Signal Score Explained: How Forenta Measures Builder Reputation

Most reputation scores work until they stop working. On major labor platforms, average ratings have climbed steadily for two decades, not because work is getting better, but because the rating system itself creates incentives that push scores upward. When 85 percent of contractors receive the highest possible rating, the score tells you almost nothing about the variance in quality that actually matters.

Research by Filippas, Horton, and Golden tracking over 20 years of reputation data on a major online labor platform found exactly this pattern: the share of contractors receiving top ratings rose from roughly 33 percent to 85 percent over the period.¹ The signal did not improve. It inflated until it collapsed.

Why rating systems inflate

The inflation problem has a structural cause. On most platforms, reviews are sequential: one party submits their rating, the other sees it before responding. This creates a strong incentive to rate strategically rather than honestly. Fradkin, Grewal, and Holtz studied this on Airbnb and found that switching to simultaneous review submission, where neither party sees the other’s review before writing their own, significantly reduced positive bias and increased the informativeness of reviews.² The mechanism is simple: when you cannot see what the other party said about you, you rate what you actually experienced.

Gaming compounds the problem

Fake and incentivized reviews compound the inflation further. Research by Hollenbeck and colleagues documented systematic gaming of review systems on major platforms, with sellers generating positive reviews and suppressing negative ones through coordinated requests, review exchanges, and in some cases direct purchase.³ Luca and Zervas (2016) found similar patterns on Yelp, where competitive pressure drove a meaningful share of review fraud.⁴

When the share of top ratings goes from 33% to 85%, the score stops telling you anything useful about the bottom third. Rating inflation is not a bug in poorly designed systems. It is the predictable outcome of most rating system designs.

What Signal Score is built from

Signal Score takes a different approach by drawing from behavioral signals rather than relying on a single self-reported rating.

Trial completion is the strongest single factor: did the builder finish what they committed to, in the agreed timeframe, within the agreed scope? This is measurable, difficult to fabricate, and directly relevant to what project starters care about. Collaboration ratings add a second layer covering communication, responsiveness, and the quality of the working relationship. Both parties rate after the trial completes, simultaneously, without seeing each other’s rating first.

Profile completeness contributes a smaller weight. A complete, honest profile is a low bar, but it signals engagement with the platform. Consistent, relevant platform activity matters more than login frequency alone. Referrals from other high-signal members carry weight proportional to the referrer’s own track record: a recommendation from someone who has completed multiple successful trials means something different from one given by a new member.

What Signal Score deliberately excludes: self-reported credentials, paid visibility boosts, and endorsements not connected to real collaboration. The score reflects what a builder has done on the platform, not what they claim about themselves.

Tiers, multipliers, and what they mean

The tier structure, Rising, Active, Trusted, reflects cumulative track record rather than a current snapshot. A builder moves from Rising to Active through sustained good work across multiple trials. Reaching Trusted means demonstrating consistency over time, not just one strong engagement. A single exceptional collaboration could be an outlier. A track record across many is less likely to be.

Plan multipliers (1x for Lite, 1.5x for Pro, 2x for Max) affect how Signal Score is weighted in matching. The important thing to understand about what the multiplier does not do: it cannot substitute for an underlying track record. A Max-plan builder with a thin genuine signal will not outrank a Lite-plan builder with a strong one. The multiplier amplifies what is already there.

Why it takes time, and why that is by design

Every new builder on Forenta starts at Rising, regardless of external reputation, years of experience, or credentials. This is deliberate. A score that could be seeded with external credentials would be gameable from the start. Someone could import a curated selection of prior work to inflate their starting position.

A score that requires real collaboration on the platform to build means that when you see an Active or Trusted builder, the score reflects something that actually happened here, with real counterparties who rated the experience honestly. The honest trade-off is that new builders with genuinely strong skills face an initial period of building track record. The alternative, fast-tracking scores based on self-reported credentials, would undermine the signal for everyone.

What this analysis does not cover

This post describes Signal Score’s design principles and the research that informs them. It does not disclose the specific weighting formula for each input, which will be refined as the platform learns more about what actually predicts good collaboration outcomes. It does not address edge cases, such as how the score responds to disputed trials or incomplete trials that end by mutual agreement. Signal Score is version one. Expect updates.