The Influencer Campaign With No Attribution
You run marketing for a small DMC selling 5–9 day Japan packages. You spent ¥1.8 million on a three-month campaign: ¥1.2 million with two travel influencers (one 180k followers, one 40k) and ¥600,000 on paid search. Bookings traced to a unique promo code from the influencers total 9 bookings at an average gross margin of ¥85,000. Paid search shows a 4.1 ROAS on last-click, but your booking system has no view-through or assisted-conversion data, and the influencers posted discount codes that some customers used after finding you via search. Your cost per acquisition on organic/repeat is low, and the owner is asking whether to repeat the campaign.
1. Evaluate the campaign's performance and decide whether to repeat it, being explicit about what the data can and cannot tell you.
Read the worked answer
What the data shows:
- Influencer-attributed gross margin: 9 × ¥85,000 = ¥765,000 against ¥1.2 million spend → roughly 0.64 return on gross margin, below break-even on a last-touch basis.
- Paid search: 4.1 ROAS last-click on ¥600,000 → about ¥2.46 million revenue attributed, but ROAS is revenue, not margin; convert to margin before comparing.
What the data cannot show:
- Assisted conversions. The influencers' codes were used by people who also found you via search, so search is likely over-credited (last-click bias) and the influencers under-credited - or the reverse if codes leaked. Without view-through or a holdout, attribution is unreliable.
- Lifetime value. A 5–9 day package customer may rebook or refer; 9 bookings may be worth more than their first-trip margin.
- Brand vs. direct response. The 180k influencer may drive awareness that pays back slowly.
Decision logic: don't repeat at the same spend on the same measurement. The influencer leg is not demonstrably profitable on gross margin, but the measurement is too weak to condemn it. Recommend a smaller, better-instrumented test: unique codes per creator, a post-purchase 'how did you hear about us' question, a geo or audience holdout, and margin-based (not revenue-based) ROAS targets. Scale what clears a margin hurdle; cut what does not after the test.
2. Design the measurement plan for a repeat test and set a decision threshold.
Read the worked answer
Measurement plan:
- Unique, non-shareable tracking per creator (distinct codes/links) and a mandatory attribution question at booking.
- Margin-based KPIs, not revenue: target gross margin per acquisition against a defined CAC ceiling. If repeat/organic CAC is, say, ¥30,000, set the influencer CAC ceiling at or below that plus a brand premium you are willing to pay.
- Holdout or staggered launch so you can compare a treated audience against a control where feasible.
- Assisted-conversion capture via the booking survey and a longer attribution window (e.g., 30–60 days view-through).
- LTV tracking on the 9 existing bookings to see if repeat/referral changes the math.
Decision threshold (example): repeat only if the test clears a blended gross-margin return of at least 1.0 within the campaign window, or shows a credible path to it via measured assisted conversions and LTV. If it clears, scale the better-performing creator; if not, reallocate to channels with proven CAC.
Key point: state the threshold before the test so the owner's decision is rule-based, not a debate over ambiguous last-click numbers.