Balkrishna Attal

Blog / Product Management

Product ManagementAugust 14, 2026·6 min read

How I Decide What Not to Build

Every roadmap has ten good ideas and room for two. Here's the filter I actually use — not the tidy, framework-shaped version I'd give if you asked me in an interview.

— views
Idea A Not now Idea B Not now Idea C This quarter

A few months ago I sat in a planning meeting where three people, in the space of twenty minutes, each made a genuinely good case for a genuinely good idea. One wanted daily reminder notifications, tuned better than the ones we already had. One wanted a friend leaderboard, because a competitor had just shipped one and early numbers looked promising. One wanted to fix a broken step in onboarding that was quietly losing us a fifth of new signups before they ever saw the product.

We had capacity for one of the three, for one quarter. All three people were right that their idea would help. None of them were wrong. That's the part nobody tells you about prioritization when you're starting out — the hard cases are never "good idea versus bad idea." They're good idea versus good idea versus good idea, and you have to pick.

The question I stopped asking

For a long time my first question was "is this a good idea." It's the wrong first question, because almost everything pitched to you by a competent team is, in isolation, a good idea. The question that actually narrows things down is different: what does building this replace?

Every yes is a no to something else, even when nobody says the second part out loud. The leaderboard wasn't just "should we build a leaderboard" — it was "should we spend six weeks of engineering time on a leaderboard instead of six weeks fixing the onboarding leak." Framed that way, in that meeting, it stopped being close.

Three questions, in order

Here's the actual sequence I run through now, roughly in this order:

1. What breaks if we don't build this?

Not "what improves" — what actually breaks. The onboarding fix passed this immediately: a fifth of new users were leaving before activation, every single week, and that number wasn't moving on its own. The leaderboard failed it cleanly. Nothing was breaking without it. Things might get better with it, eventually, maybe — but nothing was actively bleeding.

2. Who specifically asked, and how many others have the same problem silently?

The reminder notification idea came from one enthusiastic stakeholder who'd seen one competitor's changelog. The onboarding fix came from a support ticket pattern — the same failure, reported by different users, in their own words, week after week, none of them coordinating with each other. One loud voice and one quiet pattern are not the same evidence, even when the loud voice is more fun to listen to.

3. What's the two-week version?

If an idea can't produce something testable in two weeks, I want to know why, because it usually means nobody has actually broken the problem down yet — they've pitched a destination, not a first step. The onboarding fix had an obvious two-week version: fix the one screen where people were actually dropping off, measure it, then decide whether to go further. The leaderboard's two-week version was "build a database schema for a leaderboard," which tells you nothing about whether anyone will use it.

None of this is about being right. It's about making the trade-off visible enough that a room full of smart, well-intentioned people can disagree about it on purpose, instead of drifting toward whoever pitched last.

A civil engineering habit that stuck

Before product, I spent two years on construction sites, tracking budgets and schedules. Every rupee spent gold-plating a fitting that nobody would ever see was a rupee not spent reinforcing something load-bearing. Nobody ever regretted the boring, structural spend a year later. Plenty of projects regretted the decorative one. Software has the same shape, just with attention and engineering time standing in for the budget line.

What we actually built that quarter

We fixed onboarding. Activation went up meaningfully within the first month, and it was one of the more boring launches we'd shipped in a while — no press release, no announcement banner, just a number that stopped leaking. The leaderboard came back into the conversation two quarters later, once we had actual evidence — a support pattern of its own — that people wanted a social, competitive layer. That's usually how it goes: the idea wasn't wrong, it was just early, and it needed its own evidence instead of borrowing someone else's.

If you're staring at your own version of that meeting right now — three good ideas, room for one — that's exactly the kind of hour I try to be useful for.

Stuck on something like this yourself?

I take on a small number of 1:1 calls, career conversations, and resume reviews each month.

See how we could work together
BA

Balkrishna Attal

A product leader with a background in civil engineering. Writes about product decisions, careers, and moving into product from somewhere else.