ROI isn't a number you discover at the end of a campaign — it's a constraint you design into the brief on day one. Here's the framework we use to keep campaigns honest from kickoff through retro.
Section 01
Set the payback window before the creative brief
Decide whether this campaign is judged at 30, 90 or 180 days, and what blended CAC qualifies as a win. Every creative and channel choice downstream is judged against that one number.
We watched a DTC brand burn $180k on a 'brand awareness' campaign that was secretly being judged on revenue — by a CFO who never saw the brief. Six weeks in, the campaign was killed for underperformance it was never designed to deliver. Naming the payback window in writing — and getting the CFO to sign off — would have either reframed the campaign or replaced it with the one the business actually needed.
Section 02
Pick a single primary KPI per campaign
Awareness campaigns optimize for reach and recall; demand campaigns optimize for qualified pipeline; lifecycle campaigns optimize for repeat rate. Two KPIs means no KPI.
When a campaign has two KPIs, the team optimizes for whichever one is easier to move, which is almost never the one that matters. Pick the harder one. If the goal is qualified pipeline, accept that CPC will look worse than a clicks-optimized campaign — and that's fine, because clicks aren't on the scoreboard.
Section 03
Build the measurement plan before launch
UTMs, event names, conversion definitions and the dashboard view all live in a one-pager that ships with the brief. Measurement bolted on after the fact is measurement you can't trust.
A one-pager that lives next to the brief: every UTM template, every event name, every conversion definition, the exact dashboard view, and who's responsible for reading it each Tuesday. Without this, the post-mortem becomes a forensic exercise where half the data was never captured and the other half is contested. With it, the retro takes 30 minutes.
Section 04
Stage the budget in thirds
Spend the first third learning, the second third doubling down on what worked, the last third scaling the single winning angle. This prevents the classic 'spread it evenly and learn nothing' pattern.
A client of ours had a $90k Q1 campaign budget and the instinct to deploy it evenly across twelve weeks. We argued for $30k in weeks 1-4 (across five tested angles), $30k in weeks 5-8 (concentrated on the top two), and $30k in weeks 9-12 (poured into the single winner). Final ROAS came in 2.4x what the even-spread model had projected — same budget, smarter sequencing.
Section 05
Iterate weekly, not at the end
Tuesday read on Monday's data, one decision per week: kill, hold, or scale. Campaigns die from indecision more often than from bad creative.
The discipline is one decision per week, not five. Look at last week's data on Tuesday morning, decide what to kill, what to hold, what to scale, and then leave the rest alone until next Tuesday. Daily tinkering creates noise; weekly decisions create compounding learning. The teams that resist daily fiddling almost always outperform the ones that don't.
Section 06
Run a written retro within seven days of end
What worked, what didn't, what we'd take into the next one. The retro is the deliverable that turns one campaign into a compounding library.
The retro is the deliverable that turns one campaign into a library. Three sections: what worked and why, what didn't and why, what we'd carry forward. We've watched teams who ship retros consistently get measurably better at campaigns inside two cycles, and teams who skip retros repeat the same mistakes for years.
The takeaway
ROI is a design constraint, not an outcome. Decide the payback window first, measure one KPI, and stage the spend so the third dollar is smarter than the first.