Working methods for the questions that sit between a traffic source, a tracker and a campaign report.
New technical notes
Three practical methods: creative–page fit, incremental retail contribution and conversion-event identity. Fictional examples, explicit limits.
Creative and landing-page diagnosis
A creative winner belongs to the conditions in which it was tested. Before carrying it onto another landing page, check whether the message and destination work as a pair.
Read the complete method →Retail measurement
Before repeating a promotion, identify what changed, what the control would have bought, and whether the additional contribution covers the intervention.
Read the complete method →Tracking and event identity
A repeat postback is not always a repeat conversion. Define the business event and its identity before you change duplicate handling.
Read the complete method →Use the measurement decision brief (plain text) · No email required.
A real implementation story, with the operating limits and evidence status visible.
Real project in progress · Overall shadow mode
How I defined execution ownership, limited actions and human approval in a Claude-assisted workflow. Independent validation remains incomplete.
A field guide and three deeper methods for traffic quality, budget exposure and retail economics. Worked examples are fictional.
Practical method · 25 October 2026
Start with the evidence that blocks the decision. A practical route through tracking, traffic quality, testing, budgets and retail economics.
Practical method · 25 October 2026
Find the first demonstrated failure between a source click and a meaningful action before judging traffic quality.
Practical method · 25 October 2026
Separate what you can still allocate from what the evidence is mature enough to tell you.
Practical method · 25 October 2026
Reconcile attributed sales with order economics and product availability before increasing media spend.
Three practical methods for creative fatigue, constrained delivery and campaign-change verification. All examples are fictional.
Method demonstration
A falling campaign CTR does not tell you that an ad has worn out. First check whether the same creative is reaching the same kind of traffic.
Method demonstration
An unused budget is capacity, not an instruction to buy more traffic. Find the constraint on delivery before changing the bid or opening the targeting.
Method demonstration
A job marked “success” can still leave the wrong setting in place. Verify the intended state, then observe delivery separately from the request response.
Three practical methods for placement reviews, source postbacks and controlled testing. All worked examples are fictional.
Method demonstration · 25 October 2026
A decision table for mature cohorts, spend caps and the next bid or budget change.
Method demonstration · 25 October 2026
A downstream troubleshooting sequence: ref ID, event mapping, outbound request and source reporting.
Method demonstration · 25 October 2026
A two-stage test brief with one changed element, fixed controls, mature outcomes and an honest inconclusive result.
These demonstrations show a process. They do not claim completed client outcomes.
Illustrative reconstruction
Define the event, trace the ID and reconcile the same reporting population.
Hypothetical plan
Separate the question, write the test and agree when to stop.
Useful URLs remain available. These pages are retained for compatibility while their examples and numerical claims are reviewed.
Reconciliation, duplicate postbacks, missing identifiers and reporting windows.
Keep cost definitions separate and avoid treating a cheaper upstream event as proof of profitability.
Read a general measurement distinction separately from the historical deposit CPA case.
Share the role, the team and where the work happens.