One of the beautiful exercises all companies go through after reaching a certain size is forecasting: the art of predicting how the company will do a month, a quarter, a year from now. Despite the guarantee that the outcome will vary (in many cases, wildly) from the model, this exercise is necessary for planning purposes and represents something the company as a whole can work from.
Presumably, your goal is to find an accurate forecast, so I'll go through some modeling principles and share a reasonable structure for a good outcome. My favorite quote is transparent about this: “All models are wrong, some are useful.”
I spent a number of years on a “quant” team on Wall Street. One of the most important things I learned was the benefit of simplicity in human-built models oriented around the future. You should strive to have as few variables in your model as possible. The variables you do have, make sure you have a fundamental reason for having.
Let’s say you want to predict the temperature tomorrow. How would you do it? You don’t have 10 million dollars in cloud credits, just an excel spreadsheet. No PhD in meteorology, just your wits. A decent place to start would be to look at the temperature today and predict that our temperature tomorrow will be simply that. Magic! We now have our first model. We’ve established what’s known in statistics as “autocorrelation” principle—that is, the timing of what you’re predicting matters. This is extremely relevant to corporate forecasting. In fact, if you take away nothing else, understand that autocorrelation is the fundamental principle that is applied throughout forecasting. It’s applied to every component of the process.
We know that the temperature won’t be the same every day. How can we predict the change in temperature? We can look at the change in temperature and simply predict that forward, but we have a problem. We also know that temperatures can fluctuate. We can zoom out and look at the average temperature from 60 days ago, 30 days ago, and then compare that to the average temperature 30 days ago to now. That difference, that trend, would be a good estimate of how the temperature is changing over time. So our new model is the temperature yesterday plus the “trend”. We can apply this directly to companies. How should we predict the future? Well, what revenue did we do last year, and what percentage did we grow last year? Good enough and have a wonderful Christmas.
Honestly, this isn’t a terrible outcome. But let’s see if we can do a bit better. Let’s go one level deeper and take into account knowledge of how the weather works. We don’t even need to have any knowledge. Let’s pretend we were aliens looking down upon New York. (Aliens always love New York, must be the pizza.) They would see the data and notice that the temperature seemed higher during a specific time period, and lower during another specific time period every year (if they’ve been hanging around, which, again, pizza). If they wanted to predict the temperature for December in July, they wouldn’t simply extrapolate forward how much hotter it’s getting. Instead they would note that there’s a repeatable trend and account for it. And that’s what we shall do as well. Now we’re introducing a concept called seasonality. Things happen, not just in a linear fashion, but in a repeatable and predictable way. Modeling a few of these concepts can get you to a decent place.Take into account how you’ve done in the past, look at time of year and perhaps one other factor, call it good enough and move on with your busy workday. And yet, despite these simple concepts, companies—and in particular startups—don’t come close to getting anything right.
So why are companies—in particular, startups—bad at forecasting? There are a few answers. Perhaps the most direct one is that forecasting is simply hard. Not only that, but everyone is aware of the forecast. It’s not an isolated process, and as a result, forecasting is subject to the push and pull of multiple people, processes and politics.
In the broad corporate planning process, forecasting is simply one of an enormous number of decisions. If there was an easy way to predict the future, everyone would know it. The very best forecasters in the world operate on probability likelihoods, but that doesn’t work for corporate purposes. Corporations need to put a line in the sand in order to plan. This is how much revenue we’re going to hit, therefore this is how much budget each department has to spend. Most humans struggle keeping multiple scenarios in their heads. They want a number. And so, a number must be given.
Startups are notoriously difficult. When the rates of change are so drastically accelerated, it’s hard to have confidence in any longer term outcomes. There are many ways to achieve strong growth early on because the dollar amounts are so small, but few of them are sustainable enough to result in a continual growth trajectory. Projecting exponential growth is scary. Predicting a decay in a previously exponential growth path is, in some ways, worse. Ultimately, startup forecasts become in many ways more political calculations than best projections of truth.
The forecasting session is built on pain tolerance. It’s a bull rush, a waving of the red flag. Who blinks first when the silence draws thick as the CEO asks who will push the numbers to their proper place? It’s an act of high-stakes hot potato that will determine the course of the year. Who can convince the other to take on maximum risk for minimum gain while keeping the inverse for themselves? Gambit, and counter gambit.
“It seems like the sales team is only covering half of the required goal. Given their compensation structure, perhaps it would make sense for them to commit to a deeper top-line.”
“Sales would be happy to commit to the full top-line if marketing signs up for a full increase in leads. Sales qualified, of course.”
“Marketing can certainly hit lead goals as long as we qualify on the marketing level. We’ve seen sales be a bit, unfortunate, in their judging criteria when they’ve had issues closing leads.”
And around and around it goes.
The saying goes that everything is politics, but it’s never more true than when you have large scale dollars, livelihoods and reputations at stake. One of the deepest and most powerful indicators of a strong executive is someone who can “hit their numbers.” Show me an executive that hits their numbers, and I’ll show an executive that’s really, really good at setting those numbers. The first thing you learn as an executive is how to set your team, but most importantly yourself, up for success. No one wants to be seen as not a team player, but there’s a certain special skill set in those who can lay contingencies upon their plans and lay excuses that aren’t seen as excuses well in advance. If one can make the goal hard to achieve, they win in both scenarios. Miss, and the groundwork has been laid well in advance. Hit it, and be seen as a hero who overcame impossible odds to do so. The smart executive is consistently evaluating these things. Everyone wants to play heads I win, tails you lose—only a sucker takes a real honest bet.
So what are you left with after all this? If the game is known and the players are dealt a fair playing hand, you’ll end up with a decent forecast, all things considered. After all, it’s the big leagues. No one makes it without leaving a fair number of corpses in their wake. In startups, though? Bigger issues arise. When you have an uneven playing field, each headcount actually matters. And, the game is not well-known to the players, particularly not to the OG true believers who are just now encountering the corporate beasts known as planning sessions.
This is where the Founder/CEO needs to step in and be strong. Lay out the ground rules. Go in knowing exactly what you’re all there to discuss, and set the expected result of a solid answer that follows from your methodology. Make sure conversations are structured such that the people who matter are in them and the ones who don’t, aren’t. Even the playing field to the extent you can, and above all else, arrive at an outcome.
Be crystal clear in your expectations, and call out each and every member of your team as the sessions go on. One of the foundational solutions to a Prisoner’s dilemma is extra situational enforcement.
Fundamentally, the planning and forecasting process is a CEO exercise, and it should be treated as one. Until IPO glory, the CEO needs to be strongly hands-on across key pieces of the company. Forecasting is most certainly one of those. Lead the process, and hold yourself accountable for the results.