Portfolio Risk: Aggregating Risk Across Studies

The portfolio view aggregates the three risk scores across every study in the firm with a finished analysis: the counts, the averages, and the per-study detail that tells the partner which file to open first.

What these numbers are: averages and counts over rule-based heuristic indices, not over measured audit rates. They rank your own files against each other; they do not forecast what an authority will do.

What the summary shows

Plan feature: This capability requires the predictive_risk plan feature (see Plan Features).

Every study whose stored analysis result is marked successful is assessed; studies without one (draft studies, failed runs, runs whose result never landed) are skipped, not zeroed.

Measure	Meaning
Total Studies	The studies with a stored successful analysis in scope
High Risk	Studies whose audit index is 60 or above
Moderate Risk	Studies whose audit index is 35 or above, below 60
Low Risk	Studies below 35
Avg Audit Probability	The mean audit index across the assessed set
Avg Defensibility Score	The mean defensibility score (0-100) across the set
The study buttons	One per assessed study, badged with its audit band — selecting one opens its three scores and their breakdown

The bands count the audit index only. A study’s litigation level and defensibility grade do not move it between the high / moderate / low cards, and there is no portfolio average for litigation — the two averages returned are the audit index and the defensibility score.

The response does carry each assessed study’s complete breakdown (audit factors, litigation detail, defensibility criteria) — the page renders the selector and the gauges from it, which is why nothing else loads when you switch studies.

Reading it as a fix-first order

The page gives you a distribution and a selector, not a sorted table: there is no ranking control, the badges show the audit band only, and a study’s grade appears once you select it. So the fix-first list is something you assemble:

• High audit plus a weak grade (D/F) — the front of the list. The heuristics read the file as exposed, and the criteria say it would not withstand scrutiny.
• High audit plus an adequate grade — document and keep current; the issue is the exposure, not the file.
• Low audit plus a weak grade — the latent debt: a low index today, and a file that would not hold up if it were read.

With more than a handful of studies, read the payload rather than clicking through: every study’s audit_probability.score and defensibility.grade.grade come back in the single portfolio response, and sorting those two columns is the queue.

The re-benchmark queue

The D and F studies are your re-benchmark queue — a working practice, not a feature of the screen: nothing on the page flags, sorts or filters them, so the queue is the assessed studies whose grade you have read, by selecting them or from the payload. A D or F is a documentation package the criteria score as not holding — typically a small or variable pool, a wide range, or a margin outside it. The path out is the same for each: revisit the screening pipeline (the pool, the filters, the dispositions), re-run, then re-read the grade in the study drill-down.

Scope and freshness

• Tenant-scoped. The rollup queries the studies in the caller’s tenant. A Superadmin’s portfolio is scoped to the studies they created; a tenant-less platform account sees the tenant-less studies, since there is no tenant to scope to. Nothing crosses firms.
• Plan-gated. The whole surface sits behind the predictive_risk plan feature. Without it the sidebar entry is not shown and direct navigation is redirected, so the page never mounts and fires no API call; the backend feature check on the endpoints is the actual security boundary.
• Computed on read, cached per scope. The result depends only on the studies’ stored analysis, so it is cached and cleared on study transitions rather than kept in a table — no background job, no stored score to drift.

Reading the averages

Both averages are plain arithmetic means over the assessed set: every study counts once, whatever its revenue or materiality, and one unusual study moves them. They average the heuristic indices, so neither is an exposure figure in any monetary or probability sense — read them against the counts.

• Avg Audit Probability — where the firm’s files sit on the exposure heuristics. It rises when the mix rises: a jurisdiction with a higher table value, a more variable pool, more loss-making comparables, a study priced further from the median.
• Avg Defensibility Score — the more actionable of the two, because its criteria are checkable from the file itself (sample, dispersion, range, placement, acceptance rate). A falling average, even with stable risk counts, says the pools are getting thinner or more variable.

Read together: a high audit average with a high defensibility average is a portfolio the rules read as exposed but well-built; a high audit average with a falling defensibility average is the one to work on across the board, not study by study.

FAQ

When does a study enter the portfolio? The moment its stored analysis result is marked successful. A draft or a failed run is not in it until it completes.

Can I exclude or filter a study? Not through the view. The endpoint takes no date, client or jurisdiction parameter — it returns your whole assessed set. To change a study’s position, change its analysis (re-run, re-benchmark); to work a slice, filter the studies in the response yourself.

Why did the average move when no study changed? A study entered or left the assessed set, or a run rewrote a stored result — either changes the denominator or the inputs of both means.

How is this different from the dashboard? The dashboard shows workflow health (states, queue, throughput); the portfolio shows how the settled studies score against one fixed rule set. One is motion, the other is where your files stand.