Market Analytics

How far back the corpus reaches

Reports per quarter, plotted on the date the problem was reported — the source post’s own timestamp, not when we classified it. Useful for judging whether a date filter will return anything; the corpus thins out fast before about 2022.

This is not a market trend, and 2026 is excluded. Crawling began in the second quarter of 2026 and back-filled whatever was still reachable, so that single quarter holds 20,518 reports against fewer than 400 in any quarter before it. Plotting it would say “business problems exploded in 2026” when it means “we switched the crawler on”. The same recency bias shades the earlier years more gently: older posts are simply harder to find. Read this as coverage, never as demand. For comparisons that do hold, use the share-shift figures below, which divide the collection bias out.

Where people say they would pay

Ranked by the share of problems where someone stated outright that they would pay for a fix — the strongest demand signal in the corpus, and a rare one at 3.7% overall. Industries with fewer than 50 reported problems are excluded, since a single comment would otherwise top the table.

Where industry meets problem type

The densest intersections in the corpus. A dark cell means many independent people in that industry described that kind of problem — which is exactly where a narrow, vertical product has room. Click any cell to see the underlying problems.

Gaining and losing ground

Change in each category’s share of all reports, last 12 months against the 12 before. Share rather than raw count on purpose: total volume grew enormously as collection ramped up, so a count comparison would show everything “growing” several thousand percent and mean nothing. This is a modest signal on a small prior base — read it as a hint, not a finding.

What people say they are using when they describe something failing them. Product-name variants are folded together, so QuickBooks Online and QBO count once.