Deal coverage grid

A coverage grid for the Notion enterprise team: twelve open accounts down the side, the five committee roles from the workspace’s own buyer-persona definition across the top, and every cell green (two or more people held), amber (exactly one — the deal hangs on a single thread) or red (nobody held). Accounts sort most exposed first. Clicking a cell opens the account and names who is held for that role, who Trayo found and is not held, why each of them matters and their work email. A second screen is the one-click list: the twenty people who close the twenty biggest gaps across the whole pipeline, one person per gap, red gaps above amber ones.

Trayo supplied all of it. The workspace was empty, so the account set came from POST /v1/find in direct mode against the saved ICP (software, internet publishing, financial services and IT services; US-headquartered; 800–15,000 employees) and twelve were imported with /v1/accounts/batch. One batched /v1/stakeholders/search/batch against the workspace’s own stakeholder definition scanned 229 candidates and returned 108 people with a per-persona coverage verdict for each account, in twelve seconds. That produced 31 missing roles, 22 single threads and 7 covered ones out of 60 cells, and 32 gaps with a named person behind them; 10 roles came back with nobody at all, which the screen states rather than fills. /v1/contacts/lookup found 31 of 32 work emails — one came back on a different domain and is flagged with Trayo’s own identity confidence, one returned nothing. /v1/people/{id}/ask was run 8 times, for the held contacts at Equinix and Morningstar, and its answers are the per-account email-history drill-down.

Two honest caveats. The workspace held no people, so 38 of the people Trayo returned were written into it as the roster a rep would already have in CRM — real names, titles and faces, but the choice of who counts as ‘already held’ is mine and is called out on the method screen; everything the grid reads downstream of that roster is genuine. The published version of this recipe scored ‘people I have actually emailed’, which needs a metered ask per person and does not survive a pipeline-wide grid, so coverage is held-versus-should-hold and the email-history read is a per-account drill-down; the grid is capped at twelve accounts because each costs one metered 5–20 second stakeholder search, with refresh offered per account. A lookalike search seeded from the workspace’s named customers failed twice with lookalike_search_failed and the criteria search stood in for it. Add to sequence, export, push to Salesforce, draft a first touch, refresh and read-email-history are previews that say what they would have sent and send nothing; the app reads a JSON snapshot and has no backend. On the palette: the library’s brand module made #000000 the primary and #FFFFFF the page, left #727272 out as an unused grey, and in dark mode filled the hueless primary with near-white #e8e8e8 so the accent reads — both reported on the method screen.