The Excel Workbook: Every Sheet Explained

The Excel workbook is the primary deliverable of a benchmarking study: the complete, styled .xlsx record of the run — the parameters, the screened population, every disposition with its reason, the financials, the PLI distribution and the arm’s-length range — produced from the settled results on approval and stored against the study. Where the Word master report is the narrative document, the workbook is the data document: the artifact a reviewer opens to check a number, and the artifact the report’s appendices refer to.

How the workbook is produced

• On approval, not inside the run. The analysis pipeline ends at Final Analysis and hands the study to review; the workbook is built by a separate task on the report queue, dispatched when the study is approved out of In Review. Report generation is deliberately not a pipeline stage.
• The state follows the file. An approval records that a report was requested and leaves the study in In Review; the study moves to Report Generated only once the workbook has actually been built and stored. A failed build leaves the study where it was, so the badge never claims a deliverable that does not exist.
• Best-effort by design. The analysis is committed first; a workbook failure is a report problem to retry, not a study problem.
• Reviewer overrides are authoritative. The build applies the effective verdicts recorded in the comparables grid first — the accepted and rejected sets are rebuilt from them and the statistics recomputed — so the deliverable matches the approved record. Flagged and almost-accepted companies are flushed from the final deliverable; they are review work, not report content.
• Values, not formulas. Everything on every sheet is computed in Python and written as a literal value. The workbook carries styling, number formats and conditional formats (data bars, arrow icon sets, red highlight on a crossing ratio), but no live spreadsheet formulas: what you read is what the engine computed.
• Stored on the study. The workbook uploads to storage under the reports/{tenant}/{user}/{uuid}.xlsx key and is linked to the study as its report file id; the stored filename is TP-Benchmarking Analysis-{entity}.xlsx.
• Retention gate. The rebuild works from the study’s raw dump; if the dump has passed the tenant’s retention window, the build fails with a clear DUMP_EXPIRED result — the study must be re-run to capture a fresh dump before the workbook can be rebuilt. Once the workbook is built, the stored dump is cleaned up.

Downloading the workbook

The workbook downloads from the study: both the study detail view and the results analysis view show the action once a report file is linked to the study, and each press fetches a presigned download URL for the stored file on demand and opens it. While a report job is still running the action is hidden rather than shown dead. The file on disk is the one linked on the study — there is no separate “export” step, and no second copy.

Sheet by sheet

The workbook contains six styled sheets, in this order:

#	Sheet	Contents
1	Strategy	The study’s parameter record and outcome, as “Parameter / Value” rows: jurisdiction, function type, nature of business, margin type (PLI), averaging method, the tested party’s margin, the screens (max RPT %, min/max revenue in USD m, min employee cost %), the configured lower and upper percentiles — then the headline result: total companies, accepted, rejected, the arm’s-length conclusion, median PLI, mean PLI, standard deviation and CV. The parameters a reviewer holds the run to
2	Search Result	The dump record for the screened population, one row per company, with an autofilter: sixteen fixed columns (S. No, Company Name, NACE Rev. 2, Website address, Country, BvD ID number, Status, Date of incorporation, Primary code, Trade description in original language, BvD Independence Indicator, Trade description (English), Full overview, Main products and services) plus one ticked column per search criterion (the study’s function types and nature of business), then Result (ACCEPT / REJECT / FLAG) and Remarks (the recorded rationale or rejection reason). The input record, before dispositions
3	Accept Reject Matrix	Accepted versus rejected companies with the recorded reason per company — the disposition record, the sheet the review actually turns on
4	Financial Results	The accepted companies’ financials with ratio analysis, laid out to the reference report format: per-metric group headers (for example, “Intangible to Total Assets Ratio” and “Stock Turnover Ratio”), P&L percentage metrics displayed as scaled percentages, balance-sheet ratios as raw decimals, a gross margin computed as gross profit over revenue, and a red conditional highlight where a ratio crosses the 5% display threshold
5	Margin Analysis	The accepted set’s per-year base metrics, the study’s PLI as a ratio column, the weighted average, and the distribution block at the foot (below)
6	Web Analysis	Per company: BvD ID, country, External Sources — the URLs behind any web-research findings and their extraction confidence, empty when that step did not run — products-and-services text, corporate structure, and functional profile. When the optional Web Research step ran, its findings are appended to those three columns under a Web research: label, so the workbook’s own prose and website prose stay distinguishable; the Result/Remarks come from screening

Margin Analysis: the ratio follows the PLI

The Margin Analysis sheet displays the three base metric groups (operating income, operating profit, total cost) for each fiscal year shown — the three most recent detected years at most — plus one ratio column selected by the study’s margin type, and a single weighted-average column beside it:

Margin type	Ratio displayed	Formula
OP/Sales	Margin on total revenue	Operating profit / Revenue
OP/OC	Mark-up on total cost	Operating profit / Cost
TP/Sales	Margin on total revenue	Operating profit / Revenue
EBIT/Sales	Margin on total revenue	Operating profit / Revenue
operating_margin	Margin on total revenue	Operating profit / Revenue
EBITDA/Sales	EBITDA margin	EBITDA / Revenue
Berry Ratio	Berry ratio	Gross profit / Cost
EBIT/Total Assets	Return on assets	Operating profit / Total assets

The weighted average follows the Golden Rule: sums of numerators over sums of denominators across the years shown — never an average of per-year ratios. That is the same aggregation the analysis applies to the tested party and the pool, so the sheet and the statistics agree by construction.

Beneath the company rows, a five-line block summarises each ratio column, labelled exactly Lowest value, Lower quartile, Median, Upper quartile, Highest value. These are the quartiles of the ratio columns on the sheet; they are a display of the accepted-set distribution, not the study’s configured arm’s-length percentiles.

Using the workbook as a reviewer

• Start with Strategy. The parameters are the contract: PLI, percentiles, years, averaging, scope. Everything downstream is a consequence of this sheet.
• Turn on the Accept Reject Matrix. The dispositions are the record of why the pool is the pool; a rejection without a specific, factual reason is a workflow gap the matrix exposes (the reference is the Accept-Reject matrix).
• Check Margin Analysis. The range and the tested party’s position, on the same scale as the on-screen results — and the weighted average that the multi-year treatment produced.
• Keep Financial Results for the citations. When a number in the narrative or the review needs its source, it is here, per company, per year.
• Web Analysis is the qualitative trail. The dump’s own trade-description and products-and-services text, the corporate structure recorded on the comparable (the vendor independence indicator where your file carries one) and the functional profile, next to the screening Result and Remarks — the material the qualitative dispositions rest on. It is your reference data and your dump text; web pages enter the sheet only through the gated Web Research step, and only as lines carrying the URL they came from.

The workbook is the record the master report’s annexures are built from. They are schedules, not pointers: Annexure E prints the accept–reject matrix, Annexure F the tested party’s profit level indicator computation, Annexure G the same computation for each accepted comparable and Annexure H what each of those comparables actually does — the same rows these sheets hold, re-rendered into the document. Where a study has no accepted comparables or no screening record, the annexure says so in one line instead of printing an empty table.

FAQ

Does the workbook regenerate when I override a disposition in review? The next build of the report does: report generation applies the effective verdicts, rebuilds the accepted/rejected sets, recomputes the statistics and produces the workbook from the approved record — flagged and almost-accepted companies are not in it.

Where does the file live, and can I get it later? On object storage, linked to the study as its report file id. Each download mints a fresh short-lived presigned URL for that stored object — the link is fetched on demand, never embedded in the UI ahead of time. Each rebuild stores a new object and relinks the study to it, so the download always returns the latest generated workbook.

The workbook build failed with DUMP_EXPIRED — what now? The study’s raw dump has passed the tenant’s retention window, so the workbook cannot be rebuilt from it. Re-run the study (a fresh run captures a fresh dump) and the workbook generates from the new record.