Only three or four out of every ten movies made in America breaks even or earns a profit. Yet the deci­sion to green-​​light a project is usu­ally based solely on “expert opin­ions” — in other words, exec­u­tives’ intu­ition sup­ple­mented by stan­dard regres­sion analysis. There’s got to be a better way.

We think we’ve found one. In a recent study, two of us (Dan and Carmina) used a tech­nique called “sim­i­larity based fore­casting” to pre­dict box office rev­enues for 19 wide-​​release movies. Here’s how it worked. Non-​​expert movie-​​goers were asked via online sur­veys to judge how sim­ilar each movie was to other, pre­vi­ously released movies, on the basis of a brief sum­mary of the plot, stars, and other salient fea­tures. We then fore­cast the rev­enues for the new movies by taking similarity-​​based weighted aver­ages of the pre­vi­ously released movies’ rev­enues. On average, those pre­dic­tions were twice as accu­rate as ones driven by expert opinion and stan­dard regres­sion fore­casting. They were par­tic­u­larly good at iden­ti­fying small revenue-​​earning movies. This type of case-​​based deci­sion analysis is a great way to tap into crowd wisdom.

It’s impos­sible to elim­i­nate risk from strategic deci­sion making, of course. But it is pos­sible to sig­nif­i­cantly improve your odds by under­standing which decision-​​support tools work best for which deci­sions. Most com­pa­nies – including the movie stu­dios in Hol­ly­wood – over-​​rely on basic tools like dis­counted cash flow and net present value. These tools are great if you’re working in a stable envi­ron­ment, with a busi­ness model you under­stand. But if you’re on unfa­miliar ground – if you’re in a fast-​​changing industry, launching a new product, or shifting to a new busi­ness model – they can be down­right dangerous.

Read the article at Harvard Business Review →