Our course satisfaction score is 4.71. But there is another number we trust more

In an earlier article, we wrote that surveys and behavioral logs measure different things: a survey asks "did you try it?", while a log records "did you actually put it into your work?"

We thought it was only right to hold ourselves to the same standard.

Our course satisfaction score is 4.71 (out of 5). Our communication satisfaction score is 4.68. Over 5 years, we have delivered 116 courses and consulting engagements combined, with a cumulative total of 1,000 or more learners.

But satisfaction is a survey. It is a score given right after a course ends, while the good feeling is still there, anonymously, and without having to spend money again. Having written that people should not take other companies' survey figures at face value, it would be inconsistent for us to lead with only our own survey figures.

The number we trust more

11 organizations have asked us back. The longest relationship has lasted 4 years.

There is only one reason this number is better than satisfaction: a renewal is a decision to spend budget again.

• In a survey, individuals answer anonymously. For a renewal, the person in charge puts their name on an internal approval request.
• A survey is filled out right afterward. A renewal is decided months later, after the content has been tried in actual work.
• If a survey is wrong, nothing happens. If a renewal is wrong, the person in charge is accountable.

That is why we trust the number 11 more than 4.71: it is a small number, but one confirmed by behavior.

But 11 is not complete evidence either

We want to be honest about this.

11 is a small number. It is not large enough to treat statistically, and we have no benchmark to compare it with an industry average. That is why we did not write "our renewal rate is high": if you cannot disclose the denominator, you should not state a rate.

Some organizations did not come back. Some engagements ended after one time, and we do not know all of the reasons. They may have been satisfied but lacked budget, the person in charge may have changed, or we may have fallen short. Renewal as a metric does not tell us the reasons behind it either.

And this is a record of courses and consulting, not a record of an AI product. It should not be used as grounds for expecting the tools we are building to perform the same way. They are different kinds of work.

So what should you look at when choosing?

The purpose of this article is not to ask you to trust us. It is that whoever you choose, you can ask the same questions.

Whether it is education, consulting, or AI adoption, there are three things you can ask when you receive a proposal.

1. Ask about repeat business, not satisfaction.
"How many organizations have used this service two or more times?" Satisfaction is written as high in almost every proposal. It does not tell providers apart.

2. Ask for the denominator behind the numbers.
"How many out of how many?" If there is only a rate and no denominator, that rate has not yet been confirmed.

3. Ask about the cases that did not work out.
"Why did the organizations that did not come back decide not to?" If they have an answer ready for this question, they have actually looked into their own records.

We also try to stay in a position where we can answer these three questions. What we have written above is the full extent of the answers we have right now.

---

If you are deciding where to start applying AI
A 2-minute diagnostic lets you see one task to tackle first, a starting plan for a 2-week pilot, and the risks a person must check. Feel free to answer only as much as you are comfortable with.

STAR-T AI Business Operations Diagnostic →

---

Footnotes

① References and verification date
All figures in this article come from our own records; no external statistics or research are cited. Verified on 2026-08-11.

• Course satisfaction 4.71 / 5 · Communication satisfaction 4.68 / 5 (cumulative, from our own evaluation tally)
• Courses + consulting 116 engagements (5-year cumulative, from our own performance ledger)
• Learners 1,000 or more (cumulative, from our own performance ledger)
• Organizations that engaged us again: 11 · Longest relationship: 4 years (cumulative, from our own transaction records)

We have not disclosed clients' names, industries, or regions. This article discloses only the aggregated count (11 organizations) and the longest relationship length (4 years), and we have not included any description that identifies or hints at a specific client relationship. We believe information that involves a counterparty is not ours alone to disclose.

② Figures we did not use
These were deliberately left out because they fell short of our verification standard or were defined differently.

• Average relationship length — Our internal materials listed it as "an average of 3.2 years," but when we recalculated it from the raw data while writing this article, no calculation produced that value. With 11 clients, an average of 3.2 years would require the durations to add up to 35.2 years, but the durations in our records are whole numbers of years, so such a total is impossible. We also could not find any document showing how it was calculated. So we left out the average and included only the longest relationship of 4 years, which matches the raw data. Having written in this article "do not use a rate if you do not know the denominator," we could not gloss over our own average. We have put that figure on hold for internal citation.
• Renewal rate (%) — Because we decided not to disclose the denominator (total number of clients), we cannot state a rate.
• Consulting for 110 or more institutions — This is a registered figure, but its definition differs from "11 organizations" (all clients including one-time engagements vs. repeat engagements). Placing both in the same paragraph would make 11 look small or 110 look inflated, so this article uses a single basis: repeat engagements.
• 17 financial-sector institutions where we have taught — This is a registered figure, but it is on a different axis from this article's topic (repeat engagements), so we left it out.
• Number of projects we built and operate ourselves — This figure has not been validated by outside companies, so we do not use it as evidence of trustworthiness.
• Revenue figures — Withholding tax on lecture fees is handled separately, so this is not total revenue, and it is unrelated to this article's argument.

③ How this was made
The draft was created with AI, then verified and edited by people. Only figures confirmed against our own records in ① above were included in the body, and before publication the article went through three reviews: facts, tone, and legal. No generative images or audio were used.