How One SaaS Startup Reduced Churn 71% Using “Red Flag” Metrics

As the Groove team sat around my kitchen table, the feeling I was experiencing most was worry.

It was January of 2013, and we were meeting to discuss our alarmingly high churn rate, which was hovering around 4.5%.

We were getting new customers at a steady, predictable rate, but users were leaving much too fast for sustainable growth.

As we pitched solutions back and forth, one thing became painfully clear: we had no definitive answer for why customers quit.

Sure, we had ideas. The app had bugs. We had too many features. Our onboarding process was complex.

And we were working hard to fix those flaws, but we still couldn’t point to any direct connections between a customer’s behavior within our app and their likelihood of leaving.

We had committed ourselves to a data-driven approach to user acquisition – readers of Groove’s blog have read about dozens of things we’ve learned from that undertaking – but we hadn’t, as one of our developers bluntly pointed out, applied the same principles to our retention efforts.

And so it was decided that, if we were to systematically reduce churn, we’d need to systematically study it and find its source.

After the meeting ended, that’s what I immediately set out to do.

What I found was illuminating.

There were very strong differences in the behavior of users who had abandoned Groove and the behavior of those who stayed.

Below, you’ll find some of the most valuable “Red Flag” Metrics (RFMs) we’ve found to identify churn before it happens and how we’ve used them to reduce churn to 1.6% over the past several months.

Which Engagement Metrics Were the Biggest Red Flags?

Disclaimer: Every business is different, and your customers will behave differently from our customers. My hope is not that you apply these exact metrics, but that you apply these principles and find your own RFMs. It takes work, but trust me (and see our results for yourself), it’s worth it.

When we looked at the numbers, there were two metrics that seemed to be the most significant in the first 30 days after a user signs up – length of first session and frequency of logins.

We found a large difference between the behavior of our churning users and the behavior of those who continued to use Groove.

There was a difference in total number of logins as well (and I’d recommend tracking it), but it wasn’t as disparate as the other two, so we focused on those.

Looking at one cohort (thanks to KISSmetrics for making this kind of data analysis simple), the average user who did not quit after 30 days spent three minutes and 18 seconds using Groove in their first session, and logged in an average of 4.4 times a day. The average user who quit spent 35 seconds using Groove in their first session, and logged in an average of 0.3 times per day.

abandonment metrics

What Abandonment Looks Like

That’s a massive chasm.

We began to send targeted emails to users who spent less than 2 minutes on their first session, as well as to those who (regardless of first session time) logged in fewer than 2 times a day in their first 10 days.

The first group was offered help with setting up their mailbox:

example email 1

That email got a 26% response rate; and more than 40% of the users who walked through their signup with us were still Groove customers after 30 days.

The second group – many of whom had already set up a mailbox but had simply tapered off usage afterward – received a slightly different email, this time offering a personalized strategy session:

example email 4

The response rate on this email was just over 15%; and nearly 50% of those users remained after 30 days.

In fact, the success of both of these campaigns has led us to implement similar tactics for onboarding all new users.

Takeaway: Look at your metrics and find the disparities between your most engaged users and the ones who’ve quit. Then, use what you learn to identify at-risk users and get involved right away.

Another (Less Obvious) Red Flag Metric

According to the White House Office of Consumer Affairs, 96% of customers who have a bad experience will not tell the business.

Instead, they’ll simply take their business elsewhere.

As a support company, we’ve used that insight to educate our users to keep open communication with their customers, so that issues get flagged long before they become big enough to drive people away.

But while developing our RFMs, we found a way to determine, with surprising accuracy, when customers were having issues they weren’t telling us about.

Here’s how:

We know how long certain actions within Groove are supposed to take.

Creating a rule generally takes between 10 and 30 seconds.

Integrating a Twitter account takes around 20 seconds.

Customizing a support widget is 2 to 3 minutes.

We looked at a number of in-app tasks and how long users were spending on those pages.

Sure enough, some of the users who had quit had spent significantly longer than average on (usually) a single task.

When a user got stuck on a task – for example, creating a rule – they wouldn’t always tell us. In some cases, that session would be their last.

So we began to reach out to users who we thought might be stuck on a particular task. When a user spent far longer than average on a particular page in more than one session, we reached out and asked if they needed help.

example email 2

The response rate to this email, while much lower than the emails above, was still around 10%; and around 30% of the users who responded were still customers after 30 days. That number may seem low, but the data suggests that there’s a very high chance they would have abandoned Groove otherwise.

Takeaway: Your customers probably won’t tell you when they hit a snag. Dig into your data and look for creative ways to find those customers having trouble, and help them.

Bonus: Using RFMs to Find Your HAPPIEST Customers

While RFMs can help you spot your “flight risks,” you easily can use the same data to leverage your most engaged customers.

Simply look at the outliers on the opposite end.

We created a segment of customers who, based on frequency of logins, were Groove power users. Then, we sent an email asking those customers for a referral.

last email

We also sent the same email to a random group of customers without any pre-qualification and without the reference to usage.

The result?

Our qualified users sent us nearly 400% more referrals than the second group.

Takeaway: RFMs can help you find your best customers, too. Use that data to get referrals and grow your business.

The Importance of RFMs and Other Analytics

Now, more than ever, we (SaaS businesses) have access to amazing data.

And we’re always finding new and amazing ways to leverage that data to grow.

RFMs (which you may already use by a different name) provide an opportunity to keep our most at-risk customers on board, and to make our happiest customers even happier.

If you don’t use them, I hope you’ll start. It’s had a massive impact on Groove, and I’m confident it’ll have a massive impact for you.

Finally, a Request

Speaking of data, most startups and small businesses keep it private.

And while there are good reasons for doing so, it means that there’s precious little real-world data out there to benchmark ourselves against.

That’s why we’ve launched the 2013 SaaS Small Business Conversion Survey.

We’re asking SaaS startups and small businesses to (anonymously) share data on conversion rates, user acquisition strategies, and other facets of their businesses, to make us all more educated, more savvy, and more successful.

Plus, we’re offering THOUSANDS of dollars in prizes from partners like KISSmetrics, Unbounce, Mixergy, Buffer, Clarity.fm, and more.

If you gained value from this article, I’d really appreciate it if you would take five minutes to complete the survey here.

Or, to read more and learn about the prizes, check out our announcement here.

Thank you!

About the Author: Alex Turnbull is the Founder and CEO of Groove, a customer support app for startups and small businesses. You can follow him on Twitter or read more at the Groove Blog.

  1. Alex how exactly do you track when a user attempts to make use of a feature but did not complete it? I am asking regarding the section of your guest post titled “Another (Less Obvious) Red Flag Metric”

    • Thanks for the comment, Owen!

      We’re using “time on page” as an indicator…it’s not perfectly exact, but when a user spends higher-than-average amounts of time on a single action for multiple sessions, we’ve had a pretty good hit rate of guessing that they’re having trouble with that particular action.

  2. What analytics program are you using to track users interaction with your app?

  3. Nice work putting this into practice, Alex, and congrats on the results. I am curious about the process you undertook to find the initial two red flag metrics. I suspect that frequency of logins was an easy enough place to start checking into just based on your intuition– but how did you know to focus on the first 30 days instead of 14 or 60? And in particular, how did you arrive at the second RFM– amount of time for the first session. Was it just trial and error across a bunch of possible usage metrics, or did something lead you there? I’d be curious to hear about other metrics you considered but rejected– such a usage of particular “sticky” features, or support interactions, etc? Again, not to try to replicate your specific situation, but to understand your process for developing and validating hypotheses about churn indicators.

    • Hey Tom, thanks for the comment!

      That 30-day figure was actually tied to the length of our 30-day free trial at the time that we started using RFM’s. Now, we use a 14-day scale to reflect our 14-day free trial, as that’s when we need to convert users by.

      There are a few metrics that we studied pretty closely, but found that there was either little correlation between those metrics and conversions, that we couldn’t tie them closely enough together, or that we simply failed at moving the needle on those metrics.

      Examples of RFM metrics we’ve put on the back burner for now are sign-up autoresponder open rates and click-through rates, downloads from the Groove App Store and customization of widgets.

      There’s definitely plenty of trial and error involved :-)

  4. Alex,

    I think you win the “most transparent entrepreneur”. Were working on email triggers based on usage and this really confirmed the direction we are headed. I’ve yet to read an article you’ve written that wasn’t packed with timely, quality content and necessary for just about every SaaS team to read. Kudos for sharing the knowledge and for driving the SaaS SMB conversion conversation.

  5. Love your blog Alex and appreciate your transparency. Really enjoyed reading this article. Shows your great understanding of KPIs and dedication to utilizing data to take proactive action. So many companies just want to throw money at their problems to keep revenue up, when they aren’t willing or able to address their fundamental internal deficiencies that would prove so much more valuable in the long run.

    • Thanks, Michael!

      Couldn’t agree more; throwing money at business problems can sometimes stop the bleeding for a while. Doing the hard work of truly understanding what’s causing the problem, and then re-engineering your way you do things to ensure that the problem never occurs again is critical for long-term growth. Especially for us startups.

  6. WOW Alex it is the most brilliant post i ever read!
    I read all of you’r blog’s posts but this one give me so much insight about KPI and success metrics so its really really awesome.

  7. Thanks Alex for this great article. Lots of insights there!

    I have two qustions for you as we also use Kissmetrics and would definitely need to focus on these RFMs as well!

    1st: How did you obtain that first screenshot where you get the lenght of the first session and the average sign-in per day during the first 30 days. I’ve checked our metrics tab and can’t figure out how to do this! I may miss something. Or is this something that requires a special plan?

    2nd: How do you manage to send targeted emails to a specific batch of users yu identify through Kissmetrics, do go ahead, run the people search, export as excel, import in (mailchimp?) and the send the email? Sounds like an awfully painful process to me. Or is there a secret way to retrieve those emails automatically and automate that process? Really curious about this one because the inability to automate that process is really what makes us not do this kind of things using kissmetrcs.

    Thanks!

    PS: I’m a big fan of your blog :-) Too bad we were already using a competing product before I discovered you (and are happy about it), but at least, now, I am mentionning you guys everytime I am asked what support system to use. So it works, even for people not using you!

    • Thanks for the comment and the great questions, Emeric!

      1) That data is definitely in KISSmetrics, just not in the way we’ve mocked up (for visual purposes to make it easier to see). We simply run a custom report…feel free to email me for help :)

      2) We handle all of that through intercom.io

      And thank you! I’m happy you find the blog useful, and that you tell your friends about us. I hope you’ll give us a try, too ;)

  8. Great to see such an informative article on how solutions were practically applied to solving critical issues. An email is not to be underestimated! We put together a quick intro guide to SaaS startup metrics (5 of the most well-known creators and their methods) that might help readers set their metrics right up from the beginning and gather as much of the right data as possible: http://blog.twoodo.com/288/best-intro-guide-to-saas-startup-metrics/?utm_source=blog&utm_medium=comment&utm_campaign=techblog

22 comments

Please use your real name and a corresponding social media profile when commenting. Otherwise, your comment may be deleted.

← Previous ArticleNext Article →