D2C Home: $1.9M of Additional Revenue on a Flat Ad Budget

Background

Hallowell Home spent approximately $424,000 per month across six paid channels: Meta, non-brand Google Search, branded search, retargeting, TikTok and YouTube. Blended return on ad spend, as reported by the platforms' own attribution, was 3.6× — and the growth plan extrapolated that figure linearly.

The situation

Platform attribution assigns credit to the advertisement closest to the sale. Two channels benefit structurally from that convention: branded search, which captures customers already searching for the company by name, and retargeting, which reaches customers already on the site. Together they held 38% of the budget.

The company ran two geographic holdout experiments — switching each channel off in matched markets and measuring what sales actually did against a pre-period baseline.

Branded search: reported 5.0×, measured 0.87× [95% CI 0.65–1.09]. Retargeting: reported 3.9×, measured 0.32× [0.07–0.57]. Both were predominantly harvesting demand created elsewhere. The mirror error also surfaced: YouTube, held at 6% of budget, reported 1.8× under last-click while the calibrated model estimated its true return near 3.0×.

The statistical model alone illustrated a further point: uncalibrated, it fitted two years of weekly data well (R² 0.94) while producing channel estimates that diverged badly from experimental results on smaller, correlated channels. Estimates were therefore anchored to the experiments, and untested channels carried no decision authority pending a testing rotation.

The intervention

1. 01Cut branded search from 22% to 8% of budget, retained for competitive-term defense only.
2. 02Cut retargeting from 16% to 6%.
3. 03Reallocated in two stages: half on the model's direction, the remainder only after experimental confirmation.
4. 04Scaled YouTube from 6% to 16% on the corrected measurement, with a confirmation test scheduled.
5. 05Refreshed creative on the largest channel.
6. 06Held total budget flat; installed a quarterly experiment rotation to keep measurements current.

Results — twelve months following

Revenue rose $1.9M on an unchanged budget — roughly $0.8M of contribution profit at the company's margin structure. Reallocation on a genuinely flat budget produced low-teens growth, a realistic ceiling at prevailing saturation levels and materially below what a linear reading of platform figures had implied. The saturation analysis also showed the media program near its efficient frontier, informing a decision to direct incremental capital to creative development, new channels and conversion improvement rather than more spend on existing channels.