The Two-Ledgers method

Every account we run reports two numbers, every week: the one Meta's attribution claims, and the one your revenue ledger shows. The whole method exists to make those two numbers converge and stay converged. This page is the engagement, phase by phase, including the parts most agencies keep vague.

Phase 1: the audit (days 1–14, $1,400)

Fixed scope, fixed price, no obligation to continue. We review structure, spend allocation, creative fatigue, and the event pipeline: every Purchase and Lead event traced to its source, Event Match Quality scored per event, deduplication verified. You receive a written findings document with the restructure plan and the projected timeline. Accounts we audit typically arrive with match quality between 4 and 6 and double-digit campaign counts.

Phase 2: counting rules, in writing

Before anything launches, both sides sign one page that states: which events count as results, which attribution window applies (and that it will not move), what "stable CPA" means (14 consecutive days inside a ±15% band of the 7-day average), and what happens when Meta's number and yours disagree. Boring on purpose. Every later argument is settled by this page instead of by whoever talks fastest.

Phase 3: the measurement build (weeks 2–4)

Pixel plus Conversions API with shared event IDs, hashed customer parameters, and a seven-day live verification. Target: match quality above 8 on the events that gate delivery. Incremental attribution, the holdout-based view Meta added in April 2025, goes on as a reporting default. Full detail on the measurement page.

Phase 4: restructure and creative cadence (weeks 3–8)

Campaign sprawl collapses into one to three consolidated ad sets feeding Meta's post-Andromeda delivery, with 10 to 20 live creatives rotated on a weekly calendar. Median time to a stable CPA across our accounts is 24 days from this phase starting. The reasoning and the evidence live on the management page and in the track record.

Phase 5: two-ledgers reporting, weekly, forever

The weekly report has three columns: Meta standard attribution, Meta incremental attribution, and your revenue. Gaps get one of three labels: expected (window effects), investigating (with a date), or resolved (with the cause). Once a quarter we run a holdout or spend-pause test on one segment to recalibrate. If the report ever stops matching reality, that is a fire, and it outranks every optimization task in the queue.

What the weekly report looks like

One page, three columns, four rows. The columns: Meta standard attribution, Meta incremental attribution, your revenue source of truth. The rows: spend, results, cost per result, and the gap between claimed and real, each with last week beside it. Under the table, at most three sentences of interpretation and one decision request. Reports that need a meeting to explain are reports designed to need meetings; both founders sign this one, and disagreements between them ship as footnotes rather than getting smoothed over.

What happens from month four onward

The engagement goes month to month after the first 90 days, and the cadence settles: four to six new creative concepts monthly, budget steps of roughly 20% when the stable band holds, a quarterly incrementality check, and a quarterly re-read of your costs against the current benchmark data, because a CPM that inflated 13% in a year quietly redefines what "flat performance" means. Leaving is one email; the account, the pixel, and every report stay yours.

What the method refuses

• No moved attribution windows to make a month look better.
• No "blended ROAS" reporting without the paid-only view beside it.
• No scaling while CPA is unstable, whatever the calendar says.
• No creative judged before it exits the learning phase.
• No account without a written restriction playbook: Meta disables over 2 million ad accounts a year, and appeals resolve fastest inside the first 48 hours.