Agentic advertising runs on real-world context: what every place is like and who is around, hour by hour. We feed it to your agents so they find your next customers, with no cookies and no device IDs.
Talk to us about 168 VEAgentic advertising is here.
Agents now decide who to reach, in real time, with no cookies and no device IDs. The IAB Tech Lab calls it agentic advertising. An agent is only as good as what it knows about the real world, and the real world is exactly what we provide.
IAB Tech Lab · Agentic AudiencesThe structured layer is for explainable targeting: audiences you can open up, audit, and defend to a client. The embedding layer is for performance: the model learns what converts and optimizes toward sales, leads, and cost per action. Both come from the same context.
Who is around, by sector, hour by hour. It lands as plain BigQuery columns. You can audit every number, show a client exactly why an audience is what it is, and trace it back to public records.
The same context, compiled into a vector. 168 VE gives your model a learned representation of where and when, so it learns what actually converts and optimizes toward sales, leads, and cost per action. The Master Embeddings Model brings the data you already own into the same latent space, so your vectors and ours speak one language.
The same idea the rest of the industry is adopting for content and audiences, applied to the real world. What we learn is the cohorts of buyers inside every place and hour.
Every place at every hour gets read for the cohort that is there: who they are and how they behave, from years of history out to a forward forecast.
168 VE learns the complex cohorts inside every context, the full mix of who is there and how they behave. The cohorts that act like your buyers sit close together, even in cities it was never calibrated on.
The Master Embeddings Model brings the data you already own into the same space: your first-party audiences, your creative, the IAB taxonomy. Your cohorts and ours end up speaking one language, ready for any screen or bid.
Send us a file of what already worked: sales, leads, or store visits, each with a time and a place. We learn the cohort behind your best customers, find the cohorts that behave like them across the country, and hand back the places and hours to reach them. No device IDs, no personal data.
A simple file: each sale, lead, or visit with when and where it happened. Nothing personal leaves your side.
168 VE turns each conversion into context, and the Master Embeddings Model lines it up against the rest of the country to learn the cohort your best buyers share.
The cohorts that behave like your best customers, and the places and hours to reach them, ranked and ready to run. Expand into new markets and keep the lead pipeline full at a lower cost per action.
Times Square · Tuesday
At 2pm it is a finance and office crowd. By 11pm it is hospitality and tourism.
168 tells you which, by sector, by hour, with no device IDs. Buy the right screens at the right hours, and prove the audience from the same table.
Buy the screens your lookalikes pass, by hour and by sector. Audience verification without device IDs.
A pre-bid context signal that holds in markets where panels have gone thin.
Context for non-addressable inventory, mapped to the taxonomy buyers already use.
Feed 168 VE to your bidder so it optimizes toward sales, leads, and cost per action.
The priors were built on public New York data, then tested against two cities they were never calibrated on. Five independent public sources with no shared lineage, so the validation stays independent. The dataset is built without device data, which means it strengthens as privacy tightens instead of breaking.
Read the validation report168 is a verified vendor on Google BigQuery Marketplace. Subscribe in a few clicks and it bills through the GCP account you already use. Nothing to move: query it right where your data lives, and your own data never leaves your side.
250 Census attributes per cell. The free context floor, on BigQuery Marketplace, no signup.
The full hourly context for any US metro. Every cell, every hour of the week. Manhattan is free.
The vector embedding of real-world context. A learned representation your model queries directly.
Brings your own data into our latent space and aligns it with 168 VE, so your vectors and ours speak one language.
The new way to target. Start free with Manhattan, on Google BigQuery Marketplace.
Book 20 minutes
