Willow & Whisker: Reading a Six-Week Paid-Search Pilot Before Committing $35,000
Willow & Whisker operates six pet-supply stores in a mid-sized metro and has $60,000 for third-quarter marketing, its entire discretionary budget. Same-store foot traffic is down 8% year over year while online orders through a basic website are up 22%; average ticket is $62 and gross margin is 41%. The marketing manager ran a six-week paid-search pilot in two stores: 34 conversions attributed to search versus 29 in matched control stores, at a cost per click of $1.15 and a 3.2% conversion rate. She wants $35,000 for paid search, $15,000 for a title sponsorship of a regional dog show expected to draw 8,000 attendees, and $10,000 for a loyalty app. The owner notes that 41% of customers who bought twice in a year went on to buy a third time within 14 months, and asks whether the pilot proves that search works.
1. Evaluate whether the pilot justifies a $35,000 paid-search allocation.
Read the worked answer
The pilot does not prove that search works. Thirty-four versus twenty-nine conversions over six weeks is a five-conversion gap on a base near thirty. Treating conversions as roughly Poisson, the standard deviation of a count near 29 is the square root of 29, about 5.4, so a five-count gap is under one standard deviation from noise. A credible test needs hundreds of conversions per arm, or a much longer window.
Two further problems. The pilot ran in two stores, so store-level shocks, such as a local competitor closing or a weather event, are confounded with the treatment. And attributed conversions depend entirely on the attribution window; a 30-day last-click rule will claim credit for sales that would have happened anyway, especially for a brand with existing foot traffic.
The honest conclusion is that the pilot is a signal worth investigating, not evidence for a $35,000 commitment. It justifies a properly powered geographic holdout, and the spend should be defended on unit economics rather than on this result.
2. Compute the cost per acquisition and the first-order contribution. Then assess whether the 41% repeat-purchase statistic can be used to justify the spend.
Read the worked answer
CPA = cost per click divided by conversion rate = $1.15 / 0.032 = $35.94. First-order contribution = $62 x 0.41 = $25.42. The first order therefore loses $10.52 before any repeat business.
The 41% figure is the trap. It is the share of customers who had already bought twice and then bought a third time within 14 months. It is not the probability that a newly acquired first-time buyer makes a second purchase. Applying a conditional rate observed on a two-purchase cohort to a first-purchase cohort commits a base-rate error and would justify almost any CPA.
To break even at $35.94 with $25.42 per order, search needs about 1.41 orders per acquired customer, undiscounted. The correct framing is that search is an acquisition channel whose payback depends on second-purchase rate and repeat cycle, and those are currently unmeasured for search-acquired customers. Before scaling, instrument first-to-second purchase rates at 90 and 180 days for search-acquired versus holdout customers. If fewer than roughly 40% of search-acquired buyers repurchase within a year, the $35,000 destroys value.
3. Recommend a specific allocation of the $60,000 and the metrics that should govern the next reallocation.
Read the worked answer
A defensible split: $18,000 to search, run as a two-market holdout over a fixed 90-day window with a pre-committed CPA ceiling near $30; $15,000 to the dog-show sponsorship only if it buys named lead capture, meaning a booth scan, a text offer with explicit consent, and a unique redemption code, not logo placement; $12,000 to loyalty and email marketing aimed at the existing customer base, where marginal cost per incremental visit is far below paid acquisition; and $15,000 held in reserve, released only if the search holdout clears its threshold.
Test the sponsorship against a hurdle: 8,000 attendees at a realistic 4% capture is 320 leads, or $46.88 per lead at $15,000. That is above the search CPA, so the sponsorship must convert at a materially higher rate or be cut at renewal.
Governance metrics: cost per incremental visit adjusted by the holdout rather than last-click, first-to-second purchase rate at 90 and 180 days, average ticket by acquisition cohort, and loyalty enrollments per dollar spent. Report monthly, and require the search spend to stay under its ceiling or be reallocated.