Escaping the Roadmap Trap

If you're interested in the mathematical background, I can highly recommend: https://medium.com/@kentbeck_7670/decisions-decisions-or-why...

This article proposes several alternatives to feature driven roadmaps.

But it ignores the main reason that roadmaps exist: as communication documents to the rest of the organization. This is what I used roadmaps for as a PM for a decade, but it was driven home after talking to hundreds of PMs after founding my fourth company[1] and coaching PMs and product/dev teams to level up for another decade[2].

The real roadmap trap in my experience is that the org takes a roadmap as a promise of when specific things will be delivered. The best approach to avoid this is to not do roadmaps. But that’s also unrealistic for most orgs. The next best approach is to show the org short-term work that will definitely get done, and caveat the heck out of anything longer term. I’ll happily commit to 6-8 weeks out, tentatively commit to 10-20 weeks, and add large asterisks to anything longer than that.

1- https://www.savio.io

2- https://www.Reemer.com

Ever since I worked in multi stage stochastic optimization I think of strategies in terms of decisions at each nodes of a tree of scenarios. Basically you want to optimize on the long term objective function, and take decisions at the optimal time based on the cost of indecision and the estimated benefits of having more information. Having probabilities associated to each branching in the tree of scenario also makes it have baked in risk management. This is not so easy to apply in practice but it is worth doing and allows having a much longer horizon: else how could you make a strategy with decisions in 6 months not being dependent on the results of the decisions you take now ? As another comment pointed out, the main goal of product roadmap is communication, strategy is something else. In this case the roadmap could be the most likely path to take in the scenario/decision tree, with diminishing confidence over time based on the probabilities observed.

Multi stage stochastic optimization is often used in finance/trading (e.g. portfolio optimization), I highly recommend to check it out.

Interestingly there is no product on the markt to collect feature ideas and classify them. Kanban boards etc. Are nice, but if you store many ideas you will completely lose focus.

Ideas, which may or may not be implemented one day have so many dimensions: time required to build it, dependencies which must be built first, usefullness in several aspects, module the software needs to be implemented in and interacts with.

The best i have yet seen was a good crafted excel/airtable with a lot of columns. But deciding which one to build was again not easy as you could not see dependency chains etc.

> After all you don't want your employees telling customers something's coming promising a future that may not come.

Every companies do that all the time. And there is no way to prevent that. It is impossible to predict 100% accurately things that you are going to build in 6 month