Methodology
The method behind every number.
How we capture and classify every ad across seven media, and how we measure the response your spots create — documented with our partners and audited against evidence.
2015
Measuring advertising since
24/7
Continuous signal capture
7
Media in one taxonomy
100%
Records backed by proof
Quality loop
Every record goes through the loop.
Follow one television spot from the signal to the published record. Each step adds what the next one needs — and anything uncertain stops for a person.
01 / 06
Capture
The broadcast is recorded around the clock, with the channel and the exact second it aired.
- Medium
- Television
- Channel
- Channel 2 · Evening news
- Aired
- 20:14:08 · 30s
- Creative
- Waiting
- Match confidence
- Waiting
- Taxonomy
- Waiting
- Review
- Waiting
- Evidence
- Waiting
- Version
- Waiting
Two methods
What ran, and what it delivered.
Monitoring and Media Attribution answer two halves of the same question. Each has its own method, and both follow the same rule: show the work.
What ran across the market
- 01CaptureBroadcast and streams recorded around the clock. Pages, feeds and publications read the way people see them.
- 02RecognizeImage, sound and language read together, so the same creative is matched across media, formats and markets.
- 03ClassifyOne six-level taxonomy, from sector to product, applied the same way to every medium.
- 04ValueEstimated investment from the rate cards agreed in each market.
- 05ProveEvery record ships with its video, audio or screenshot, its position and its context.
What your spots delivered
- 01Spot logDate, time, timezone, channel, creative and program for every airing — ideally with cost and reach.
- 02ResponseWeb and app events by the minute, from the Decidata tracker, Google Analytics or a file.
- 03BaselineWhat each minute would have looked like without the spot.
- 04IncrementalResponse minus baseline, split into direct and indirect — so a busy hour is never mistaken for a great spot.
- 05Credit and costShapley, first, last, position-based or linear models across paths, and the cost of every incremental result.
Principles
What we never trade away.
Evidence first
No record without proof. Every insertion is kept with the evidence that lets anyone check it.
Multimodal by design
Image, sound and language are read together. A single signal is never enough to understand an ad.
Humans in the loop
Models propose, analysts confirm. Anything uncertain is reviewed before it is published.
One taxonomy
The industry's own classification, applied the same way to every medium and versioned over time.
Independent
We do not sell media. We measure it.
Our own technology
From signal capture to platform, the technology is ours — so we can explain and improve every step.
Transparent with partners
Coverage, methods and corrections are shared with clients and industry bodies, with adjustment windows and a change log.
We measure ads, not people
Viewing profiles are synthetic. No personal data about audiences is collected.
FAQ
Questions about our methods.
01Where does the evidence come from?
From our own capture. Broadcasts and streams are recorded, and pages, feeds and publications are read the way people see them. Each record keeps its video, audio clip or screenshot.
02How are new creatives found?
When a new spot airs, the system discovers it as it repeats across channels — no manual seeding. Anything the models are not certain about is confirmed by an analyst before it is published.
03How is investment estimated?
From the rate cards agreed in each market, applied insertion by insertion and kept comparable across media.
04What happens when a record is wrong?
It is corrected, logged and versioned. Clients and industry bodies can see the change log, and corrections follow adjustment windows agreed with the market.
05How does attribution separate a spot from normal traffic?
It models the baseline you would have had in each minute without the spot and credits only the response above it, split into direct and indirect.