Skip to content

Rating and Review System

Platform artifact — instantiates Reputational Signal Governance

Collects ratings and reviews from counterparties after each interaction and publishes them as an at-a-glance standing signal.

Rating and Review System is the front door of the reputation pipeline: it elicits first-hand feedback from the very people who just transacted, and publishes it where the next person can see it. Its whole job is to manufacture the raw signal by asking at the right moment — a star rating, a thumbs, a written review tied to a specific interaction — and then to control how that pile of feedback is displayed. What makes it this mechanism and not its siblings is that it neither computes a single number nor checks whether the reported facts are true; it captures subjective counterparty experience and renders it. Everything downstream — scoring, verification, abuse screening — consumes what this artifact collects.

Example

A home-services marketplace pairs homeowners with independent plumbers. The moment a job is marked complete, both sides get a prompt: rate 1–5, optionally leave a few sentences. The review is stapled to that specific booking, then surfaced on the plumber's profile — average shown up top, most recent reviews below, filterable by job type. A homeowner comparing three plumbers for a burst pipe doesn't read a résumé; they skim twenty recent reviews and pick the one with steady, detailed, on-time reports. The signal that decided the hire never existed until the system asked for it at the moment the work finished.

How it works

What distinguishes it is solicitation and display, not calculation:

  • Prompt at completion. Feedback is requested at the moment of a real, logged interaction, so each review is anchored to something that actually happened rather than volunteered out of nowhere.
  • Structured plus freeform. A comparable core (stars, tags) sits alongside open text — the first is easy to skim and aggregate, the second carries the context a number can't.
  • Render to an audience. The system decides sort order, recency weighting, and what a given viewer sees — the display side of the signal.

Tuning parameters

  • Solicitation timing and pressure — when and how hard you ask. Harder prompting lifts response rate but pulls in low-effort or coerced ratings.
  • Structured vs. freeform weight — stars and tags are comparable and gameable; written reviews are rich and hard to summarize. How you balance them shapes what the signal rewards.
  • One-sided vs. two-sided — letting both parties rate each other curbs one-sided smear, but mutual visibility can make each side afraid to post an honest negative.
  • Recency vs. volume in display — foreground the latest reviews (adapts fast, noisier) or the all-time body (stable, slow to reflect change).
  • Audience scope — public, logged-in-only, or parties-to-the-transaction. Wider visibility makes the signal more useful and raises the stakes of gaming it.

When it helps, and when it misleads

Its strength is that it is the cheapest way to turn thousands of private experiences into one shared, browsable signal — and it gives a newcomer choosing a stranger something concrete to go on.

It misleads mostly through who bothers to speak. Reviews self-select toward the delighted and the furious, so ratings pile up at the extremes and the quiet middle goes unheard; a brand-new subject with no reviews faces the cold-start problem[1] and can't be judged at all; and fear of retaliation quietly suppresses honest negatives. The classic misuse is curating the displayed set — quietly gating or reordering which reviews appear — to manufacture a flattering profile, running the mirror backwards as marketing. The discipline that guards against it is to publish every qualifying review, bind each to a verified interaction, and hand authenticity screening to a detector rather than deciding case by case what to show.

How it implements the components

  • behavior_trace_capture — solicits and records the rating or review attached to each completed interaction; this is the capture step for the subjective side of the signal.
  • visibility_and_audience_boundary — governs what is shown, to whom, and in what order — the display and audience-scoping side.

It does NOT weight evidence quality or attribute verified facts (that's Verified Transaction History), aggregate the feedback into a score (Reputation Score or Standing Index), detect fake or coordinated reviews (Sybil, Collusion, and Brigading Detection), or adjudicate disputed ones (Appeal and Correction Workflow).

Notes

Keeping capture separate from scoring is deliberate: because this artifact only gathers and shows the raw signal, a team can fix a bias in how feedback is collected — change the prompt, add a verified-purchase flag — without re-tuning the score that sits on top of it. Conflating the two is how a display tweak silently becomes a scoring change nobody reviewed.

References

[1] The cold-start problem — a new subject (or a new platform) has too little history to be rated meaningfully, so early signal is thin and volatile. It is why rating systems lean on verified interactions, provisional badges, or borrowed evidence until a real track record accrues. ↩