Early on the morning of September 23rd, 1999 the Mars Climate Orbiter was entering the last leg of its 250 million mile journey from Earth to Mars and was beginning its orbital insertion when contact was lost. Two minutes later the craft passed to the far side of the planet and for 21 minutes scientists on earth waited to see if it was alive. When the Orbiter was expected to emerge from the far side no contact was made. It was soon determined that the craft had experienced a catastrophic failure and on September 25th it was declared a loss at a total cost of $327.6 million.
For two months an investigation board researched the incident and in November of 1999 they released a draft of their report. In the end it wasn’t the rigors of interstellar travel that doomed the craft, nor the complex equations necessary to get the craft to the right location at the right speed. All of that was fine. The problem? An engineer at Lockheed Martin forgot to convert US units to metric units. They had mixed apples and oranges.
I was working through a financial model the other day and I ran into a common apples & oranges issue in startup economics. As with any model, I started with simple assumptions and built up from there. I modeled a company with 100,000 active users that was gaining new users at a steady rate of 10,000 per month (note: all numbers have been changed to protect the innocent…). This yielded a growth rate that was a nice straight line going up and to the right.
I added revenue from a base product ($10/mo) to the equation, and next was revenue from an up-sell (+$5/mo). Then I built in some conversion metrics moving users from the base to the up-sell at a rate of 5% per month. I was feeling pretty good about things.
But you want to make sure a model has a good reflection of reality and that means you don’t just gain users, you also lose them too. That, my friends, is called churn. You hate to see them go but the reality of life is that they won’t all love you forever.
So I figured that this super-awesome company making this super-awesome product would lose about 5,000 of its users at the outset on a monthly basis. But there’s a problem, you really don’t want to assume that you’ll be losing a constant number otherwise that would imply that your churn — which starts at 5% when you have 100k users — would drop to 1.5% by the time you have 340k users in 24 months. That doesn’t hold up, if anything churn has a tendency to grow over time, not fall (as you move from early adopter true-believers to a wider audience, the likelihood of gaining a user less likely to stick around tends to grow for most products). So you can’t use a constant and 5% sounds like a good place to be. Let’s take a look at our user base now:
You can start to see the problem. That line is going asymptotic (eg: it’s getting flatter). What happened to my nice pretty up-and-to-the-right growth? Well my friends, it got eaten by churn. That 5% churn in month 1 was 5,000 users and my new user acquisition was 10,000 yielding a net of 5,000 new users. So even in my first month, adding a 5% churn cut my absolute growth in half. But it gets worse, the more I grow, the larger 5% is in absolute terms. Whereas in my earlier model I was at 340k users by the end of the 24th month, in my new model I begin churning out 10,000 users/month at 200,000 users which means that is my maximum possible size and it’ll take me a long, long time to get there. By month 24 my net user growth per month has shrunk from 5,000 down to 1,500 and it’ll just keep shrinking as that growth line continues to flatten.
The core issue here is that I was growing in absolute numbers and I was churning in percentages (well the core issue for my business was that I wasn’t growing fast enough and/or I was losing too many users, the core issue for my *model* was absolute vs percentage). That was my fatal apples to oranges mistake. The reality is that I need to peg both my growth and my churn to either percentages or absolutes. We already saw the problem with pegging to absolutes yielding a shrinking churn over time which is implausible so really we should look at percentages on both sides.
So, let’s run the model again but this time, instead of 10,000 users per month starting with 100k users, let’s do 10% growth and 5% churn month over month, shall we?
Now things are starting to look a whole lot healthier. But as you breathe that sigh of relief that your model is suddenly making sense again, it dawns on you that you’re going to need to be acquiring 60% more users each month at the end of 12 months and 170% more by month 24.
Get out of here, you’ve got some work to do. But first, make sure you’re dealing with the right units to avoid crashing and burning on the far side of Mars.