By Artem Korsakov ·
The review "everything was great, I recommend" is pleasing to the performer. For a buyer of a complex project, it leaves almost all questions open. What was done? Who performed the work? What was considered a good result? Why is the review author satisfied? Without these details, five stars are difficult to turn into a decision.
When rating truly influences choice
In 2011, Michael Luca from Harvard Business School matched Yelp ratings with the revenue of restaurants in Washington state. An increase in rating was associated with an increase in revenue for independent restaurants, while ratings had almost no effect on chain restaurants: buyers already knew enough about them. Source: Harvard Business School, 24 October 2011.
The study is old and relates to restaurants, so its conclusions cannot be transferred directly to B2B contractor selection. The core idea is useful: a reputational signal is more important where the buyer knows little about the performer. It remains to understand what information they actually receive from five stars.
Why identical stars hide different experiences
A small website and the integration of several working systems may receive identical ratings. However, in one case, the client evaluated launch speed, and in another, data exchange reliability and support. Comparing average ratings without the task and scale of risks is dangerous.
Even project duration requires context. The work might have been delayed due to missing access, team changes, or the performer's error. A general mark of "delivered on time" does not show how the parties counted the deadline and what happened to additional requirements.
Which fields make a review useful
Ask the author to answer specific questions:
- what task was the project solving;
- which part of the work was the contractor responsible for;
- what was accepted by the client;
- what constraints were encountered;
- what was convenient and what should be improved;
- who a similar solution would suit.
It is not necessary to turn every review into a long questionnaire. A short text can be shown, with details revealed upon request. However, key context must remain available next to the rating.
How to confirm roles without excessive promises
If a specialist led the implementation, the card should state exactly that. Developing it themselves, process design, contractor management, and acceptance control provide different experience. All these roles are useful if the buyer does not have to guess them.
Confirmation from the client must relate to specific claims. For example, a client representative approved the task description and team participation, but did not confirm the calculated savings. These fields receive different statuses. The system saves a version of the text for which confirmation was given.
If the author adds a new number after that, the old confirmation should not automatically extend to it. Otherwise, the product creates trust in what no one has checked.
What to do with negative experience
Instead of automatically deleting an unpleasant review, a clear procedure is needed: a request for explanations, the ability to respond, verification of the author's connection with the project, and a moderator's decision with reasons. A dispute over quality does not prove fraud.
For the first version, it is sufficient to limit reviews to confirmed participants and specific questions. Someone who understands conflict situations will still be required regardless. AI can group complaints and find messages related to them, but should not independently assign blame.
How to turn reviews into a working selection tool
Offer buyers filters by tasks, roles, and systems. Show several relevant projects instead of a general rating place. If there are few confirmations, display this honestly. A new performer does not become bad simply because they have not yet accumulated a public history.
Look at the experience verification prototype. Its usefulness should be verified by whether the buyer was able to select relevant questions and narrow down the list of performers. The number of stars on the screen says nothing about the quality of the choice.
Related reading: Contractor work quality: why it is difficult to prove before a deal; Business transparency: why disclose clients, connections, and terms.
Частые вопросы ### Is it necessary to completely abandon stars?
No. They can be a quick reference if accompanied by the number of ratings, period, and context. But for a complex purchase, it is worth opening the actual projects and understanding what lies behind the final mark.
Can we entrust AI to rate a contractor?
For the first product, it is better to use AI to search and briefly retell confirmed information. An automated final rating requires clear criteria, error checking, and the ability to appeal. A beautiful number without this rather complicates the choice.