# The first ninety days of a retainer, as it actually goes

> Month one reads the codebase before touching it. Month two clears the queue that built up while nobody owned the system. Month three is where it starts looking like progress instead of maintenance.

The first month looks unproductive from the outside and is the most valuable one: reading the code nobody documented, mapping what talks to what, learning which parts are load-bearing superstition. The deliverable is not features — it is a written map of the system and a sorted list of what is actually on fire versus what merely smells of smoke.

Month two spends the map: the fixes that were queued for 'when we have a developer', the dependency updates everyone postponed, the backup that was never tested. Month three is where steady-state starts — small features landing weekly, the monitoring catching things before users do, and the monthly call shifting from 'what broke' to 'what is next'. That arc is the thing being bought, not the hours.

## Week one to four: the map

Access, credentials, architecture notes, the deployment nobody remembers, the cron jobs that run the business silently. Everything learned goes into writing — because a retainer whose knowledge lives in the engineer's head is the same single point of failure it was hired to remove.

## Month two: the queue

The bug list everyone knew about, the security updates postponed twice, the 'temporary' workaround from 2023. The queue is triaged by blast radius: what breaks the business if it fails, then what costs hours weekly, then what is merely ugly. Not everything in the queue gets fixed — but everything in it gets a decision.

## Month three: steady state

Small improvements shipped weekly, monitoring tuned until alerts mean something, and a rhythm established — a standing call, a board both sides can see, a monthly note on what happened and what is queued. The system starts to feel owned. That feeling is the product.

## What to take away

- Month one produces a map, not features — and that is the valuable part.
- Month two clears the queue by blast radius, not by loudness.
- Month three is steady state: weekly ships, meaningful alerts, a rhythm.
- Everything learned goes in writing — knowledge in one head is the problem being replaced.

## Tags

Retainer, Maintenance, Process

## We build this for clients

https://dfieldsolutions.com/en/services/full-stack

## More from the lab

- https://dfieldsolutions.com/en/lab/custom-software-cost.md — What custom software actually costs
- https://dfieldsolutions.com/en/lab/client-portal-vs-email.md — Why clients get a progress page instead of status emails
- https://dfieldsolutions.com/en/lab/public-status-page.md — The public status page replaces the status meeting
- https://dfieldsolutions.com/en/lab/custom-vs-off-the-shelf.md — Custom software or off the shelf — the honest test

---

Source: https://dfieldsolutions.com/en/lab/first-90-days-retainer
DField Solutions — Dunakeszi, Hungary — dezso@dfieldsolutions.com
Booking: see https://dfieldsolutions.com/en/contact