"Keep, watch, or replace: how the verdict is reached"

One verdict per sub

When you run the Subcontractor Analysis app, every sub on your books gets one of three verdicts: keep, watch, or replace (shown on screen as "Review or replace"). The verdict is not a vibe or a black box. It is a short, fixed ladder of checks over your own data, your logged performance on their jobs, the backcharges you have absorbed, their certificate of insurance status, and what they cost against the value of the jobs they touch. Each verdict comes with a one-line reason that names the dollars behind it, and a dollars-at-stake figure that ranks the list worst-first, so the sub most worth your attention is always at the top.

Nothing here is peer data. The verdict is built entirely from your own book: your invoices, your COIs on file, and the engagements you have rated. See how a sub's reliability score is built for the score the first rungs of the ladder read.

The ladder, top to bottom

The checks run in a fixed order, and the first one that matches decides the verdict.

1. 1Replace: reliability is at risk. A sub whose reliability score has fallen below 50 reads as at risk, and that alone is a replace verdict. The reason line reads like "Reliability at risk, $4k in backcharges" when backcharges exist, or simply "Reliability at risk." when they do not.
2. 2Replace: the COI is expired on a meaningful amount of work. An expired certificate of insurance is a replace verdict once the sub has billed more than $5,000 of work. Below that, an expired COI still flags them (step 4 territory does not apply, but the compliance views surface it), it just does not by itself justify the replace call. The reason reads "COI expired on $18k of work."
3. 3Watch: reliability is slipping. A reliability score from 50 to 74 reads as watch. The reason is "Reliability slipping," with the backcharge total appended when there is one.
4. 4Watch: the COI is expiring soon. A certificate inside its renewal window produces "COI expiring soon."
5. 5Watch: no COI against real spend. A sub with no certificate on file at all, who has billed any work, produces "No COI on file on $7k of work." You are carrying their liability on every job they touch.
6. 6Watch: they cost a large share of the jobs they touch. When a sub's billed cost reaches 40 percent or more of the combined value of the jobs they worked, the reason reads "Costs 43% of the value of the jobs they touch." That is not automatically bad, some trades are simply the job, but it is a margin exposure worth knowing about.
7. 7Keep. Everything else is a keep. A rated, compliant sub reads "Reliable and compliant." A sub you have never rated reads "No issues flagged; not yet rated," which is an honest keep, not an endorsement.

Dollars at stake, and how the list is ordered

Every verdict carries a dollars-at-stake figure, a rough read on what leaving things as they are could cost you:

• Reliability at risk: the backcharges you have already absorbed, plus about 15 percent of everything that sub has billed you, the share of their work treated as exposed if the pattern continues.
• COI expired on real spend: about 20 percent of the work they have billed, the exposure of running uninsured trades on your jobs.
• Reliability slipping: the backcharge total itself.
• The other watch reasons and every keep carry no dollar figure, because there is nothing costed to put behind them yet.

The verdict list sorts replace first, then watch, then keep, and inside each band by dollars at stake, then by how much the sub has billed. The result is that the top of the list is always the most expensive problem, not the loudest one.

Where the verdict shows up

The Verdict tab of the Subcontractor Analysis app is the home view: a keep-watch-replace mix bar across your roster, then every sub as a ranked bar with its verdict, its costed reason, and its billed total. Clicking any bar opens that sub's detail card. The same drop-candidate logic also drives the Replace tab on a sub's card, covered in replacing a sub, and it feeds the decisions that surface in your Feed, covered in sub decisions in your Feed, so the verdict you see in the analysis and the card that reaches your Feed never disagree.

Data sources