There is a particular kind of agency relationship where everything is green until, suddenly, it is three months late. The weekly status updates were all fine. The burndown chart looked healthy. Then integration week arrived and nothing fit together.
The fix is not better reporting. It is making the increment real.
Demoable means deployed
Our rule is simple: if it cannot be clicked in a deployed environment, it does not get demoed. No local machines, no screenshots, no "imagine this connected to the API".
This is a harder constraint than it sounds. It means the deployment pipeline is built in week one, before the first feature. It means feature flags, so half-finished work can ship without being visible. It means the boring infrastructure work happens first, when it is cheap, rather than last, when it is a crisis.
What a fortnight actually contains
A two-week cycle for us breaks down roughly like this:
| Phase | Duration | Output |
|---|---|---|
| Shaping | Ongoing, one cycle ahead | Written scope with explicit trade-offs |
| Build | 8 working days | Merged, tested, deployed increments |
| Demo | 1 hour | Working software in a real environment |
| Adjust | Remainder | Re-scoped next cycle based on what we learned |
Shaping running a cycle ahead is the part teams most often skip, and it is the part that makes the rest work. Engineers should never start a cycle by figuring out what the work is.
Trunk-based, always
Long-lived feature branches are where integration risk accumulates quietly. We merge to trunk at least daily, behind flags where needed. The merge conflicts that would have been a painful week-long reconciliation become five-minute problems.
This requires a test suite people trust. If the suite is flaky, developers stop believing failures, and the whole model collapses. We treat a flaky test as a broken test and fix it the day it appears.
The uncomfortable side effect
Shipping visibly every fortnight means your stakeholders see the pace honestly. If a cycle produces less than expected, there is nowhere to hide it.
We think that is the point. A project that is going to be late is going to be late whether or not you find out in month two or month five. Finding out in month two is the only version where you can still do something about it.
What we ask of clients
This model needs one thing from your side: a decision-maker in the demo. Not a note-taker who relays feedback to someone else the following week. The value of a fortnightly cycle is that course corrections happen in the room, and that requires someone who can make the call.