A sample, not a client. Sample Software Company is a composite: every name and number is illustrative. What the $1k Map includes
Sample Software Company
A B2B software company with a subscription product. 34 people, 12 engineers in three squads, a monolith and two services, weekly release target, two AI pilots under way.
This is a sample of the Engineering lens. The company is a composite and the numbers are illustrative, sized the way a real Map sizes them. Your Map uses your systems, your team, and your numbers, and stays confidential to you.
1 · Your goals
"Ship on the date we say, every release."
"Stop finding bugs after the customer does."
"Know which AI work is worth funding."
2 · Your systems today
| System | What it does today | State |
|---|---|---|
| Monolith | Core product, four years old. Billing and permissions live inside it. | Tangled few tests around billing |
| Two services | Reporting and integrations, split out last year. | Clean own repos and owners |
| Pipeline | Builds and tests on every merge. Deploys to production. | By hand one engineer runs the deploy |
| Tracker and reviews | Tickets, pull requests, release notes. | Uneven each squad works its own way |
| Support desk | About 600 tickets a month, sorted by hand. | By hand no link to the tracker |
| Two AI pilots | A support triage assistant and a code review assistant, each four months old. | Stuck no owner, no measure |
3 · How work moves today
Green steps are where the time goes. Median cycle time, first commit to production, is 19 days. 5 of the last 12 releases shipped on the planned date.
4 · What is fragile
- One engineer can deploy. When they are out, releases wait, and the last two slipped for that reason.
- Billing has almost no tests. Every change to it is a held breath, and it is where 5 of 14 monthly escaped defects start.
- The AI pilots have no owner. Each has a champion, neither has a decision maker, and neither has a measure of success.
- The squads grew faster than the standards. Three definitions of done, so quality depends on who reviewed the change.
5 · Where the opportunities are
| Opportunity | What changes | Worth | Measured by |
|---|---|---|---|
| A release cadence that holds | Scope freezes at a set point, a second person can deploy, and a release checklist replaces memory. | 6 hrs a week | Release predictability, 5 of 12 on time up to 10 of 12 |
| Faster review and merge | Review turnaround is set at one working day, and changes stay small enough to review. | 5 hrs a week | Cycle time, 19 days down to 8 |
| Defects stopped before release | Tests go around billing first, then an automated regression pass replaces the manual one. | 11 hrs a week | Escaped defects, 14 a month down to 6 |
| One standard across squads | One definition of done and one review checklist, with a named owner for each service. | Risk, not sized | Engineers able to ship a release, 1 up to 4 |
| One AI pilot to production | The support triage assistant gets an owner, a target, and a rollout. The code review assistant waits. | $24,000 a year | AI initiative return, hours returned against run cost |
| Total found | 22 hrs a week $24,000 a year | ||
How the numbers were sized. 22 hours a week over 48 working weeks at an illustrative $85 an hour is about $90,000 of capacity. Defects: 8 fewer a month at about 6 hours each is 11 hours a week. Review waits: 12 engineers each lose about 25 minutes a week, 5 hours in all. AI: 600 tickets a month at 6 minutes of sorting, cut by two thirds, is about 480 hours a year at $50.
6 · What to do, in order
Quick wins · next two weeks
- Train a second engineer on the deploy and write the checklist. small
- Set a one day review turnaround and a pull request size limit. small
- Name one owner for each pilot and write down what success is. small
Bigger bets · in this order
- Tests around billing. Stops the most defects first. 3 to 4 weeks
- A release cadence with a scope freeze. Makes the date believable. 4 weeks
- One standard across squads, with service owners. 4 to 6 weeks
- Support triage assistant to production, once the cadence holds. 3 to 4 weeks
7 · What to leave alone
- A rewrite or a move to microservices. The monolith's trouble is missing tests and unclear owners, and a rewrite carries both over. Fix those first and revisit in a year.
- The code review assistant. Review is slow because of queue rules and large changes. Fix those and there is little left for AI to solve.
- More tooling for the tracker. The three squads need one way of working before any new tool can help.
8 · How to get there
Recommended next step: Lead, with a Build for the billing tests. Several efforts need one owner who can set the standard, run the cadence, and decide what to fund and what to stop. It needs the CEO to back one standard across all three squads. The triage assistant starts with a person approving every action, and trust is extended one action at a time as the numbers hold.