Measurement of sports sponsorship
Every sponsorship euro, proven.
Sponsor Proof finds your brand across the case’s channels, scores its quality with human-verified computer vision and estimates its value with a public methodology. Every number traces back to its proof.
Illustration: AI-generated photos and fictional brands; the boxes show how the detector locks onto sponsors, and the scores follow the public weights.
Social, web, creator live streams and online press
- TikTok
- YouTube
- X
- Twitch
- Threads
- Websites
- Online press
The problem
The market sells you a number: a single figure, a secret formula, accuracy claimed but never shown. Sponsor Proof hands you the proof, from the report total down to the post, the frame and the box around the logo.
- provenance levels on every number
- 5
- value is always a range
- P10–P90
- a versioned methodology, fingerprinted on every report
- v1.0.0
- data residency by default
- EU
How it sees
From frame to euro, in five steps.
lupi.official
Reel · 14/03
01 · Collection
1,284,902 views · Observed
- 01
Collection
We collect posts, videos, reels and live streams from the case’s channels, from declared sources: connected accounts, official APIs, third-party providers where needed, the rights holder’s original media. Each with its terms-of-use status.
- Source: connected account · L1
- Published within the case window
- 1,284,902 views · Observed
- 02
Detection
Computer vision finds the brand wherever it appears: pitchside LED boards, pitch graphics, big screens, banners. Even when it is small, angled or partly covered.
- Dedicated detector (vision-detect 2.0.0)
- One box per appearance
- Confidence per box
- 03
Quality
Every appearance gets a Quality Index: legibility on a phone, share of screen, clarity, position, clutter from other brands. A logo in the background is not worth a close-up.
- QI = L × S × Q × cap
- Public, versioned weights
- Illegible on a phone: zero value
- 04
Human review
An analyst reviews a statistical sample and every doubtful detection. False positives are dropped from the calculation; confirmed detections become Observed, and the verified share stays in plain sight.
- Stratified sampling
- False positive: discarded
- Accuracy gate before publishing
- 05
Value
Impressions, quality and the country’s market rates become an estimated media equivalent value. Always a range, never a single figure. And always next to cost, never called ROI.
- Country rate cards, with sources
- Monte Carlo with a saved seed
- Media value / window cost
Traceability
Click the euro.
Every figure in the report opens up. From the total to the channel, from the channel to the content, from the content to the frame where the logo appears. No number is left without proof.
Demonstration case with sample data
Media equivalent value attributed to the sponsor
Estimated62%
of the value rests on observed impressions
- Observed impressions · 62%
- Estimated impressions · 38%
Detections verified by an analyst94%
Declared uncertainty
Value is not a number. It is a range.
Every parameter carries uncertainty. We show it: low, base and high, and how much each assumption weighs on the total. Move the parameters: the simulation really runs, here in your browser.
Low · P10
€138k
Base
€168k
High · P90
€212k
What moves the total most
- Reference CPM−€42.1k / +€54.8k
- Quality Index−€33.7k / +€42.1k
- Estimated impressions−€29.5k / +€37.9k
- Attributed share−€14.7k / €0
The base value is the deterministic calculation with every parameter at its base value. The range runs from the 10th to the 90th percentile of a Monte Carlo simulation with a saved seed: same inputs, same result. Sample data.
Provenance
Every number tells you where it comes from.
Five levels, from strongest to weakest, on every figure in the report. In the header, the share of value that rests on observed impressions. A composite value always takes the level of its weakest input.
- Observed
Read at the source or confirmed by an analyst: likes, views, posts, followers via official APIs.
1,284,902 views
- AI-detected
Found by computer vision, with the share already verified by an analyst.
VESKA logo · 94% verified
- Estimated
Computed with a declared model: photo impressions, value via CPM.
≈ 2.1m impressions
- Declared
Provided by one of the parties: contract cost, media kit data.
Window cost
- Projection
Scenario or full season: useful for deciding, never presented as a measurement.
Full season
62%of the value rests on observed impressions
94%of detections verified by an analyst
- Observed impressions · 62%
- Estimated impressions · 38%
Accuracy
Accuracy is measured. Not claimed.
Others write “99%”. We publish precision and recall per brand, computed on a hand-annotated reference set, with confidence intervals. If a case falls below the threshold, the report says so.
Accuracy card
Reference set: at least 500 images and 30 clips per brand and asset
- VESKA● OK
- TREVIQ● OK
- OLMARE● OK
- ORVENA▲ recall
The comparison
What the market does not show you.
Traditional approach
Sponsor Proof
Methodology
Confidential
Public, versioned, fingerprinted
Value
A single figure
Low / base / high range
Provenance
Measured and estimated mixed
Five levels on every number
Accuracy
Claimed
Measured on a reference set, with confidence intervals
Proof
Reports with no access to the evidence
Every euro leads to the frame
Scope
Limits in the footnotes
“What we do not measure”, always visible
Side of the deal
Only one
Brand and rights holder, same snapshot
Time
Projects that take weeks
A case set up in minutes
Data
Residency rarely declared
EU residency by default
One case, two views
Same numbers. Two perspectives.
The brand and the rights holder open the same snapshot, with the same methodology. The question changes, the data does not: the negotiation starts from common ground.
VESKA × LUPI
2026 · Q1
- Media value / window cost
- 1.4–2.0×
- Share of media value from LED assets
- 46%
- Against your other sponsorships
- 2nd of 7
- Placement recommendation
- More time on the central LED
Sample data.
Ambush radar
Who is in your content without the right to be?
Sponsor Proof checks the brands detected in the property’s content against the category competitors declared in the case. Every possible ambush arrives with the proof and the content it appears in, and an analyst reviews it.
- Confirmed by an analyst
BRIXA Energy drinks
Banner in the stands · 14 club posts
- In review
PAGORA Payments
LED board in a club video · 3 posts
- Possible ambush
KAIRO Apparel
Press-conference backdrop · 1 post
Ask the data
Answers with sources. Always.
The assistant answers by citing metrics and evidence, with their provenance level. It lives in the platform and, through an MCP server, inside your own tools. Every number comes from the data, never from the model; the MCP server is read-only.
In March the LED delivered between €38k and €52k of media equivalent value (base €44k), 21% more than in February. The growth comes from two reels with over a million views each, with the logo verified in 96% of detections.
- LED value · March · Estimated
- Reel 14/03 · 1.28m · Observed
- LED detections · 96% verified
Public methodology
Every report carries its seal.
Formulas, weights, rates and sources are public and versioned. Every report carries the version and fingerprint of the methodology that produced it: anyone can check which rules it was calculated with. When the methodology changes, we show you the difference cause by cause.
Version
1.0.0
sha256 · 9f2c7e41a0d3
Sample fingerprint
The delta, explained
Illustrative example, indexed values (previous calculation = 100).
Previous calculation: 100; Error corrections: −14; Scope changes: −9; Measurement changes: +6; Coefficients and rates: −3; Methodology 1.0.0: 80
Scope
What we do not measure. In plain words.
A number is only credible if you know what it leaves out. Every report declares its scope and actual coverage, in a box that is always visible.
What we do not measure
- Broadcast, TV and OTT match streamsOut of scope. A possible future extension.
- Videos without the original filesIn-video exposure not measured: only the cover counts.
- Platforms without a sourced rateMeasured, but with no euro value.
- Stories without a connected accountNot measured: without a connected account they cannot be read compliantly.
- Business outcomes without connectorsOnly as a Projection, never as a measurement.
Europe
Born in Europe. Built for Europe.
Reports in five languages, with local formats and each country’s market rates. EU data residency and data protection by design. Sources declared channel by channel, with their terms-of-use status.
Osservato
- Data residency
- EU, by default
- Report languages
- IT · EN · ES · FR · DE
- Rate cards
- Per country, with source and date
- Sources
- Levels L1–L4, each one can be switched off
Who it’s for
For those who invest, who sell, and who advise.
- Brands
A measure you can take to the board.
Independent and defensible. Media value against the window cost, comparison across your sponsorship portfolio, solid ground to renew or renegotiate.
- Media value / window cost
- Portfolio and normalised comparison
- What-if and return on objectives
- Rights holders
Credible recaps for every partner.
Clubs, leagues, federations and events: a view of every sponsor, the media value of every asset, branded reports that stand up to your partners’ questions.
- A recap for every partner
- Media value per asset
- White-label reports
- Agencies
One method for every client.
Multi-client portfolios with the same methodology, country rate cards, branded reports and an agency view that keeps each client’s data apart.
- Multi-client
- Country rate cards
- API, webhooks and white-label
Put to the test
Let’s put your sponsorship to the test.
Pick a sponsorship, a window and the channels: we’ll show you how we measure it and what we can prove, with the evidence.