Som · 2 November 2026 · no numbers pulled · fill the blanks after the angle is chosen (W6-6), then send to Travis (W7-2)
This template is written before any data is pulled, so the rules cannot be changed to fit a result. It has no Hypertune FPS figure. Austin still needs editor access on a Google Doc copy. This page is the draft.
One row is one PC scan, not one account. If one account has more than one PC, each PC is its own row. Repeat scans of the same PC in the window collapse to one row: the latest complete scan in the window.
[angle] Who is in the sample (for example paying Windows users, early-access testers, or a recruited set of new PCs).
Staff, test accounts and incomplete scans are out. Say how an incomplete scan is recognised before the pull (a required field missing), not after.
One window, in UTC, with a start datetime and an end datetime written here before the export. The report states both dates next to every figure. No second window is added because the first one looked small.
[angle] Start: ____. End: ____.
Write the list before the pull. A row dropped for any other reason is listed in an appendix with the reason, and it still counts in the "looked at" total.
The unit is one PC. Shares are "of PCs in the sample", not of accounts and not of scans. A settings finding and a game test are separate units and are never added together. A Windows setting change is not a frame-rate result.
Every public number is a count and a share of the sample, with the sample size beside it. Do not quote a share for a group smaller than the privacy floor. Do not add a confidence interval unless Travis asks for one in the method check. If a cell is too small, the report says "not shown", not a rounded guess.
Only aggregates. No account, no email, no machine name, no licence key. Austin sets the floor (the smallest group we will show) before the pull. Until that number is written here, no cell is published.
Floor: ____ PCs, set by Austin on ____.
A cell under the floor is omitted. It is not described as "fewer than N" if that still points at a person or a single brand.
A zero stays in. A negative stays in. They are not dropped to raise an average. If the angle reports a change, the report shows how many PCs were zero, how many were negative, and how many were positive, each as a share of the same sample. The news story does not use Hypertune's own FPS gains as the hook.
It does not name the angle, the dates, or the privacy floor. Those wait on the chosen story and on Austin. Field names for a scan export wait on W6-4. Nothing in this file is a result.