You Shipped an MVP. Now What? From Idea List to Roadmap
7 min readPublished By Ofer Regev
Shipping the MVP was the exciting part. The harder part comes next: you have a list of ideas, a few users, and no obvious order of operations. Without a roadmap, you ship whatever's fun to build, the loudest request wins, and momentum burns fast with no clear direction.
Here's a way through it in five moves.
A roadmap isn't a wishlist
Backlog
Roadmap
Everything you might build someday
Every item ties to a real goal
No order, no timing
Order and timing are explicit
Nothing says why it matters
Choosing X means not Y, right now
Grows forever
Small enough to explain in one sentence
1. Prioritize
Four methods, from rigorous to simple.
RICE scores each idea: (Reach × Impact × Confidence) ÷ Effort. Reach is how many people over what timeframe, impact runs on a 0.25-3 scale, confidence is a percentage, and effort is in person-months. It forces you to argue with numbers instead of opinions, and works best when you have real usage data.
MoSCoW sorts a single release into Must, Should, Could and Won't (this time). It forces the trade-off conversation up front. Watch out: everything drifts toward "Must." A useful cap is keeping Musts to around 60% of total effort.
Now / Next / Later uses buckets, not scores. Now is actively building and ships in weeks. Next is roughly scoped. Later is a real idea not yet detailed. It's the fastest way to communicate with stakeholders, but it tells you when, not why, so pair each item with a reason.
Kano sorts by delight versus must-have: basics (expected, punished if missing), performance (more is better), and delighters (unexpected, drives loyalty). Every delighter eventually becomes a basic. Pair it with a real survey, not intuition.
Which one?
Under about ten ideas, any method works, and Now/Next/Later is usually enough. With a long list, scoring forces real differentiation: RICE if you have usage data, MoSCoW if you don't but need structure. Re-sort every few weeks as you learn.
2. Ideate (and plot value against risk)
Ideas aren't only user requests. Reactive ideas are what users tell you: feature requests, support tickets, "it would be great if." Proactive ideas are what you notice: where people drop off, gaps competitors left, your own instincts as a builder.
Then plot them on a value-versus-risk grid:
Quick wins (high value, low risk): do these first.
Big bets (high value, high risk): worth the gamble, one at a time.
Fill-ins (low value, low risk): whenever.
Money pits (low value, high risk): cut them.
For a habit tracker, daily streak reminders and a weekly progress email are quick wins. Friend challenges are a big bet. Dark mode is a fill-in. A custom AI coaching chatbot nobody asked for is a money pit.
3. Stay lean
The build-measure-learn loop doesn't stop at the MVP. The difference now is that you have real users and data instead of guesses.
Shrink it to days. Pick the one flow that tests the idea and cut every edge case. For "friend challenges," the full version needs invites, leaderboards, notifications and privacy. The test version is two friends and one shared streak. It ships in days and answers "do people care?" Expand only if the answer is yes.
Pivot or persevere? Signs to persevere: users come back unprompted, feedback gets specific rather than polite, someone asks to pay or asks when it's ready. Signs to pivot: usage fades once novelty wears off, feedback stays polite but flat, and you're the only one still excited.
4. Check product-market fit
The question is whether users are using it or needing it. People who use it tried it and maybe liked it. People who need it would be hurt if it disappeared, and tell others without being asked.
The 40% test. Ask people who've actually used the product how they'd feel if it went away: very, somewhat or not disappointed. If 40% or more say "very disappointed," that's a real signal. Don't survey new signups, since they skew the result. It's a gut check, not gospel, so pair it with usage data.
If you're not there yet, that's information. Don't scale go-to-market spend to compensate, and don't add features hoping one sticks. Go back through build-measure-learn and talk to the users who stayed.
5. Go to market
You need 50 people, not a market. One real conversation beats a hundred impressions. Test where you're already trusted: your network, communities you're part of, an audience you've built. Skip cold ads and generic SEO for now. The rule is one channel, two weeks, real signal, and a second channel only if the first works.
A one-page GTM plan has three lines:
Channel: a habit-building Reddit community.
Message: "Streaks that don't reset when life happens."
Success signal: 20 signups in two weeks.
Do it out loud
List your top five or six ideas.
Sort them with Now / Next / Later.
Plot two or three on the value/risk grid.
Pick one quick win to start this week.
Say it out loud to someone. Saying it makes it real, and you leave with a plan instead of notes.