Customer churn: how we cured our startup cancer

Customer churn is cancer for a SaaS company. Overlooking it has been the single biggest mistake our company made, and the closest we’ve ever been to going out of business.

Churn is a silent killer that slowly but surely will eat away your business. You can cure it if you catch it early on, but if you let your churn problem go unchecked, it will kill your startup.

We learned this the hard way.

Let me give some context on how we found ourselves there: between 2015 and 2016 our company grew REALLY fast. We 5x’d our subscriptions in just 12 months and everyone was really excited.

In 2016 we raised the second part of our seed round, and it was the easiest money I’ve ever raised; it was obvious that we were onto something, and the diligence process with all of our investors went pretty smoothly. There was inevitable FOMO.

And I remember one investor came to us saying they were going to pass because our customer churn was too high. At that point nobody paid much attention, and let me show you why.

You’ll see that we were adding $6,000 to $7,000 in new subscriptions every month. In the second half of 2016, that grew to $10-$12,000 in new MRR every month, which again, got everyone really excited.

We were consistently adding $4-$5K in new Monthly recurring revenue (or MRR), and moving fast towards the $1 millions dollars in Annual Run Rate milestone.

By the way I am fast-tracking through some of this terminology, assuming that you have already seen some of our videos. If you haven’t, please check our our SaaS metrics video- I promise you it will help you make sense of all this.

So the big numbers of new customers made us ignore the number of LOST customers.If you are only focused on that pretty, growing, MRR chart- it’s very easy to miss what’s happening under the hood.

So let me show you what we did to fix it.

The first part was understanding the why.

Why we get customer churn

In general, people cancel because they are not getting enough value out of a subscription, not enough to justify the price. Period.

You can talk all you want about onboarding, aha moments and cancellation flows, it all comes down to value obtained vs price. The more value you bring, the more money you can get away with charging.

I get that question a lot: what should be the pricing of a SaaS product? From the retail world, we are accustomed to thinking about margins: we markup the manufacturing price to generate profits.

In software it’s very different. Software is very cheap to operate, so we could very much leave money on the table if we aren’t aggressive about pricing.

My take here is, the pricing should be as much as we can charge. Simple. You find that number by testing, there is no way to guess it, and user interviews or surveys are definitely not going to give you a final answer.

But balancing those two variables- value obtained and price, is some of the most complex challenges to solve for SaaS products.

Value

So let’s talk about value first.

The value that a customer gets from a product can be affected by,

• 1- The Onboarding: if you are not explaining well what the tool does, then customers won’t extract enough value out of it; because maybe they don’t know how.
• 2- The UX or user experience: the interface. If the customer can’t find what they are looking for, or if they find the experience frustrating. We are willing to forgive UX hiccups, but only up to a certain point.
• 3- Simply, how useful the tool is in their day to day. Are they using it daily? Weekly? What’s the alternative: how much of a better experience are you giving them versus the alternative.

Some examples from our own book: we use Intercom for our messaging, onboarding emails, support and newsletters- and it’s over $5,000/mo for us. It’s one of our most expensive SaaS platforms- but we couldn’t live without it: the alternative just doesn’t have the features, so we’d have to mix different tools… and it would still cost somewhere in that range.

Another service that has locked us for life is ChartMogul: it’s the platform you’ve been seeing when we refer to our SaaS metrics. We’ve been tracking our SaaS KPIs there for years, the set of features gives us what we need and we don’t want to go through the process of migrating all this data elsewhere. It’s a lot of data.

• 4- Another variable you might have noticed here is how hard it is to leave. For B2B software you want to find something that embeds deeply in the work pipeline of your customers, so that they depend on your tool. Those are the companies that thrive. Look at Slack, or heck, look at Microsoft Office.

You can also see this in consumer products. I recently migrated from Spotify to Apple Music, on their Apple One plan-  and it was a tough call, mostly because of the work required or the fear of losing some of my playlists.

So your focus as a founder or product manager (you are probably going to be both for a while) is creating a product that adds clear value to your customer’s lives. It can’t be just a ‘nice to have’; ideally, it should be an ‘I can’t live without this’ relationship.

And there are a lot of ways you can measure this.

First, you want to ask them. Get into conversations with your customers. I do probably 20-30 user calls a month, and that gives me a qualitative view, and as much as those calls are focused on helping them in whatever way I can, I’m taking notes. I’m asking which features they have found or haven’t found- or which ones are they using more frequently.

For a more quantitative measure, there’s a little survey you can use. You’ve probably seen it before. ‘How disappointed would you be if you could no longer use Slidebean’?.

Another standardized measure is NPS surveys (which stands for Net Promoter Score). The question here is ‘How likely are you to recommend Slidebean to a friend?’ and a 0-10 meter.

Any answer below 6 is considered bad, so that’s -1 point. Answers between 7 and 8 are considered average, so neutral. The only answers that really speak passion about the product are 9s and 10s.

So with the sum of your NPS answer should give you a number, usually between 20-40. 40+ is considered great, 70+ is considered World Class.

All of these indicators can be considered predictions for churn. What we do is we pull this data into ChartMogul so we can look at at the aggregated data based on their NPS score, and understand the likelihood of one of these groups churning, or our effectiveness month over month at getting people to give us a 9/10 vs as 7/ 8.

Last but not least, you can simply ask them why they are leaving. You’ve seen this quite a lot in cancellation flows. This is valuable information, but with a big catch, there’s not much you can do about it. If a user has cancelled your chances of re-engaging with them are slim.

What you can use, is plug those reasons into your data and study why they are cancelling.

In our case, most of the cancellations we saw in 2016 could be summarized as ‘I like your product, I just don’t need it anymore’.

And that was painful to hear. The story was consistent, I can still see it in my 2016 nightmares: 9 or 10 NPS score, great customer satisfaction, but ‘I just don’t need to make presentations anymore’.

But that answer will be different from company to company. You might be getting complaints on missing features or bugs, and those are generally easier to tackle, easier to address- because the customers are telling you what’s driving them away. Again, as long as you are taking the time to track it.

But our cancelation reason reflected a deeper, more fundamental problem with our product. This was combined with the fact that our second most common reason for cancellation was ‘too expensive’.

So we figured, well, maybe playing around with pricing was the solution. And boy were we wrong. I like to call this stage of our process ‘half measures’. Because I love Breaking Bad and it’s the best TV series ever made and half measures, we learned later, were not enough.

Half Measures

Pricing

Going more expensive

Remember, two variables affect customer churn and retention: value and pricing.

We have experimented with pricing A LOT. Started at $4.99, then went to $29/mo, then to $79/mo.

Our biggest lesson from all of this is that when converting, people are less sensitive to pricing than you might think. One of our biggest breakthroughs back when we were starting the company was that we discovered that the amount of people converting to a $4.99/mo plan, vs an annual-only, $160/yr plan- was pretty similar.

Conversion rate was only slightly affected with the latter pricing, and we of course made a lot more money…  with a year’s subscription up front, no less.

So our first thesis to solve this was,

• Allowing customers to pay to unlock a single presentation.
• Ramping up the price of the subscription, so that only customers that are power presenters, regular users, will choose the subscription route.

We settled at $29 for a single presentation, and monthly plans at $49 and $79.

The thesis was not bad. Looking back, I think the thesis is solid. But boy did it fail.

At first, we saw a nice spike in non-recurring revenue, which ChartMogul reports separately.

And while the number of subscribers was reduced, we still saw a decent amount of people picking that option.

But the real problem: Customer churn, would come a couple of months later.

The new higher pricing spiked churn to 25%. People were sort of accepting the new pricing to convert, but they were really sensitive about it and would cancel one or two months later. Of course. The value this product was bringing was not $79 worth.

Again, we thought only power presenters would pick that plan, but that wasn’t the case.

Even worse, our NPS score dropped a lot. People were suddenly more sensitive to bugs and stuff that they couldn’t solve, and they expected more hands-on support.

One detail we did not foresee here was how much revenue shifted from recurring, to non-recurring. We had to chase those sales every month, and a bad month would seriously endanger our runway.

Going cheaper

After a failure in that approach we considered going with a lower pricing. We came up with a plan we called Starter, which was priced at $12/mo.

And that also kind of worked. Churn on the $12/mo plan was better, under 10% for the first time in a while… but the next problem is lifetime value, and cash flow.

On a $79/mo plan with 20% customer churn, your average lifetime value per customer is $395. That means the average customer stays on the platform for about 5 months. But more importantly for cash flow, you collect $395 in 5 months.

Now don’t get me wrong, that level of churn will kill any business, but at that time we were acquiring customers for about $50. So we spend $50, we recover them in the first month, and then we make $345 worth of profit within the next 5 months.

On a $12/mo plan with a 5% churn rate, your lifetime value is $240. It’s less. And worse than that, it takes you 20 months to make that money. It takes you 4 months to recover the $50 you spent bringing that customer in.

So we had sort of solved the churn problem, but the pricing created other serious cash flow problems.

The lesson with these pricing changes is,

• Experiment, yes; but
• Always keep in mind how it will affect your cash flow and your other metrics.

Your job as CEO is not running out of money. In our case, these metrics broke our profitability but in a company looking to raise funding, they could affect the story you are telling investors.

The good thing is we learned A LOT from the behaviour of our customers. And it allowed us to eventually solve the problem.

Other Half Measures

By the way, pricing experiments weren’t the only effort we were making to improve retention.

1- We set up an aggressive dunning campaign: dunning emails are used to re-engage with customers that churn, both voluntarily and involuntarily.

Voluntary churn comes from people that cancelled. Involuntary churn comes from people whose credit cards failed or expired and failed to update them. It’s churn either way, but you tackle it differently.

2- We also iterated a lot over our onboarding process, both visually and with emails. We became really good at getting great open rates, which we achieved by sending really personalized (but automated) emails that included details from the customer’s activity on the platform.

Those are good efforts, and we could accurately measure how much churn was stopped thanks to those, but it was not enough.

Full Measures

What we had failed to see through all these years, through all these experiments, was that our problem was a product-market-fit problem.

Product-Market-Fit is this elusive achievement for startups: it happens when you find the ideal balance between product and target audience. When you find that sweet spot, suddenly your marketing starts working better, your NPS gets higher and your retention goes up. A very Cinderella moment.

And once again, the answer, oftentimes, is in the data.

When we set out to build Slidebean we created this alternative approach to designing presentations, content first, automated design. We thought of it as a platform that a lot of different types of customers could use: it worked for teachers and professors, it worked for sales teams, for consultants, for marketers and account executives. Essentially anyone who needed to present often and who cared that those presentations looked good.

While we didn’t know what the customers were doing on the platform (because we don’t have access to their slides) we could still see what templates they were picking. So we could classify them in startups, marketers, salespeople, academic use.

And lo and behold how our churn numbers were drastically different for the startup group. They were staying longer, using the platform more, rating us better on NPS scores!

This was the beginning of a company-wide pivot.

Over the course of the next few months, we embraced how the ideal use for Slidebean seemed to be ‘pitch decks’. That required us to refocus our marketing (into something like these videos) and even to refocus the product, re-do and re-name some features.

These customers were already, clearly, better customers for us than the rest. So now it was just a matter of improving the experience. Doubling down.

So aside from features and marketing we made a key change in our business model after we understood who the audience was. If you are using Slidebean for pitch decks… Why would you have a monthly product?

Pitching investors is not a monthly thing. You don’t make a pitch and are done and finished raising money in 30 days. Or in 60 days. Raising a round takes, easily, 6 months… So what if instead of thinking of months, our pricing reflected the length of the fundraising process.

That would have made up for 6 month or 9 month plans, but in the end we settled for 12 months. An annual plan that would make sure founders were covered for the whole extent of the fundraising process.

So we switched to annual only plans. And you might think we are just cheating on customer churn: forcing people into a 12 month commitment vs the ability to cancel monthly. But if that were the case, then the retention at the 13th month would be similar to the monthly plans, right?

Well we improved that by about 4x. The number of customers that were still with us after their first chance to leave, was 4x higher than with monthly plans?

The reasons behind that are a complex combination of variables that we probably haven’t fully understood yet.

It certainly has to do with us having more time to make sure they get value out of the product. It definitely has to do with customers having more time to use the product, more time to get more presentations into the platform, and making the cancelling a tougher choice.

And the change is not without its set of problems. We know the conversion barrier is high, we know the pricing is steep for customers outside the US and trust me, we are trying to figure it out.

So the lesson of all of this is, don’t be afraid to experiment- but use your data to make better decisions.