Audience intelligence
Understanding who an audience really is, using consented data and measures we can explain.
We don't guess who an audience is, we ask, record the answer and show our working
Understanding an audience used to mean buying a cross-platform identity graph and trusting it. That model is ending. Google shut down most of the Privacy Sandbox APIs in October 2025 and kept third-party cookies in Chrome by default, which sounds like a reprieve but is not: Safari and Firefox block cross-site tracking by default, so 18.6% of global browser traffic is cookieless in 2026 regardless of what Chrome does, and deterministic identity across platforms is no longer something you can assume. The durable asset is first-party creator data gathered with consent that a person can withdraw.
How we work
We build audience intelligence from measures we can explain to the brand and to the person whose data it is. That means consented audience data held in one governed place, described in plain terms rather than opaque scores, and analyzed with methods whose logic we can show. When platforms report their own numbers, we read them for what they are: useful directional signal about that platform's slice of the journey, not truth about the whole. Meta's Andromeda and TikTok's ranking systems each see only their own surface, and each has an incentive to claim credit.
- Only data the creator has agreed to share is used.
- Comparisons are against a creator's own baseline.
- Estimates are labeled as estimates.
- Every figure can be traced to its source.
Where this is going
The direction of travel is toward modeled, privacy-preserving measurement, and toward audience understanding that survives a consent withdrawal without breaking. We design for that now. Data minimization is the default, purpose is specified before collection, and a Global Privacy Control signal is honored as a binding opt-out because in a growing bloc of US states it now legally is one. The result is an audience picture that is smaller and more honest than the identity graphs of the last decade, and considerably more defensible.
How we help
Consented data
Every data point we use comes with a record of the consent behind it. If a creator withdraws consent, the use stops, and the report changes to say so.
That discipline is slower than scraping. It is also the only version a client can defend in a review.
Explore related issues:
Audience profiles
Who watches, from where, at what hour and for how long, read across platforms rather than inside one. We separate reach from attention and say which one a number describes.
Profiles are refreshed on a schedule, with the date of the last reading on every page.
Explore related issues:
Decisions
Intelligence is only useful if it changes a choice. We tie every reading to the decision it informs: the creator shortlist, the channel mix, the timing, the budget.
When the data does not support a decision, we say that too.
Explore related issues:
Our thinking
