Note
Published
2026-08-10
Reading
6 min
Data
Public sources
Headline figure
High
Claim
When not to hire a cost mechanic
Eight situations disqualify this studio, and in every one of them saying so on the first call is cheaper for both sides than discovering it in week two. Two of the eight are about the studio being too small, three are about the work being someone else's, and three are about the buyer not being ready.
Topics: method
Built from public provider documentation and sourced research. Worked figures inside are labelled synthetic. No client data was used.
Argument
The disqualifying question first
Can you name the event driving this, and its date? A board deck with a gross-margin slide. A pricing change. A credit expiry. A margin floor committed in an enterprise contract. Series B diligence. A bill growing faster than revenue.
If nothing has a date, the work is real but it is not urgent, and an engagement bought without a date is the one that gets deprioritised in week two and resented in week four. Read the methodology instead and come back when something has a deadline attached to it.
The eight
| Situation | What happens instead |
|---|---|
| Your security review requires a SOC 2 report | None exists this year. The compensating path is in the access appendix, and if you cannot waive the requirement the answer is no |
| You spend above roughly 50M a year | Your own hires and a Tier 1 firm win on arithmetic at that size. This ladder cannot cover an estate that large, and pretending otherwise costs you a quarter |
| You want hours, or a body in a seat | No hourly rate exists. Every rung is a fixed price against a named artifact. Hourly pays us to be slow |
| You want commitments purchased for you | Automation ladders commitments better than a person clicking Buy. ProsperOps, Zesty and similar are named, referred, and pay no commission here |
| You want a dashboard, or a tool resold | OpenCost, Kubecost, the AKS Cost Analysis add-on and Infracost already exist. We wire up what you own and correct it for your estate |
| Your exports cannot support a defensible number | The $900 export check says so in writing, before a teardown is sold. Tag activation never backfills, so sometimes the honest answer is 30 days of clean export first |
| Nobody can name the event, and nothing has a date | Not urgent enough yet. You get the methodology pages and no follow-up |
| You need references, logos or a bench | None are offered. What is offered instead: published arithmetic, fixed published prices, and credentials you create and revoke |
Two of these are permanent, six are not
The 50M ceiling and the absent staffing redundancy are structural. Capacity here has a limit that is arithmetic, not modesty, and the caps are published in the access appendix.
The other six can change on your side. A security review can waive SOC 2 when the compensating controls are enforced in your own account. Exports can be fixed in 30 days. An event can arrive. When one of those changes, the answer changes with it.
Why publish this at all
A studio with no track record has no reference to call, so the only credible signal available is the willingness to name the cases where the answer is no. A vendor that qualifies nobody out is telling you it needs the deal more than it needs to be right.
Sources are listed at the foot of this note with their confidence stated. Medium and low confidence figures say so in the copy.
Every query here runs against a read-only role you create, scope and revoke. Access policy.
Close
More notes
Check the arithmetic yourself
The full derivations, with the queries written out so they run under your own read-only credentials, in your own console.