← All playbooks·Digital attribution

What you need first

• Daily TV GRP or delivery log for the flight by day and daypart, from the TV buying platform or ad server
• Daily brand search and direct traffic time series covering 4 weeks pre-flight through 1 week after the flight ends, from GA4/Search Console and the brand paid-search campaign
• Pre-flight baseline growth trend (week-over-week rate), computed from the same source over the 4 weeks before the flight
• Media calendar listing any other campaigns running in the same window, from the planning team, to flag confounders
• Brand search conversion rate and average order value, from the ecommerce/analytics platform
• TV flight net media spend and its own last-click conversions, from the invoice and campaign tracking

The procedure

1. Pull the daily brand search and direct session series for 4 weeks pre-flight, the flight window, and a 1-week tail
2. Project the pre-flight trend forward day by day to build the counterfactual baseline for the flight and tail weeks
3. Overlay the daily GRP delivery and compute daily uplift as actual minus baseline for every flight and tail day
4. Net out uplift explained by concurrent campaigns flagged on the media calendar
5. Sum the adjusted daily uplift across the flight and tail window to get total incremental sessions
6. Apply the brand conversion rate and AOV to the incremental sessions to get incremental conversions and revenue
7. Divide TV net spend by incremental conversions to get the halo-adjusted CPA, and set it beside TV's own last-click CPA
8. Log the baseline method, trend rate, and decay window length used, so the number is reproducible

Worked through with numbers

Pre-flight baseline is 4,000 brand search sessions/day with a 2% week-over-week trend. Flight runs 2 weeks (1,200 GRPs, €180,000 net spend), plus a 1-week tail. Trended baselines: week 1 = 4,000×1.02 = 4,080/day, week 2 = 4,080×1.02 = 4,162/day, tail = 4,162×1.02 = 4,245/day. Actuals: week 1 = 6,000/day, week 2 = 6,400/day, tail = 4,900/day. Daily uplift: week 1 = 1,920, week 2 = 2,238, tail = 655. Weekly totals (×7): 13,440 / 15,666 / 4,585, summing to 33,691 incremental sessions. The media calendar shows a paid social brand campaign overlapping week 2 at low intensity, estimated to explain 5% of that week's uplift: 15,666×0.05 = 783. Adjusted total = 33,691 − 783 = 32,908 incremental sessions. At an 8% brand conversion rate that's 2,633 incremental conversions, and at €65 AOV that's €171,145 incremental revenue. Halo-adjusted CPA = €180,000 / 2,633 = €68.4. TV's own last-click conversions (from a vanity URL) were 300, for a last-click CPA of €600. Combined credited conversions = 2,933, blended CPA = €61.4. Read it as: last-click tracking alone made TV look nearly 10x more expensive per conversion than it actually was once the brand search halo is counted.

Where it goes wrong

• Use a trended baseline, not the flat pre-flight average, since a flat baseline overstates uplift on a channel that was already growing month over month
• Extend the measurement window past the last airing with a decay tail instead of cutting off on the last flight day, since brand search response lags and trails for several days after the last spot
• Net out concurrent campaign activity from the media calendar before crediting all of the uplift to TV, or the same lift gets double-counted as halo for two channels
• Report the halo-adjusted CPA next to the channel's own last-click CPA instead of replacing it, since stakeholders comparing across channels need both numbers to trust the adjustment

How to know it is right

Terms used

• Halo Effect
• Zero-Click Search
• Conversion Attribution

← All playbooks Calculated in ValueMetrics →