We pointed our own research engine at our own market

We built a research system to answer a question conventional search handles badly: what do business owners actually say about their own companies, at volume, in their own words, without a survey in the way.

This month we pointed it at ourselves.

We took the categories of work we get hired for: money that has to reconcile, customers who must never see each other, who someone is and what they may touch, work that runs unattended. Then we asked the corpus how often owners describe each one.

The answers were not what we would have guessed, and one of them reversed an assumption we had been working from for a year.

The part that matters is not the finding

It is that the question could be asked at all, and asked again later by somebody else. Every source page is preserved whole and addressed by a hash of its own contents, so each figure can be walked back to the posts that produced it months from now.

Most market research is an opinion with a chart attached to it. You cannot check it, because whatever was read to produce it is gone. This is a measurement we can hand to somebody who thinks we are wrong, along with the script, and let them re-run it.