A hotel chain ran a 3-month influencer program: 332 posts across 21 creators and 7 properties. The headline metrics are real. The composition behind them is fragile. The question this dashboard answers: now what?
Look at the totals and the program looks healthy. Look at how those totals are distributed across creators, and you see a single creator carrying twenty others. Three decisions, in the order they need to happen:
Make the contract measurable, not just paid.
They're a link shelf, not an engagement format.
Two creators per property. Then we can learn.
Of program engagement comes from Creator N at Hotel C. 15 posts — 4.5% of program output — produced 1,155,082 of 1,412,995.
The other 20 creators across 6 hotels produced 257,913 combined.
Was this just one lucky TikTok?
No — Creator N has a track record. 6 of their posts got more than 50,000 likes / comments / saves / shares each; 4 got more than 100,000. This wasn't a single viral hit — it was a creator who consistently outperforms.
Read this as a stress test: drag the slider to remove the top creators and watch how much engagement the program would still have. A healthy program would lose engagement gradually as you remove people. This one falls off a cliff the moment you remove the top creator — the top three creators alone own 88.1% of all engagement. That's the fragility.
Unobserved pairing. The per-post estimate is anchored to Creator A's own track record (60%), then nudged by Hotel C's relative performance (30%, log-tempered) and TIKTOK format effects (10%).
We have a Creator-Quality score for every one of the 21 creators × 7 hotels = 147 possible pairings (the matrix in Fig. 02). The optimizer picks the subset of pairings that maximizes total CQ — under three executive rules: (1) drop creators below the Pareto floor; (2) cap how many hotels any one creator can carry; (3) make sure every hotel gets enough creators to A/B test. Move the sliders on the left to see the picks update.
Last cycle, 81.7%of engagement came from one creator. That isn't a portfolio — it's a bet. After dropping creators below the Pareto floor and capping any one creator at 3 hotels, the optimizer picks 6 creators covering all 7 hotels, with effectively ~4.9 independent creators in the program (up from 1.5). If any one of them underperforms next quarter, we lose ~24%, not ~82%. Creator N stays — they're still the highest-quality pairing for Hotel C — but their slate is capped so we can finally run cross-property A/B tests with newer creators and learn faster.
Read this as: how big is the audience that sees a typical post (left-right), and how often does that audience save / share / comment(top-bottom). Carousels are in the sweet spot — small reach but the audience really interacts. Stories show the opposite: lots of views, almost no measurable interaction. Dot size = total engagement. TikTok's big dot is one creator's viral hits, not a consistent pattern.
For each post we predict expected engagement from the creator's, format's, hotel's, and platform's typical performance. Posts that beat or missed the prediction by 2× or more land here. The hidden gems are signals to investigate: “why did Creator I's Carousel hit 8× its profile?” is the question a creative debrief should answer.
A 0–100 score per hotel that blends five signals (weights shown below). Higher = healthier creator program. Hotel C is shown twice: with Creator N (the headline number) and without (the fragility test).
Engagement and reach are OBSERVED math— derived directly from the program's median ERs and median views/post. Revenue is MODELED · DIRECTIONAL: a range, never a point. Edit the assumptions above to argue with the model.
Creator N (score 81 at Hotel C), Creator U (score 66 at Hotel A), Creator S (score 64 at Hotel B) are the three strongest creators in the data. Renew Creator N with usage rights and paid-amplification permissions, but pair every hotel with a second creator before next quarter so the program isn't dependent on a single person.
This is a recommendation, not a guarantee. The current data can't tell us if Creator N is great everywhere or just great at Hotel C — that's why we recommend a cross-hotel test next quarter.
JC × VN_ Data Set.xlsxA measurement person. Builds the spine before the slide.