Field Note · Measurement
Calibration is a system, not a study
Brands calculate a multiplier once, coast on it for months, and make budget decisions on a number that no longer reflects reality. Here's the system that keeps it honest.
Think of your marketing strategy as a stack: channels, funnel stage, and objectives at the foundation. Optimization and attribution one level up. Then bidding, creative, and budget on top.
The most advanced scaled advertisers build calibration into every layer. Every bid decision, every creative read, every budget allocation depends on its accuracy. If it's wrong, every decision above it is distorted.
Taylor Holiday frames this well: the goal isn't perfect measurement. It's progressive truth, getting systematically less wrong over time, so your reporting moves closer to the financial realities of your business.
The operational fix has five parts:
- Run lift studies on a recurring cadence. Quarterly at minimum for brands spending over $500K/month. Start with single-cell studies at the market and funnel stage level, then multi-cell studies for tactical decisions. Each produces a calibration multiplier for that segment. No direct study? Use the closest available multiplier and flag it for the next cycle.
- The math is simple. (The organizational buy-in to care about the math? Less simple.) Incremental conversions divided by attributed conversions. That's your calibration multiplier, your incrementality adjustment factor. It can be above or below 1.0. Either direction is valuable. The multiplier on conversions may differ from the multiplier on revenue. More on that another time.
- When a multi-cell test identifies a winner, you get two outputs. The winning strategy and an updated calibration multiplier for that tactic. Put differently: if you change your strategy, change your multiplier. Multi-cell studies split audiences, which inflates iCPA. A multi-cell multiplier is a conservative read. Real-world performance should only improve.
- Track multipliers over time. A simple table works: date, channel, funnel stage, tactic, lift result, attributed result, multiplier. Watch for drift. When conditions change, flag your multipliers as stale and prioritize a new study.
- Use calibration as a sanity check before acting on any result. Before you adopt a winner, ask: does my measurement still reflect reality? A stale or misapplied multiplier will distort the read.
The most scaled brands are moving toward continuous calibration, where multipliers refresh automatically. But the discipline starts with the fundamentals: run the tests, calculate the multipliers, track the drift.
Once you have calibrated multipliers across your key segments, use them. Allocate budget based on calibrated results, not attributed results. Fund the campaigns that drive the highest incrementality. That's where calibration becomes a growth lever.
Your incrementality test tells you the truth once. Your calibration multiplier tells you the truth every day.