TP Benchmark: Transfer Pricing Comparables screening done by AI

Hey Product Hunt 👋 Every transfer pricing benchmarking study has the same unglamorous middle step. You run a quantitative screen in Orbis, it hands back a few hundred candidate companies, and then somebody opens each one's website in a browser tab and decides, by eye, whether it is actually comparable to the tested party. That step is where the study is won or lost, and it is the step that almost never gets written down properly. The reasoning lives in the analyst's head. When a tax authority asks why company 147 was rejected two years later, the honest answer is often "we don't remember." TP Benchmark does that step, and writes down its reasoning. Upload your Orbis export. Optionally, drop in your existing TP documentation and let it extract the tested party profile for you. Then: 🔍 It finds each company's real website. Orbis domain first, then scored web search, with job boards, company directories and registries filtered out. You approve the discovered URLs before a single analysis runs, so you never pay for a screen of the wrong companies. 📑 It reads the site and applies the OECD 2022 framework. Full FAR analysis, the five comparability factors, and the independence check. Every conclusion has to quote the text it came from. If a finding cannot be supported by something on the page, it gets recorded as an information gap instead of a confident guess. ⚖️ You get ACCEPT, REJECT or REVIEW. REVIEW is a real outcome, not a failure. A borderline case, a thin site, an ambiguous ownership phrase: those go to a human rather than getting a decisive-sounding verdict they don't deserve. 🗄️ It captures proof. Archive snapshots of what each site said on the day it was analyzed, so the evidence still exists when the site is redesigned. 📊 Then it exports. Interquartile range with your tested party's position in it, and an Excel workbook you can hand to a reviewer or attach to the file. What it deliberately does not do: it does not replace your judgment. It is the qualitative layer between the Orbis screen and the professional signing the study. Financial screening stays in Orbis, sample construction and the final call stay with you. The output is a recommendation with its evidence attached, which is precisely what makes it reviewable. Pricing is per company analyzed, €0.50, minimum 150. No seats, no subscription. I would especially love feedback from TP practitioners: where is the analysis too permissive, and where is it too strict? The comparability calibration is the part I most want to get right.