all notes
2026-08-20Naman Barkiya

4 builds, 4 cuts: what we didn't ship, and why each cut held.

Across four shipped builds — ONETAPP, SIT Manager, Tirth, and HospiHealth — we cut a mobile app, a leaderboard and API, a mid-build feature, a full booking platform, and a marketplace, all before or during the build rather than after launch. None of the four cuts has been reversed; the discipline is timing the cut before the feature is built, not deciding whether to build it at all.

Across ONETAPP, SIT Manager, Tirth, and HospiHealth, four features got cut in week one — before a line of code existed. None have been rebuilt since. Here's the receipts.

Four shipped builds, four features cut in week one, zero of them rebuilt since. Across ONETAPP, SIT Manager, Tirth, and HospiHealth, the pattern is the same: the cut wasn't a compromise made under deadline pressure — it was a decision made before a line of code existed, and every one of those decisions held.

We publish build postmortems one at a time. This one is different — it's the cross-section. Four named builds, four concrete things we didn't ship, and the reasoning that survived contact with real users. If you're scoping your own first version, these are the receipts.

1. ONETAPP: cut the mobile app, the leaderboard, and the publisher API

ONETAPP is a gaming rewards platform — Electron desktop app, landing site, user panel, admin dashboard — built from a one-line brief to a live cohort of thirty-two users in ten weeks. The original brief assumed mobile. We spent three days of week one arguing it out of scope: gamers sit at a desk with Steam open, and a background app competing for a phone screen they weren't looking at made no sense.

Three things got cut in that same week-one session: the mobile companion app, a public social leaderboard, and a publisher API for game studios to add challenges directly. None were cut because they were bad ideas. They were cut because none had users yet to validate them — a leaderboard with forty entries reads as abandoned, and a publisher API with zero publishers is a maintenance burden with no one using it.

2. SIT Manager: cut real-time notifications, mid-migration

SIT Manager replaced eight years of fragile legacy PHP for Suvaidya Institute — two prior agencies had stalled for nine and six months respectively; we shipped a working first version in fourteen days and the full system in under eight weeks. The cut here wasn't a feature struck from the original spec. It was scope we let creep in and then pulled back out.

We introduced real-time notifications in week six, mid-migration, and it added risk during the most fragile part of the engagement — the cutover from a fifteen-month-stalled legacy system was the actual product at that point, and notifications weren't it. The lesson wasn't "don't build notifications." It was: don't add scope during a cutover, no matter how small the feature looks.

3. Tirth: cut the booking platform, kept the trust layer

Tirth is a platform for religious travel across India's faiths, a market running almost entirely on WhatsApp, cash, and a trust chain between priest, guide, and pilgrim. After three weeks of research — forty pilgrims, twelve guides, six temple-side operators — the obvious build was a booking platform: inventory, payments, reviews, listings. Every prior attempt in that market had built exactly that. Every prior attempt had died inside two years.

We cut the booking platform before we built it. Tirth shipped as a guide-first directory instead — named, vetted guides with testimonials, ending in a WhatsApp handoff, not a closed transaction. The guide network grew 4x in the first quarter through guide-to-guide referral, which doesn't happen inside a stripped-down transaction layer.

4. HospiHealth: cut the marketplace, built custom admin instead of a SaaS tool

HospiHealth needed a founder-story page, a jobs portal, and an admin dashboard for a hospital consulting firm — eight weeks, brief to live. The original brief wanted a job-board marketplace: public listings, candidate profiles, a matching interface. We cut it in week one. A marketplace needs two-sided volume to read as alive, and a new firm with one employer and no candidate base yet would launch a marketplace with five listings and zero applicants — worse than no marketplace at all.

The cut saved six weeks. The jobs portal that shipped instead — a structured intake form — has since taken in forty-three applications across three roles, all through a custom admin we built instead of routing sensitive CVs and salary data through a third-party SaaS tool.

What holds across all four

BuildWhat got cutWhyResult
ONETAPPMobile app, leaderboard, publisher APINo users yet to validate any of the threeShipped in 10 weeks; none rebuilt since
SIT ManagerReal-time notifications (mid-build)Added risk during the cutover, the actual product momentFull migration in under 8 weeks
TirthThe booking platform itselfMarket runs on trust, not transactionsGuide network grew 4x in Q1 via referral
HospiHealthJob-board marketplaceNo volume yet to make two-sided matching workSaved 6 weeks; intake form got 43 real applications

The common thread isn't "cut aggressively." It's timing. Every cut above happened before the feature was built, not after it shipped and failed to get used. A cut made in a scoping session costs an afternoon of disagreement. The same cut made six weeks into a build costs six weeks, plus the maintenance tax on code nobody asked for. We wrote the general version of this rule in the three-question test we run on every feature before it ships — these four builds are what happens when you actually run it.

None of these features were wrong ideas. ONETAPP's publisher API and mobile app are both real roadmap items, just not week-one ones. That's the whole discipline: not "no," but "not yet, and here's the evidence we're waiting on."

Written 2026-08-20 by Naman Barkiya.

FAQ

Questions this usually surfaces.

What's the most common feature that gets cut from an MVP?
Anything that needs two-sided volume to work before you have users — marketplaces, public leaderboards, and matching interfaces. HospiHealth cut a job-board marketplace in week one because a new firm with one employer and zero candidates would launch with five listings and no applicants, which reads worse than no marketplace at all. ONETAPP cut a social leaderboard for the same reason.
Is it better to cut a feature before or after launch?
Before, every time. A cut made in a scoping session costs an afternoon of disagreement. The same cut made mid-build costs the build time already spent, plus the maintenance tax on code nobody uses. SIT Manager's real-time notifications, added in week six of an eight-week migration, is the counter-example: scope that crept in and had to be pulled back out mid-cutover.
Do cut features ever get rebuilt later?
Rarely, and only once real usage data asks for them. None of the four cuts across ONETAPP, SIT Manager, Tirth, and HospiHealth have been rebuilt. ONETAPP's publisher API and mobile app remain real roadmap items — just not week-one ones. The rule isn't 'no,' it's 'not yet, and here's the evidence we're waiting on.'