The carrot “^”

I've made my share of financial models for lots of business ideas but I never found myself having to use a "^" to forecast revenue growth. It's not that I don't like the ^ it's just that for my past businesses the math for revenue never really seemed to call for its use. Moreover, I think you have to be pretty confident to use it and hope to be taken seriously. Well my current business the math called for its use – and I cautiously applied it to the model. If you look at the growth rates of social web sites and fads they both have exponential-like growth rates (early on) when viewed in the read view mirror; predicting these growth rates before launch is impossible and makes any modeler look insanely optimistic. The problem with financial models for startups is they are only good at estimating expenses – revenue never matches what you plan. I looked back at our model for Link-VTC (1993) and within the first 12 months we exceeded our 5-year plan for revenue (we weren't that ambitious apparently). For Raindance we always seemed to be shy of our forecast but did well anyways. This new gig I can't deny the math behind our sales model so the ^ is in. While I'm trying to appear safe and sane with our forecast I'm more aware of how fragile these types of businesses are to certain market factors.

In simplistic terms, a function for viral growth is similar for product fads and social sites: you have X users at the start, each user on average tells at least Y friends about the product, you hope that r% of people (Y) convert to a user/buyer so that rY <= Y (if rY < 0 you got problems), and then the cycle repeats every n days. The faster the cycle the lower number you can model for r. Every new user then is added to your X to start the process over. And then if you want to get sophisticated you model the decay in r and eventually the decay in X. In a model, you need to make assumptions for every number: X at start, Y how many friends they tell, r for the rate friends convert to users/buyers, n how fast the cycle repeats, and how all these numbers change over time. A slight change in one direction or the other for any number and it can send your model to moon or plummeting to the ground in a nasty fireball. It's hard to manage such sensitivity in a business model and that is probably why most people find it difficult to believe until examining it after the fact. Nevertheless, a revenue/growth model, while maybe not perfect upfront, helps identify the key factors that drive a viral growth engine. It is also a useful exercise to understand how all the elements of your business interact - even though you're probably wrong about something. Here is an interesting analysis of MySpace. The exponential growth is only for a short period then it changes – which is true for most successful "viral growth" companies.

Comments

[Rob Biedenharn]

The type of modeling that you describe is "System Dynamics" and can be learned in an intense one-week course from the Sloan School of Management at MIT. http://mitsloan.mit.edu/dynamics

I worked for Prof. Sterman some 20 years ago and was able to sit in on the course for free (it was only about 1/5 the price then, but still more that I would have ever been able to afford as an undergrad).