Operating model
Instrument, then optimise — never the reverse.
Every engagement, regardless of which capabilities it draws on, runs through the same three phases in the same order. The order is not a preference. It is the entire reason the work holds up under audit.
Why the order is fixed
Optimising before instrumenting optimises the wrong thing
A budget reallocation based on unreconciled numbers does not fail randomly — it fails in the direction the unreconciled numbers were already biased.
We have run this sequence — instrument, find the constraint, compound and hand back — on every engagement for long enough to know what happens when a client asks us to skip straight to optimisation. The account improves against its own reporting and gets worse against the business. Platform-reported numbers are structurally optimistic, so any optimisation run against them before reconciliation simply moves budget toward whichever channel is best at over-reporting itself, not whichever channel is actually performing.
The three phases are not independent projects. Instrumentation in weeks 1–3 produces the reconciled numbers that make constraint analysis in weeks 4–12 possible at all — you cannot isolate a binding constraint from data you do not trust. And the handover phase only works because the first two phases were documented as they happened, not reconstructed from memory once a client asks what was actually done.
The three phases
Read each one in full
01
Instrument
Weeks 1–3. Nothing is optimised until it can be measured — collection rebuilt, platform numbers reconciled to your ledger, first.
02
Find the constraint
Weeks 4–12. Every account has one binding constraint. We isolate it and put the whole budget behind relieving it.
03
Compound and hand back
Quarter 2 onward. Systems get documented and transferred, reducing your dependency on us by design.
A fixed rule
No budget moves before phase one closes
This is the one rule in the operating model we do not negotiate on. Clients under pressure to show quick movement sometimes ask us to start testing budget changes in parallel with instrumentation. We decline, because a test run on unreconciled data cannot be trusted regardless of its result — a positive result might be a reporting artefact, and a negative result might be hiding a real gain. Three weeks of patience at the start of an engagement is what makes the following nine months of decisions defensible.
Continue