancellation rate is often hard to shift compared to reducing CAC (smarter campaigns, optimized landing pages, self-serve sales/on-boarding) or increasing average MRR (different pricing tiers, add-ons for new product or service).

For example, my company WP Engine‘s cancellation rate is under 2% per month.  That’s low for the hosting sector. Furthermore, when we poll customers who cancel after the first 90 days, the most common reason for cancellation is “project ended.”  Meaning, it’s not something we can affect by changing something internally.

Therefore, pay-back period is a smarter place for us to focus in terms of reducing COC.  In fact our CAC is also already very low, due to tremendous word-of-mouth that our lovely customers bestow upon us. (And now you see how much we appreciate that!) But maybe we should do more to increase word-of-mouth activity while we continue to optimize our paid advertising campaigns. And what about MRR? We’ve started selling SSL certificates to customers who want secured sites, which results in incremental revenue (and a better customer experience, because we handle that mess for the customer, including renewing and re-installing the certificates each year).

As another example, take the company above with the 5% cancellation rate that traditional metrics literature would say creates “impossible headwinds for growth,” and yet they have a better COC than a typical enterprise SaaS business, demonstrating that a very good pay-back period can overwhelm a crappy cancellation rate.

Important COC Lesson #2: Zero net churn trumps both CAC and MRR

The standard SaaS metrics literature does give an important way to combat “headwind” from cancellations: Up-selling existing customers.  If you have 2%/mo cancellations, but if on average you increase MRR by 1%/mo with things like customers graduating to larger tiers, adding more “seats,” buying add-ons, buying premium support, etc., then effectively you’re only losing 1% of your MRR per month, not 2%.

Thus, c above is not really cancellation rate, but rather what is called “net churn,” meaning cancellation rate minus up-sell rate.  The strongest SaaS companies have negative net churn!

You can see the effect of approaching zero net-churn in COC: If c is zero, COC is zero, which means “getting back to even” costs nothing at all.  Phew!

Another fact pops out: In terms of “not shrinking,” suddenly p — and therefore CAC and MRR — doesn’t matter at all!  Of course they do matter for healthy revenue and inexpensive growth, but at least in the “headwind” sense they fall away.

This highlights the fact that getting to zero net churn is the strongest thing you can do in terms of COC.  And since we just talked about “cancellation rate” having a floor, that means you must develop paths for up-sells.

Of course negative net churn is a bonus.  It sends COC negative too, which can be interpreted as adding directly to profit margin, just as a positive COC sadly takes a tremendous bite out of profit margin.

COC: A new standard SaaS metric

Typical SaaS metrics literature characterizes it this way: “As a SaaS company scales, growth rate diminishes, but cancellation rate doesn’t.  That means it gets harder stay ahead of the headwinds created by cancellation.  So you need to work on getting to zero net-churn and, if you don’t, you cannot build a large SaaS company, and certainly not a profitable one.”

That’s true, but thinking about it in terms of COC is an easier way to understand how to define “headwinds,” exactly how big your “headwind” currently is, the true components of “headwind,” and thus points the way to how you will reduce the headwinds besides reducing net-churn.

Special thanks to consummate SaaS-metrics blogger Joel York (go read!) for reviewing drafts of this post.

  • Justin

    Great in-depth post Jason. I’m hoping to run the COC calculation later today for us.

  • Ruben

    This is interesting. I’ve been focusing on churn a lot lately but haven’t really looked at it like this. That said, I tend to look at payback period by channel and each of those varies. I do get a large number of customers from existing sources that don’t require monthly spend (and had low spend to start with), so if include all new customers and average everything out, my payback period is under a month. Obviously, I want to look at this by channel, but what’s the value of averaging it all out if a larger percentage is from sources that I’m not actively spending money in?

    • Jason Cohen

      Definitely you should be measuring CAC by channel. I like using pay-back period as the measure of that effectiveness. That’s how to do it operationally, i.e. day-to-day.

      The reason it’s also useful at a whole-company perspective is to understand the macro-scale effects on growth and profitability.

  • Ian

    Great post!

    I think the punchline is increase your MRR to offset, not eliminate costs and churn. Which makes sense. But, you are proposing to do so with add-on services which will inherently increase your CAC. There are costs to managing those pesky SSL certificates. So we are assuming MMR will still outpace CAC but the gap or rate is not necessarily eliminated, but hopefully increased. Seems to me that you still need to be creative in the ways you control your CAC with the new offerings. If you have to add a new division to the company and 10 employees for a marginal gain in MMR, the overhead could cause a problem in the future if you see an unexpected spike in cancellations. So maybe part of the equation is to add extremely high margin up-sells which have low operational costs.

    I also found some of Joel’s equations interesting, especially when modeling max customer figures:

    • dskaletsky

      Great question. I had the same one. Especially on the Enterprise SaaS side, there are definitely costs associated with driving the upgrades that offset churn (I’ve seen them be as high as the sales cost – not marketing costs – associated with acquiring a new customer).

      How would you take these costs into account when calculating the impact of Zero/Negative Net Churn?

      • Jason Cohen

        Thanks both of you, good conversation.

        I agree that if CAC is high on increasing MRR, that should be factored in. However, the CAC for upsells is generally far less than getting the initial sale. There’s no way that sales cost for a new *enterprise* customer is the same as that of an upgrade; also typically upsells should be from an account manager or special sales team. That could be true for SMB or consumer, but in that case you shouldn’t be using salespeople for upsells anyway — you need low-touch sales for low-dollar sales.

        Also, remember that “upsells” are generally features that you can sell to new customers in the first place, which means increasing all MRR, including the marginal new MRR, which is quite impactful.

        It’s true that in the end, there is always a limit to how large a company can comfortably grow. Not just because of churn and other costs — although certainly that too — but because as market penetration increases there’s less to get, saturated channels increase CAC, etc..

        That’s why companies who continue to grow at scale do so with innovation outside this narrow view of optimizing the SaaS metrics. Building ecosystems (e.g. sforce for, expanding markets (e.g. now going hard at the CMO office and even the CIO office, or Microsoft creating XBox), and when they’re really smart, disrupting themselves before someone else does (e.g. Amazon with Kindle)

        • dskaletsky

          I think the lesson in this string is really a product one. Meaning, if the cost of upsell/upgrade (and eventually renewal) are high (there is probably a good way to benchmark this against initial CAC), then instead of trying to figure out a way to reflect this in the metrics, it’s probably time better spent to take a hard look at your product. Why isn’t it driving easier upsells/upgrades? Is it a feature issue? Pricing structure issue? Usability issue? Awareness issue? Etc…

          Thanks for the great post & conversation!

  • Lilia Tovbin

    It’s an interesting way to think about cost associated with growth. I find it hard to trust the monthly numbers to draw any conclusions because my business is highly seasonal. But since a descent number of my conversions are assisted, rather than attributed to a single channel, it helps to average out to an overall rate to understand the bottom line (projected growth and profitability) and the COC helps with that.

    • Jason Cohen

      Correct, for seasonal businesses you cannot use monthly numbers. One way to *sort of* do that is to first normalize for seasonal changes, but (a) there’s too much variability in small businesses to do that with precision and (b) the component numbers probably change anyway throughout the season due to different kinds of people with different expectations.

      You might need to look year-over-year, or compared with same-time-last-year. That means you don’t get data updates as quickly. But you have no choice!

      • Lilia Tovbin

        Right, for big picture I look at year-over-year and for strategic decisions I look at same-month-last year and then compare the change in a growth rate between months to see how it’s trending.

  • Walter Scott

    Shouldn’t CAC be “monthly” fully-loaded sales and marketing costs plus COGS? A number of SaaS solutions have high OEM, hosting or support costs that should be considered. Would billings instead of revenue be a better factor in MMR as revenue could be deferred based on delivery or contract terms? In SaaS valuations cash flow not EBITDA matters.

    • Jason Cohen

      No. :-) But thanks for bringing up this important point.

      What you’re getting at is that “paying back CAC” should be thought of in gross revenue, that is in “revenue minus COGS” or also known as “MRR x GPM.” Because if COGS is 50% of revenue, and an average order is $50/mo, and your average CAC is $150, the way I’ve defined pay-back period (p) above says it’s paid back in 3 months, but your point is that, no, it’s 6 months!

      So when you’re talking about “marketing payback period” then I agree with you completely. In that sense, the statement I make above about “this is the same as marketing payback period” isn’t correct, although almost all the literature talks about marketing payback in revenue-months not gross-revenue-months, so I’m being consistent with the usual definitions.

      However, with COC, the gross-revenue payback is NOT the correct calcuation! It really is “p” as defined in the article.

      The reason is that it’s computing the replacement cost of a number of dollars of TOP LINE revenue, not the replacement cost of bottom-line or gross revenue.

      REGARDLESS OF GPM, if p=CAC/MRR=3, then it costs e.g. $150 to reacquire $50/mo of top-line revenue. Therefore if $c cancel, it costs $p*c to replace, in dollars.

      Hope that makes sense!

  • Guilhem Bertholet

    Makes a lot of sense and gives us some insights to better fine-tune SaaS models as we’re pretty all aware that it’s not an easy game to play! :)

  • Pingback: TMBA39 – “Above the Fold is Dead” – 8 Design Tips For Non Designers

  • Vlasta

    Jason, it seems like you are saying that the best SaaS business is not SaaS at all. What you describe looks like you want to sell classic products to customers, because this way you extract the value instantly, and you want to sell different products or upgrades to the same customers at later time.

    If it looks like a duck, swims like a duck, and quacks like a duck…

    • Jason Cohen

      I am definitely NOT saying that.

      There’s nothing above about “extracting value instantly.” SaaS models allow for the business to improve over time and actually have better financials against existing customers — something that traditional companies cannot do. Also — at least in theory — you get more total dollars from a successful customer.

      Rather, there are of course important aspects to SaaS businesses which can sink them, or prevent them from being profitable. You have to be vigilant about that, and it can be harder to track exactly because all the financials — revenue and cost — is spread out over time, and there’s things (cancellations, but on the flip side, upgrades) which can happen along the way.

  • Brandon Doyle

    Great post. I also just read your AMA on Inbound as well. Tons of great tips there. Thanks for taking time to do that!

  • Mathias

    Great post, thx a lot! We
    emphasized the importance of monitoring and “managing” churn during our
    recent Business Model 101 presentation ( as well – this is a great follow-up reading to this!

  • Pingback: Issue 5 - Its all about growth | SaaS Weekly

  • Clint Wilson

    @asmartbear:disqus, great walk-thru of the business and I finally got our team to produce all our metrics from our biz this year. Well, one thing I learned is “The strongest SaaS companies have negative net churn!” and I will stop beating my teams up so much over churn now. (I will beat them up for negative customer reviews however)

    I do think you added a great point too on the fact that some churn is uncontrollable like “project cancellations”. Upon reflection we have similar churn with “software platform changed” as well as we started longer Enterprise trial periods.

    Great read:)


  • Megan Lloyd

    ) I was impressed by your epic guide above; you’ve really
    achieved your goal of sharable, epic content.

    Thank you for summarizing everything for us.


  • Molly Smith

    A classic Smartbear blog….

  • Pingback: Some useful resources for understanding and benchmarking SaaS metrics | Memia

  • Pingback: Why I don’t like the LTV metric by @ASmartBear

  • AlexanderRaskin

    Great article. Thanks a lot.

  • Pingback: So farewell, then, 2013… | Memia

  • Raejetsetter

    Hi Jason, thanks for the great post! Just quick question, how did you calculate the 50% in ’5%/mo cancellation means 50% of their revenue cancels each year — crazy high!’? Should it be 5%/mo x 12 mo/yr = 60% revenue cancels each year? Thanks! Rae

    • Jason Cohen

      It’s not 5% x 12. Percentages are cumulative. So, thinking about it from a “retention” perspective, in the first month you’ve retained 95%, then the second retains 95% *of that*, etc, 12 times, which is 0.95^12, which leaves you with about 54% remaining. I rounded in the text to 50%.

  • Z

    We are currently working on our 2014 budget and SAAS metrics. We have partnership/affiliate deals where we are required to pay a recurring commission each month based on MRR. When calculating the cost of acquisition should we be factoring in the full cost of recurring commission over the life time value?

    • Jason Cohen

      No, you need to take that out of your GPM, not your CAC. When it’s recurring for the life of the customer it’s not a “one time charge” which is what gets put into CAC, but rather a “recurring charge” which means GPM.

  • Arthur Clementin

    I’ve been looking for people like you going beyond growth & recruitment. Simply the idea of growth quality is interesting and opens a completely different approach. Looking forward your next thinking.

  • Pingback: Ten Resources for Understanding SaaS Business | What I'm Learning