You know what it is, how it works, why it holds up, and why the clock matters. Now the number.
You are not renting a box.
Three things move the number.
What it costs depends on what the work actually is, how much capacity sits behind it, and how fast you need it. So every figure here is a range — and we show you which factor moved yours.
The model
What actually changes the cost
1 What the work is
Keeping something running, building new capability, getting off a legacy system, or work that has to prove itself to a regulator. Verification effort per story is not constant across those.
2 How much sits behind it
How many lanes run in parallel. More is faster until it isn't — and we will tell you where that point is for your model rather than sell past it.
3 How fast you need it
Review effort scales with the amount of work, not the calendar. Compressing the same scope into half the time concentrates the human cost rather than spreading it.
A flat monthly rate would be simpler, and it would be wrong. Keeping a quiet application running and rebuilding a regulated system against a fixed date are not the same job, and charging the same for both means one of you is subsidising the other.
Move the factors
See what each one does
A build-day is one worker building for one day — not a calendar day, since several run at once. One build-day is about 11 delivered stories.
Measured on a backlog averaging 7.8 acceptance criteria per story (median 6). Your backlog sets your own conversion — which is why we quote build-days, not stories.
Starting points
How an engagement begins
Each of these is a band, sized by the drivers named beside it. None of them is a commitment to the monthly above — they exist so we can tell you something true before you commit to anything.
Application analysis
A read-only pass over an application you already have: security and access, upgrade path, model quality, performance. You keep the report whatever you decide next.
Three to five working days. Nothing in your model is changed.
Legacy assessment
We extract what your platform will give up — screens, schema, permissions, rules — and hand back a scored inventory, a migration profile, a risk register, and a date we will commit to.
Credited toward onboarding if you proceed.
Onboarding & provisioning
Your isolated instance, secrets vault, repository wiring, brand intake, and your first full approval cycle walked through with you.
Required on a new instance.
Discovery & backlog
We read your requirements, your legacy database and your screens, and hand back a written backlog with acceptance criteria, a migration profile and a risk register.
On one engagement this turned a multi-hundred page specification into several hundred estimated stories in a day.
A standing commitment
Support agents
An agent is sustained capacity rather than a person — which is why agents price through the same three factors as everything else instead of having a private rate card. Set the estimator above to keeping something running and move the lanes: one lane is roughly one agent.
What response you expect changes the answer
Business hours, best effort
Covered by the capacity you buy. Work queues and gets pulled like any other unit.
SLA-backed response
Covered, but it changes how the capacity is spent: reserved capacity waits for work, so expect less delivered from the same build-days. That trade is the SLA.
On call, outside hours
Priced separately. This buys availability rather than output — someone is obliged to be there whether or not anything breaks — and no capacity band can express that honestly.
Fixed bid
Whole projects
When you want the outcome rather than the capacity, and you want the risk on us. These are bands — the width is the point, and it narrows as we learn what drives your number.
One focused tool for one team — a form, a queue, a tracker that a spreadsheet has outgrown.
Something a department depends on daily, with real roles and real reporting behind it.
Core to how the business runs. Several user types, integrations, and data that matters.
A system of record. Downtime and data integrity matter more than delivery speed.
A band is not a quote. anything a form can give you is a range with its assumptions written beside it.
What we will not sell you
The ceiling is real
Throughput is bounded. Shared surfaces — navigation, security, the common model — are single-writer, so beyond roughly six to eight concurrent workers you are queueing, not scaling. We size to that reality instead of selling workers that would sit idle.
It is also why capacity has a price factor at all. If more lanes were always better we would just sell you more lanes.