Fixed scope with written exclusions
The exclusions section is the one worth reading and the one most proposals leave vague. We write down what is not included before work starts, not after a disagreement.
Company
The commercial terms of an engagement shape its outcome more than the methodology does. These are ours, including the ones that operate against us — because a term that only ever protects the vendor is not a term, it is a hedge.
Tell us what you need. Scoping conversations are free and sometimes conclude you do not need us.
Company
Each one is written into the engagement rather than described here and forgotten.
The exclusions section is the one worth reading and the one most proposals leave vague. We write down what is not included before work starts, not after a disagreement.
Anything we should reasonably have identified during scoping is our cost to absorb. New requirements are a change order. That distinction is contractual, not a matter of goodwill.
Every engagement starts with a read-only connection. Write access is a separate, later decision made once reconciliation has been clean for a period.
We do not recommend retiring a system until a shadow ledger has tied across three consecutive closed months. This has delayed cutovers by a quarter and we have never regretted it.
No handover from a sales team to an unnamed delivery team. We are small enough that this is not a promise requiring enforcement.
Automation engagements begin with a week of measurement from your systems rather than a workshop estimate. Without a baseline nothing afterwards can be evaluated.
We build with AI tooling, which makes the marginal cost of an engagement largely measurable. Our price is that cost times five, with a floor per engagement type. The multiple covers scoping, review, testing, deployment, warranty, and the engagements that go wrong.
Publishing the formula is not transparency theatre. A customer who understands the pricing basis argues about scope rather than about rate, and scope is the conversation that improves the outcome. Rate negotiation improves neither party’s result.
A day rate rewards taking longer. Every consultancy will tell you their incentives align with yours and the billing model says otherwise — a project that runs over is a better outcome for the vendor than one that lands early.
Fixed price on defined scope reverses that. If we do it faster we keep the difference; if we misjudge we absorb it. That only works if scoping is done properly, which is why discovery is a real phase rather than a sales call.
Billing for these is how a vendor relationship turns adversarial, and the amounts involved are never worth the damage.
We turn work down regularly, and it is worth knowing when. An integration that cannot justify its maintenance — used monthly by two people, $18,000 to build, $400 a month to run — gets neglected, and a neglected integration produces wrong numbers quietly. We recommend a scheduled export instead.
A process running under fifty times a month rarely justifies automation. A rule that never changes is a policy rather than a module: cheaper, more predictable, and it does not need maintaining. A workflow whose rules four people describe differently is a governance problem first, and building encodes one person’s version permanently.
And anything that would bypass the policy engine, write to the ledger directly, or log to a separate schema. Those we decline at any price, because the constraints are what make the rest defensible.
It will. Our error policy sets out how errors are detected, corrected as reversals rather than edits, and disclosed — including errors you would probably never have noticed.
Critical errors affecting financial data that has been externally relied upon are notified within 24 hours of confirmation, by direct contact, with the affected records listed rather than characterised and the impact quantified.
Where the fault is ours we correct it, produce the analysis your auditors need, provide a written explanation suitable for sharing with a board or lender, and fix the underlying cause — none of it conditional on anything.
Timely decisions from someone with authority to make them. That is the single largest predictor of whether an engagement lands, and it accounts for roughly a fifth of the stalled implementations we have been called into.
Honest answers about data condition. “We have not looked” is a useful answer and a common one. Optimism here is expensive, because the gap surfaces during load rather than during scoping.
And access to the people who actually do the work, for a few hours. Processes redesigned to a team rather than with them do not survive a busy month.
Discovery failures being our cost. Not billing for connector maintenance or training that did not land. Delaying a cutover until three closed months have tied, which pushes revenue recognition by a quarter. Disclosing errors you would not have found. Each of these is a real cost and each exists because the alternative damages the relationship more than the amount is worth.
Questions
Ninety minutes, a written recommendation, and no obligation. Occasionally the recommendation is to do nothing.