Industry Diagnostics: Margins, Trends and Risk Profiles

Industry Diagnostics is the cross-study view: margin data aggregated from every stored analysis the firm has run, grouped by industry. It is a read-only reference layer — nothing in it touches a study, a comparable or a range — and it is only as good as the practice’s own finished work: the more stored analyses a firm has in a sector, the more the diagnostics say about that sector. It sits in the Results area of the sidebar, next to the per-study results, and it answers a different question from them: not “where does this study’s tested party stand” but “what does this sector look like across everything we have benchmarked”. The view is open to every signed-in user of the workspace; it is not behind a plan feature.

What feeds it, stated exactly:

• Studies — every study in the tenant carrying a stored analysis whose persisted result committed successfully. Workflow state is irrelevant: a study sitting In Review contributes, and a failed or never-run analysis does not. A superadmin on a tenant-less platform account sees only their own studies.
• Companies — the accepted companies of those analyses, and only those carrying both a PLI value and an industry label. A blank industry keeps a company out of every tab.
• Year and jurisdiction — read from the study, not from the company: the financial year is parsed out of the study’s own year value, and the jurisdiction from its parameters.
• Counts are rows, not entities — a company benchmarked in two studies contributes twice. The totals are study-company rows.

Margin Aggregator

The first tab aggregates the accepted-set margin statistics per industry:

• Summary cards — total company rows, the number of contributing studies, and the number of industry sectors covered.
• The sector table — one row per industry: company count, study count, mean PLI, median PLI, standard deviation, CV, a percentile band showing the sector’s quartiles, and the full min–max range. The response also carries p10, p50 and p90 per sector.
• Sorting — the Companies, Mean PLI and CV headers toggle; the table defaults to company count, highest first, so the densest sectors surface first and the most dispersed ones follow on CV.

This is the sector’s margin profile: the level (mean/median), the spread (std dev, CV, quartile band) and the depth (how many company rows and studies back the row).

Trend Analysis

The second tab takes the same data through time: the overall picture across all industries, or one industry selected from the dropdown.

• The line chart — mean PLI by fiscal year for the selected scope, with a dashed median line beneath it.
• The year-by-year table — for each fiscal year: mean PLI, median PLI, standard deviation, CV, company count and study count.

The years are the studies’ own financial years, parsed to a calendar year; a study whose year does not parse is grouped as Unknown and drops out of the overall year series, so a gap in the chart can be a data-entry gap rather than a quiet sector.

What to read off it: a drifting median (sector margins moving level over the years), a widening CV (dispersion growing — the sector is less homogenous as it gets newer), and a thinning company count (data availability in the sector is drying up, which will bite future benchmarks there).

Risk Profiles

The third tab ranks the sectors by volatility — the profile of how hard a benchmark in each sector is to run cleanly:

• Summary cards — the counts of high-volatility, medium-volatility and low-volatility sectors.
• The sector table — one row per industry: volatility rank, company count, mean PLI, standard deviation, CV, a risk level badge (High / Medium / Low), a risk score bar (0–100) and the dominant jurisdictions in the sector’s data.
• The bands. Risk level is the sector CV: High at 50 and above, Medium from 25 to under 50, Low below 25.
• The score. The 0–100 bar is an additive index of the same CV band (40 / 25 / 15 / 5 points at CV ≥ 60 / ≥ 40 / ≥ 25 / ≥ 15) plus a depth penalty for thin sectors (30 points under five company rows, 20 under ten, 10 under twenty), capped at 100.
• The ranking is by CV, highest first; jurisdictions list up to five distinct values found in the sector, alphabetically, and a study with no jurisdiction recorded contributes Unknown.

A thin, scattered sector therefore scores high on this tab for two separate reasons — the margins disagree with each other, and there are few rows behind the row. That is the profile of a sector where re-benchmarking is normal maintenance, not a sign something went wrong. Note that this is the sector view’s own arithmetic: it is not the per-study risk and reliability scores, and a sector can read High while the individual studies in it score comfortably.

How a firm uses it

• Before a run (sanity check). The tested party’s file margin against the sector’s typical band: a file margin that sits far from the sector median before any benchmarking is a scope or profile question worth settling while it is still cheap — the PLI, the function, or the entity’s own economics.
• After a run (check the benchmark). The study’s accepted-set median against the sector band: inside, and the screen selected a representative pool; outside, and the screen over- or under-selected in a way the diagnostics make visible. The per-study reading of the same numbers is in Reading the Statistical Results.
• Across the portfolio. Which of the firm’s sectors are structurally high-volatility, and in which jurisdictions the sector’s data lives — the input to search design (geography, years, size filters) before the next study in the sector starts.

The limits, stated plainly

• The diagnostics aggregate committed runs: a sector with one stored analysis has a row, and the row is that analysis.
• It is a reference, not an input: no study’s range, score or disposition is computed from the diagnostics.
• Margins are shown as PLI statistics per the studies that produced them. The aggregation does not normalise across PLIs, so comparing sectors whose studies used different indicators is the reviewer’s interpretation, not the view’s.
• Only accepted companies carrying both a PLI and an industry label are counted, so the sector rows describe the surviving pool, not the screened population.
• The three aggregations are cached briefly. A study that just finished appears on the next cache refresh, not the instant it commits.

FAQ

Whose studies feed the view? The tenant’s own studies, wherever they sit in the workflow, as long as their stored analysis committed successfully — so an In Review study contributes and a failed or never-run one does not. A re-run replaces the stored result, and the view follows the latest run.

Does the view change a study’s results? No. It is a cross-study reference layer; each study’s range and scores come only from its own accepted set.

The sector row looks thin (one or two companies). What does that mean? Depth, not error: few accepted companies in that sector across the firm’s completed studies. Treat the row as indicative only, and weight the per-study diagnostics for any study in the sector.