How to build a good ideas machine
Product isn't about having good ideas; it's about creating a system that tells you which ideas are the good ones.
Your job is not to have good ideas.
If you’re a product person on a team of any size or stage, or a leader at an early-stage startup fumbling in the dark for product market fit, this probably applies to you.
A company is a machine that produces a particular kind of value for a particular customer. In the realm of software, the inputs to this machine are “ideas for things we could build” and the outputs are tools and software that, ideally, your users use to get value.
Now, if all your inputs into this machine were ‘good’ ideas that would create immense value for your users, your job is really easy. Just have a brainstorming session, pick the idea that everyone agrees is the best, and then build that. It’s pretty straightforward.
That’s the point here; if it were that easy to know which ideas are the good ones, many of us would be out of a job. Because your job is not to have good ideas. Your job is to build a machine that tells you which ideas are the good ones, and does so as quickly and efficiently as possible.
The first step in building this machine is to make sure your team knows which machine you’re building. If a machine is defined by the product it creates, you must align your team around the problems you are trying to solve, and thus the value you are trying to create. Ideas are really just solutions to problems, so without alignment on the problem, you’re going to get a wild variety of solutions. If you can refine your problem down to improving a specific metric, or a well-defined user problem that you’ve observed, that will serve you best.
Your machine should be geared towards weighing each idea against the highest level of evidence as quickly as reasonably possible. How do you rank your levels of evidence? It’s actually straightforward: gut feeling & managerial approval count for very little, while empirical experimental user data and MVP results count for a great deal more (Itamar Galad writes here about scoring your evidence). I’m not advocating for a machine that launches a fleet of MVPs, but instead one that selects for the best-supported ideas and pushes them one step a time down the evidentiary path, ruthlessly trimming at every turn to ensure you remain focused on testing only the most promising concepts.
When you’re thinking about this evidentiary path, it can change how you think about validating ideas at every step. You don’t necessarily have to prove that each idea is a good one. Rather you must only prove at each step that this idea has more evidence for its ultimate success than the other ones you are considering. This can help you design smaller tests that are faster to run and more tightly contained in terms of what success looks like. Smaller tests help you iterate faster, which really just means you’re learning faster about what your users want and don’t want.
The larger point here is that process is not a vice. For any product person, process is among your most potent tools. No amount of user research, GANTT roadmapping, or robust product specs will save you if your team is focused on the wrong problems, or falls in love with solutions, or builds things because an investor thinks it’s a good idea. Every idea must go through the same machine. You can create immense value for your users and your company through process – through building and tuning your machine.
Your machine can help you ensure:
Everyone is having ideas about the right things. You have laid out the most important problems to solve, in as much specificity as you can.
Your ideas can quickly move up the evidentiary path. You’re constantly optimizing for getting the highest level of evidence with the least amount of effort, and preserving your focus by eliminating ideas that are less promising (limiting your work-in-progress).
Your experiments deliver clear results as quickly as possible. You’re only targeting the next level of validation, thus keeping your tests tight and interpretable.
A well oiled machine that does these things will help you:
Rebuff HIPPOs and the other dangerous animals of product management. The people in the room with fancier titles are allowed to have ideas, but they go in the same bucket as everyone else’s.
Avoid falling in love with your own solutions. It may be really cool to build that widget, and everyone on the team is excited about it, but if it doesn’t pass the evidence test it doesn’t get built.
Avoid overcommitting. If your test plan is focused only on the next level of evidence, you reduce your risk of building big things that nobody wants.
Sometimes things are clearly bad ideas, like online shopping after midnight. And some things are just plain good ideas, like not taking that second slice of cheesecake. Sadly, most of the choices we have to make while building products fall into the chasm of uncertainty and insufficient information. That’s why we build our machines: to counterbalance our biases, impatience, exuberance, and company politics, while helping us understand where to focus on getting more information. A well-oiled product machine limits risk by stripping away the less-good ideas at every step, ensuring that by the time you ship a full product, it’s probably a good idea.