Keeping It Simple: Unit Economics

coinsBack in November, I wrote about financial modeling and wanted to pick up the topic again. I think this is very important for lean startups.  A financial model focuses your brain on key goals and assumptions. It’s great to obsess over product-market fit, but you also need to take the time to make sure your economics are reasonable as well.

When you are just starting, a fully-fledged financial model of your customer growth, revenue, and costs can feel daunting, so sometimes it is most useful to break the process into smaller tasks. For example, don’t stare at a giant spreadsheet, but rather sit down with your co-founder(s) and a piece of paper and talk through how and why you imagine headcount growing.

For revenue estimates, try starting with a simple “unit economics” model, i.e. take your primary revenue model and lay out the revenue drivers and results for a single customer. The purpose is twofold:

  1. detail the key levers for the business
  2. answer this question: how much revenue will we get every month for every active user we are able to recruit and keep? Or, in acronym form, what is our ARPU – average monthly revenue per (active) user.

In our case, I built a simple calculation with 4 inputs:

Every month, for every 1 user:

  • N% will spend money
  • They will buy O# number of item(s)
  • The average transaction size will be P$
  • Our piece of the transaction will be Q%
  • Thus, ARPU = N*O*P*Q

This ARPU figure is simplistic, but it allows us (and potential investors) to see at a glance how big we will be at 10 users or 10 million.

Now, I think most freemium products have a different ARPU in year 1 compared to year 3.  It could be because your product improves, but I remember looking at Evernote’s numbers and they had this massive tail of revenue sweeping in because of the combination of growth rate, cohorts converting over a period of months, and high retention of subscribers.  If the former, this simple approach will suffice. If your business is closer to the latter, you probably do need a monthly model to really show the effects.

Research, Guesstimates and Gossip
Where does one get the input numbers? Look for metrics from companies with similar business models and customer buying experiences – the closer the comparable the better. Then it’s your job to research, make smart guesses, and reach out and talk to people. If you’re lucky, you’ll be able to dig up articles where similar companies talk about subscriptions / virtual goods purchases / ecommerce conversions, etc etc.  Talk to your peers and network to entrepreneurs who have a similar business model. See if a comparable company recently spoke at a conference (executives are often more forthcoming about numbers during a conference speech because they know the audience is hungry for it). VCs/angels are also a great source of data because they see so *many* data points.

Side note on VCs: in a meeting, don’t be afraid to ask a VC about the data points they have seen, but make sure you’ve done your homework first. Many VCs won’t say “those numbers look wrong” — they’ll just quietly dismiss you as ignorant or naive.  It’s better to ask, and if there is a mismatch on numbers, you’ll both learn new data points (which is good!) and have a chance to demonstrate both your mental flexibility and the amount of thought you have already put into the numbers.

Example of a Guesstimate
We have one company roughly in our space that doesn’t share a lot of information publicly. Still, by reviewing their site and blog, using their service, digging into media/blog articles, plus a smidge of industry gossip, I was able to make reasonable guesses as to the number of employees, number of users (I found out from a VC friend that my first estimate was low), and their average purchase size. I also knew that they were profitable, having only taken a little angel money at the start several years prior.

To see if my own ARPU figures were reasonable, I estimated their total costs per employee at $16K — 14K to 16K seems to be considered “normal” for an Internet company.  The calculations were as follows:

  • Monthly Revenue = (# of employees * 16K) * 1.15     [note: the 1.15 multiplier adds in a 15% profit margin, which I bet is actually low].
  • Active Users = Total Users * 75%       [note: I wanted to discount the total user estimate to clean out mostly-inactive users; in some cases, this kind of  discount will be much higher]
  • ARPU = Monthly Revenue / Active Users

Base vs Best
Once I had my simple unit economics model, I plugged in really crappy numbers to see how bad the revenue picture might be while our app was crude and simplistic. Then I made a “target” version with better numbers that we believed we could hit (based on real examples) and could justify to an investor.

Converting to a Monthly Model
You can use your simple unit economics calculation as the revenue “engine” for your 3-year financial model. All you will be doing is plugging in the number of active users that month at the top of the engine. Rather than working with a single (i.e. static) set of assumptions for your key inputs (in my first example above, the key inputs are N, O, P, Q), build your model with so that every month has it’s own assumptions. This way you can model out changes over time, such as the transition from Base to Best.

By the way, when I create a new business model, I usually start with a clean slate for the top line calculations because assumptions and drivers are so business- and product-specific. If you use an existing model, just clean out everything to do with revenue (possibly even user acquisition too) so you aren’t trying to retrofit someone else’s business onto yours.

Remember, a huge reason for doing all this is going through the exercise of putting the model together, not religiously believing in the end results, so don’t shortcut it!

Avoid Over-complication
As you are going through this exercise, you will have to decide for yourself whether it is worth going into more detail for your full financial model. For example, in past I have modeled out assumptions across different user types because we wanted to guess and track virtual goods transactions for casual players versus hard-core players. However there is a tradeoff here — any sophisticated modeler knows that you can make a spreadsheet do whatever you want. More detail doesn’t always mean more utility or more truth.

Focus on your *key* metrics, i.e. the things you will actually track as a business. Ideally this model becomes a living document where you can track actual results.

More on metrics and models

Final Notes

  • Obviously my examples are all around consumer-facing Internet companies.  That’s what I’m currently working on, so most of the posts on this blog these days are written from within that context.
  • If you are struggling to create a financial model, go simple and break it down into smaller pieces.  Get away from the spreadsheet and think through things with your partner or just a pad of paper.
  • Lastly, I want to stress that the exercise of creating a financial model is incredibly important, NOT because the resulting numbers are sacrosanct, but because it will sharpen your mind, expose gaps in your business, and provide you with a useful tool when paired with your analytics for monitoring your business on an ongoing basis.  Like everything in a lean startup, this should be iterative and constantly tested and improved.