Arguably 2019 will go down as the year the fake burger went mainstream. And Impossible Foods has a lot to do with it. The buzzy Redwood City, California-based startup served up its plant-based patties in more than 17,000 locations around the country, including restaurants like White Castle, Red Robin, Burger King, and more, and launched in grocery stores for the first time. More importantly, the startup's faux burgers beg some big questions about the future of meat, which is one of the reasons why it is Inc.'s Company of the Year. Getting to that point, however, has come with gnarly challenges, fierce competition, and a fair amount of improvisation, as Inc. editor-at-large Burt Helm explains in his feature for the winter issue of the magazine. Below, founder and CEO Pat Brown explains why his fast-growing startup gave up annual planning--and the unorthodox method it uses instead.--As told to Guadalupe Gonzalez

The thing about the foodservice business is that the flow of information from the customer to us is treacherous and unreliable. We don't sell directly to restaurants. We don't always know how many restaurants are serving our product. Basically, we ship to an uber distributor that then takes the stuff, divides it into smaller lots, and ships to hundreds of local and regional distributors that deliver the product to restaurants. It's a quite complicated and semi-ridiculous system. Demand signals sort of trickle back to us.

Leading up to our product shortage earlier this year, we had not an immediate but [rather] rapidly dawning awareness that the inventories were dwindling, and we didn't have the capacity at the time to replenish them. It took a couple of months before we could really get our teeth in supply again. It's not so easy to massively ramp up production from this standing start.

We took this experience as a heads up that we need to do a much better job of anticipating and planning for surges in demand moving forward. We've categorically changed the way that we forecast and manage our supply chain compared to six months ago. What we had been doing--and what I believe seems to be the norm for most businesses--is that we were using a completely deterministic model for how things would grow: it was just a line supposedly showing demand over time.

That's complete bullshit. I've always felt that way. When we launched in Burger King, our sales were three times what they had anticipated. Burger King knew better than anyone how to guess, but the fact is you just don't know until you do it.

You have to model it probabilistically. We're now using what's called the Monte Carlo method, which is a way of running models where many of the parameters that predict demand have an uncertainty associated with them. We run a thousand trajectories and in each one we randomly sample from the range of the probabilities in different events.

A simplified way of looking at it is that we want to have a plan for production capacity to be able to support either the first percentile or the 95th percentile of demand. That obviously entails some risk, but the reason it's not that risky is that given our rate of growth, and the strong demand signals that we're seeing, we are confident that if we are not at the 95th percentile by a given month, we will be there soon. The last thing we want, and we made this mistake earlier in the year when we couldn't keep up with demand, is to plan for a conservative growth rate, and as result, not be able to fulfill it because we haven't built the capacity.

The opportunity-cost of not planning for success vastly outweighs the risk of building capacity prematurely. It's a completely different way of thinking about the business and running it. We don't have any certainty where we are going to land in that spectrum, so we better be prepared for anywhere. It hasn't gone down easy.

When looking at our projections for retail sales, it's the exact same thing. There are tens of thousands of grocery stores that are interested in having our products in their shelves. We can make estimates with their help, but the fact of the matter is--it's just a guess. It could be anywhere in this very wide range and we need to have a plan that works wherever it lands.

Rather than doing an annual re-planning of the business, we're doing it monthly, because we change more in a month than most food companies change in five years. We have to act like it. We've already seen what happens when you let demand get ahead of your production.