The grid publishes named states
The coverage matrix is a table of detector classes against capture channels, with a state in every cell and a version on the artefact. Readers often arrive wanting a partial-credit summary and a count of what is absent. The grid declines both. Its legend has three names: a detection class that runs in the product today, a path that is simulated, and a path that is roadmap.
A simulated cell means the code path is real and has been run by a test harness, and the label stays on the cell. A roadmap cell means the behaviour is specified and unbuilt. Paths outside the product are a named list under the grid rather than a softer colour in a cell.
A cell is the weaker of its row and its column
The page computes each cell: it takes the detector’s state and the channel’s state and keeps the weaker of the two. A detector that runs today, placed on a channel that is only a specification, produces a cell that reads as the channel. That arithmetic is why a strong row cannot quietly upgrade a weak column.
Eight identifier, payment, and secret classes are marked as running in the deterministic tier. Two further rows, crown-jewel exact match and source code, are roadmap at the detector itself, so every channel on those rows stays roadmap. Paste and typed input are simulated columns. Drag-and-drop, file attach, and print are roadmap columns. No capture column is marked as running on a live endpoint, so no cell in those columns can read as if it were.
Shipped means a cell, under this legend
Inside the grid, the strong detection state means the class is in the detector registry and the verdict service can run it. A production fleet and a purchase contract sit outside that cell. The word is local to the legend printed on the same page. Quoted without the legend, it should be dropped.
The artefact is versioned. A cell is meant to change with a new published version. Quoting a cell in an evaluation means quoting the version as well. A later build of the browser path, still described as renderer-assisted and weaker than the enterprise connector, does not authorise a reader to repaint a simulated column.
Simulated and roadmap are the gaps the grid is for
The simulated columns show detection on input from a harness rather than a live fleet. The endpoint inventory behind any such story is a seeded stand-in. The harness can feed the real detector. It cannot prove heartbeats, tamper detection, or coverage of machines the organisation actually owns.
Roadmap columns are the capture methods the specification names and this phase does not run: content dropped on a page, a file attached for extraction, and print. Leaving those cells visible is the point of publishing the grid. Hiding them to improve the impression removes the information an evaluator uses to scope a test. Traffic the product does not intercept, including desktop applications and direct API callers, can still be discussed as inventory from proxy and DNS logs, and in this build that inventory is fed by a synthetic log source.
Read it so the gaps stay on the page
The detector column stays put while the channels scroll, including on a narrow screen, so the class name remains in view beside a distant cell. Print and review should keep that column. A matrix that slides its gaps off the page, or that collapses to a score in the corner, is a different document from the one that was published.
There is no percentage, no “N of M channels covered”, and no single maturity mark across the grid. Those numbers would invent a confidence the cells leave unstated. An evaluation quotes the cells it exercised, names the version, and leaves the other cells in the picture. The unfinished ones are part of the evidence.
How to read this
| Read it as | Which means | Do not read it as |
|---|---|---|
| A cell | Present, partial, or absent | A contribution to a score |
| Absent | Part of the evidence | A cell to hide |
| “Shipped” | Means the cell is in this POC | A contract term by itself |