TrendHive Analytics

TrendHive Analytics is a Shoreditch adtech outfit founded by three ex-agency planners who left a major holding company after a quarterly performance review in which they were told their methods were “creative but not sufficiently data-driven,” and decided to verify that claim by becoming the data. Fifteen people now, including the original three founders, two data scientists, one full-time copywriter, and seven people whose job titles change every six months due to ongoing reorganization.

Their product line is audience-segmentation and content-effectiveness models for mid-tier advertisers, focused on the parts of the addressable market the global agencies have decided are too small to bother with. The work is unglamorous and steady. The firm makes money — comfortable monthly Cu margins, not headline numbers. The founders, when asked, will note that their former employer no longer exists, which they consider sufficient validation.

The office has, on a back wall, a printed copy of the quarterly review that prompted the founders’ departure. It has been laminated. It is mounted slightly crooked, which the founders insist is intentional.

TrendHive’s models include, by default, a behavioural QA layer specifically designed to detect “manipulation drift” — instances where the model’s outputs are being optimized for engagement in ways that the firm considers ethically borderline. The layer is disabled by client request roughly thirty percent of the time. The firm has, internally, a quietly maintained record of which clients request the disable. They have not disclosed this list to anyone. They have also not deleted it.