Methodology

Earned Media Value · how the numbers are produced

Sentia publishes two parallel EMV data sets every month. Each one answers a different question and uses a different source. Reading them side by side is intentional · it lets you sanity-check first party measurement against public consensus and vice versa.

The two data sets

Both families of values are stored in the emv_benchmarks table and refreshed on a monthly cron, keyed by the calendar month they describe (for example 2026-05).

Public consensus

Key: public_consensus:<YYYY-MM>. We average per engagement EMV rates published by named industry sources for the Italian market, currently Kolsquare, Ayzenberg, Buzzoole, IAB Italy, and Influencer Marketing Hub. Source pages are crawled with Serper and the rate cards are extracted with Gemini Pro · numbers are converted to euros, normalised per engagement type, and averaged at the platform-and-format level. We do not weight sources by provenance · each source contributes equally so that one provider cannot dominate the consensus.

Sentia portfolio

Key: portfolio:<YYYY-MM>. We compute first-party EMV from Sentia owned Meta and YouTube ad spend for Italian brands on a 120-day rolling window. For each ad we know the surface (Instagram, Facebook, YouTube, Reels, Shorts), the format (image, video, carousel), the spend, and the resulting on-platform engagements. The portfolio EMV per engagement type is the spend-weighted cost per outcome, taken across the cohort. Engagement types we measure first-party are reactions and views · comments, shares, and saves are currently mirrored from public consensus for most surfaces and will progressively switch to first-party as the ingestion pipeline matures. Note that "reactions" is the canonical counter on every surface · Meta returns six reaction types (Like, Love, Haha, Wow, Sad, Angry) that already aggregate into the field, and IG, TikTok, and YouTube use the same underlying counter for their respective heart and thumbs-up reactions.

Why two data sets

  • Public consensus is calibration-friendly. It anchors Sentia numbers against what analyst houses publish, so a CMO can compare like for like with the rate cards she already sees in pitch decks.
  • First-party portfolio is decision-friendly. Spend-weighted cost per outcome is what an actual brand pays this quarter, on this format, on this surface · it moves with the auction, with creative quality, and with audience saturation.
  • Drift between the two is itself a signal. When portfolio rates run cheaper than consensus, the buy is performing above market. When they run more expensive, either the cohort is paying a premium for quality or the consensus is stale · either way, the comparison surfaces the question.

Refresh cadence and provenance

Both data sets refresh on the first of every month, on a Trigger.dev schedule. The portfolio recalibration runs against the latest 120 days of owned spend. The public consensus refresh re-fetches each source, drops anything older than the previous quarter, and requires at least two non-stale sources before publishing a row. Every published row carries the source list and the date it was measured · you can read the file-level metadata in the EMV benchmarks table.

What you can do with this

Use the public consensus as the conversational benchmark · the number you cite in a deck or against an analyst report. Use the portfolio number to underwrite a decision · what an in-market buy-side actually costs this quarter for this combination of platform and format. The two are complementary, not redundant.