Product Management is About Prediction
Part 2 — Predictive Analogues
In the previous article in this series, we discussed how waiting until you’re certain, before following a new product direction, is often too late. The opportunity is missed unless you can make reasonably accurate predictions about what your product needs to be and what it needs to solve, in the future, with enough lead time to do the work to change it to match. Because you need to act with incomplete or volatile information, a good product manager will employ techniques and methods to minimise the risk of being wrong, long before you can establish that you are assuredly right.
The use of models to forecast likely future trends is widely ridiculed by people more accustomed to dealing with hard data and concrete facts (though a substitute is never proposed — crystal balls, tea leaves, entrails?). The uncomfortable fact is that you need some kind of forecasting model, or you’ll consistently miss the market opportunities that present themselves. Whatever you choose, it has to be better than guessing or it’s worthless. Choosing not to forecast at all is worse than guessing.
When weighing up what’s likely to be required in the future, the first obstacle to avoid is the scope insensitivity error. Usually, the error we make, as product managers is to miss the fact that exponential increases in product scope (let’s say “features”, for argument’s sake) only creates a linear increase in willingness to pay. Customers don’t pay ten times as much for ten times the features; they only pay double. This is a powerful argument against scope creep. Each additional feature bolted-on earns a smaller amount of value than the last one. Similarly, to realise twice the value of the competitor’s offering, your product has to be an order of magnitude better. “Better” usually doesn’t mean “has more features”.
Scope insensitivity errors in time frames are just as deceptive. The probability of your users requiring a completely different set of product capabilities, over the next financial quarter, is far lower than the probability of them needing a lot to change over a period of years. The converse is also true. While incremental change might work over the short term, it utterly fails to keep your product relevant over the long term. The scope of the forecast is crucial. You might get lucky, if your competitors are equally slow to change and blind to time frame scope errors, but that’s not a product strategy; it’s a lottery.
There are existing mathematical algorithms to cope with moving targets. Dynamic programming seeks to move a body in motion to where its objective is likely to be, in the future, no matter what twists and turns it takes. This is how missile guidance systems often work. When I worked in digital audio research, we used dynamic programming to make edits in one spoken dialogue audio stream, so that it lip-synced perfectly with another. The algorithm ensured that nobody could easily tell where the audio cuts and splices were made, to achieve perfect alignment.
The essence of the algorithm is successive guesses, which are refined, as new information is obtained. If the initial guesses are too far from what happens in reality, the algorithm cannot catch up and consequently can fail, because time is of the essence. However, if your initial predictions are close enough, they get better with time, until the target is met by the missile.
The hockey-playing legend, Wayne Gretzky, reportedly claimed that the secret to his success was skating to where the puck is going to be. This is an interesting and vivid idea. Skating to the wrong place is disastrous, because the puck is elsewhere. You need to consider a lot of perspectives and possibilities in the heat of game play, but time is short, so you have to process the information available and decide which direction to go in a very agile way. If you get it right, you make the shot.
This applies as much to product management as to hockey. You succeed when you are able to release what the customer is going to want, by the time they want it. That means starting development before you know for certain and getting the prediction as close to accurate as you can at the outset, correcting your course, the nearer you get to release time. The best quote I could find about this is, “When the facts change, I change my mind”. Iteration is a powerful tool. Intransigence is not.
Is there a forecasting method that works reliably? In the next part of this series, we’ll examine what Amazon does to get close to a predictive method that succeeds and how they handle failure.
About the author
Michael Topic is a freelance Product Manager with over thirty years experience delivering products that didn’t exist before. He welcomes contract enquiries to define new, competitive products, design them and deliver them. His speciality is software-based products.
Disclaimer
The opinions offered in this article are intended to describe common scenarios that sometimes occur in general product management practice. They are in no way intended to be read as referring to any particular employer, past or present.
About the “Product Management is About Prediction” Series
This multi-part series examines the many ways in which effective product management has become reliant on high speed, high quality predictions of future customer needs. It is based on psychological research and experience.
Today’s winning companies are getting better at more accurately forecasting customer needs, with enough lead time to respond to them. The series also addresses some common product strategy pitfalls and how to avoid them.
