Every number in this report has a provenance.
Two chips run the legend. OBSERVED means the number is read from the source XLSX, or is a deterministic function of it (a count, a sum, a median). MODELED · DIRECTIONAL means the number depends on at least one assumption you can edit. Modeled numbers are always shown as a range, never as a point estimate.
Every term, defined.
Total interactions on a post: likes + comments + saves + shares (organic, not paid). For Stories, also includes reply taps and profile visits.
For ~64% of rows (mostly Stories) the platform-reported total exceeds likes+comments+saves+shares because Stories add reply taps and profile visits. We trust the platform total.
How often the audience interacts with a post, as a % of follower count. Per post: (engagements ÷ followers) × 100. Group-level numbers use the median across posts unless labeled "total".
A statistical model that estimates how much each marketing channel — organic social, paid social, experiential, etc. — drives sales or brand outcomes. Vendors like Nielsen, Analytic Partners, and Marketing Evolution build them. Most MMMs need a year of clean data to calibrate.
Ch. II of this dashboard is about preparing the dataset that will eventually feed the MMM.
0–100 score measuring how likely customers are to recommend the brand. Survey-based, runs alongside the marketing data as an outcome variable.
Saves + shares + comments. Excludes passive likes — what the model treats as a user signaling intent rather than reflex.
High-intent engagements ÷ followers × 100, per post. Plotted as the Y axis of the Discovery × Intent matrix.
Median views per post by format. The X axis of the Discovery × Intent matrix — read as a reach proxy, not as click-through.
A 0–100 health score per hotel. Blends five signals: creator count (25%), engagement rate (25%), post volume (20%), format diversity (15%), comment tone (15%). Higher = healthier creator program at that hotel.
The composite weighting is a tunable choice — the sub-scores are observed. Hotel C is intentionally shown both with and without Creator N to expose the fragility.
A Lorenz-style curve of cumulative engagement share by creator, sorted descending. The slope is the program's concentration risk.
Removing top-1 (Creator N) cuts engagement to ~257k. Removing top-3 cuts it nearly to zero. That is the brittleness story.
Per post: positive / neutral / negative comment counts ÷ total comments × 100. Per-format figure is the mean of those per-post percentages.
Stories are excluded — they generate negligible comments.
Median ER with vs. without each creative attribute (Music, Voice Over, Text Overlays, Speaking to Camera). Computed on posts excluding Creator N to avoid one creator dominating the signal.
Cells with n < 10 on either side are flagged "thin sample" — directional only.
How much engagement the program would get under a new format mix. We scale the actual engagement total by the ratio of (engagement rate of new mix) ÷ (engagement rate of current mix). No assumptions — just real engagement rates by format.
How many views the program would get under a new format mix. Same logic as projected engagement, but using median views per post by format.
Projected engagements × your booking-rate assumption × your average-booking-value assumption. Always shown as a range, not a single number.
Both the booking rate and the average-booking-value are guesses — no actual booking data was provided in the case file. This is a planning tool to test "what would the upside be if X% of engagements convert", not a forecast.
What this organic reach would cost if you bought it as paid ads. (Views ÷ 1,000) × CPM (cost-per-1,000-impressions). A way to put a dollar number on organic reach for budget conversations.
Useful for ranking organic against paid investment. Not a ROAS replacement — paid and organic interact in ways this number doesn't capture.
Creator follower size band: Nano (<10k) / Micro (10–100k) / Mid-Tier (100–500k) / Macro (500k–1M) / Mega (1M+). As assigned in the source XLSX.
What this read does not claim.
- No paid amplification, hotel revenue, occupancy, booking, brand-lift, or post-level click data was provided. Recommendations sit on engagement, reach, and sentiment alone.
- The reporting window is 2024-05-18 → 2024-09-01. Three calendar months (June – August) are the analysis spine.
- Creative-spec lift can be small-N for some attributes — flagged "thin sample" when n < 10 on either side.
- The Hotel CQ composite is a tunable index, not a benchmark. The intent is to expose Hotel C's fragility under the loss of Creator N, not to claim a universal hotel-scoring methodology.
- Projected revenue is built from user-editable assumptions and is shown as a range. Every revenue number carries the MODELED · DIRECTIONAL chip.
File and reporting window.
JC × VN_ Data Set.xlsx