DABYTE INDEX · ARCHIVE
Every measurement DABYTE has published, and the raw file behind each one
Answer. DABYTE has published 9 measurements of the SaaS & AI Tools index, covering 2026-07-23 to 2026-09-07. Every one is kept at a permanent address as the exact JSON it was published as, so any change we report can be recomputed from the two files rather than taken on trust. Nothing here is edited after publication; corrections appear as a new measurement.
Latest: ·
full series: /api/history.json ·
how it is measured
Published measurements
| Measured | Brands | Top brand | Panel | Brands that moved |
|---|---|---|---|---|
| 2026-09-07 | 20 | Slack (52.0%) | v3 | 13 |
| 2026-08-31 | 20 | Slack (45.8%) | v3 | 12 |
| 2026-08-26 | 20 | Slack (43.7%) | v3 | 14 |
| 2026-08-17 | 20 | Slack (45.8%) | v3 | new panel |
| 2026-08-10 | 20 | Salesforce (31.3%) | v2 | 17 |
| 2026-08-04 | 20 | Slack (33.3%) | v2 | new panel |
| 2026-08-01 | 20 | Notion (41.7%) | v1 | 10 |
| 2026-07-28 | 20 | Slack (25.0%) | v1 | 11 |
| 2026-07-23 | 20 | Notion (33.4%) | v1 | — |
These files are the evidence for every delta shown anywhere on this site. If a number on a brand page disagrees with the archive, the archive is right and we want to hear about it.
Why the panel column is there
A share of answer is a share of a specific list of questions. Change
the list and every number changes without a single engine having changed its mind. So each
measurement records which version of the panel produced it, and we refuse to compute a change
across two different versions — the archive marks those rows new panel rather
than printing a movement that did not happen. A comparable series therefore starts at the
first measurement of the current panel, not at the first measurement ever.
This archive spans 3 panel versions (v1 first measured 2026-07-23, v2 first measured 2026-08-04, v3 first measured 2026-08-17). Rows measured under different versions are never compared to each other anywhere on this site. The current question list is on the methodology page and inside every dataset file.
Why the series matters more than any single measurement
A single snapshot tells you who an engine named this month. It cannot tell you whether that is a position or an accident, whether a brand is climbing or sliding, or whether anything a company did actually changed what the engines say. Only a series does that, and a series cannot be bought or backfilled — it accrues or it does not exist. Most published AI-visibility benchmarks are one-off snapshots; this page is the part that is hard to copy.
Machine copies: /api/history.json
for the full per-brand series, and /archive/<date>.json for any single
measurement exactly as published. CC BY 4.0, no key.