Judicial approach
How this tribunal has reasoned on the point in its published judgments: the lines of authority relied on, and how procedure has been handled. Always with the count behind it.
Outcome odds · Precedent · Judicial approach · For litigators, counsel & funders
The odds in cases like yours, the authorities this tribunal reaches for, and how it has handled the point before. From the published record, in sixty seconds, on a method built and tested by a dual-qualified solicitor.
I · The problem
The night before a hearing, the one variable you can’t look up is the person deciding it. So you ask around chambers and hope someone has appeared before them.
The obvious preparation, and on a listing that landed last week, all you have time for. Then you walk in.
Hundreds of published judgments. Which authorities get cited unprompted. How late evidence has been treated. How often relief from sanctions is granted at this stage, and on what reasoning. None of it is in the bundle.
Priced for a £10m trial, not a two-day application. So nobody does it, and you argue on instinct. So does the funder underwriting your case.
So we read the published record once, over 500,000 English judgments so far, and keep it growing.
II · The briefing
Eight sections, each on its own terms, then weighed against the rest. The useful signal hides where the judge’s record and the textbook disagree.
How this tribunal has reasoned on the point in its published judgments: the lines of authority relied on, and how procedure has been handled. Always with the count behind it.
Success rates for cases like yours, with the cohort definition, the confidence interval and the count. Every case behind the number, listed.
How often decisions of this kind are taken up, how often reversed, and on which grounds. Useful before you decide whether to fight at first instance.
The authorities actually cited in comparable published cases, ranked by how often they are reached for unprompted, not generic keyword matches.
Which lines of argument succeeded or failed on this point in the published record, by outcome, never by who ran them.
How late evidence, adjournments and relief applications have been dealt with at this stage, as reported in the judgments themselves.
Recent appellate movement, Law Commission work and sector signals shaping the issue at the hearing date.
Where the record suggests you press, where you concede early, and what to have ready for the points this court has pressed before.
III · Jurisdictions
Court analytics only work if the method is built by someone who has practised before those courts. We start where we are qualified, and add jurisdictions the same way.
Commercial Court, Chancery Division, King’s Bench and the Court of Appeal: over 500,000 judgments published by The National Archives’ Find Case Law service, growing every week. That record is not complete, and every figure we show says how many decisions sit behind it. Live now for UK firms, chambers and funders.
Live and in daily use, including at the largest law firms in the country. Over 700,000 decisions of the courts of first and second instance, every claim traced to its paragraph, plus proprietary datasets of our own that no public database holds. Supreme Court, Supreme Administrative Court and Constitutional Court coverage on request.
Slovak courts must publish their final decisions by law, and the Ministry of Justice releases them as open data through a public API. The legal system shares its roots with the Czech one, so the Czech method carries over closely.
Joint waiting listPoland, Germany and Austria are next, on one shared waiting list. Join once, tick the countries you need, and you hear first when each one goes live.
Join the waiting listIV · The team
You could feed a judge’s name into a chatbot yourself. You would get a fluent answer, no source, and no way to check how it got there. That is a guess, not analytics.
So we split the work the way it should be split. The solicitor builds the method: the ontology that decides what counts as a claim and an outcome, what a comparable case is for each claim type, what a “win” means at each procedural stage, and how a quotation must match the original text before it is allowed in. He tests that method against decisions he has read himself, jurisdiction by jurisdiction, and it does not ship until it holds.
The software then runs that method across the published record, more than 1.2 million decisions across our live jurisdictions and counting, at scale and without tiring. That software is built by a senior data engineer who has spent a career on graph databases and large-scale data pipelines, which is what turns a million judgments into one connected, queryable record rather than a pile of PDFs. Anything it can’t verify against the original text is thrown out. When you ask a question through Claude or ChatGPT, the answer comes from that verified record in seconds, not from a solicitor reading your matter. That is what keeps a briefing cheaper than an hour of a trainee.
It is also why every number carries a confidence interval and a count, and every line links to the paragraph it came from. The method has a name behind it. The evidence is one click away. Your own judgment is the part that stays yours.
When the hearing is real, you want to know who built this. Here’s who.
V · Claude & ChatGPT
The platform is ready to connect to your Claude or ChatGPT through MCP, the open Model Context Protocol. No new app to learn. Ask a question in plain English and get a cited answer back.
One connector, one key per client. Your assistant gains a set of tools it can call: judicial approach, comparable-case odds, precedent scan, decision lookup. The data stays on our servers in the EU. Your assistant only sees what it asks for.
// you Odds on relief from sanctions in the Commercial Court, late witness statement, this stage? // freedia → get_odds(court, claim_type, stage) Comparable cohort: 61 applications, 2019–2026. Relief granted: 44% (CI 32–57, n=61). Division average for this stage: 57%. Published record only; counts are not the full caseload. Relief was granted in 19 of 27 cases where the delay was under 14 days, and in 8 of 34 where it exceeded that. Cases behind the number (61) attached, each linking to the judgment on Find Case Law.
VI · Pricing
A briefing costs less than an hour of a trainee. Every plan includes all live jurisdictions and the MCP connector for Claude and ChatGPT.
For a barrister or sole practitioner who wants the court read before every hearing.
For litigation teams and chambers. Shared matters, shared briefings, one key per person.
For national firms and litigation funders underwriting a book of cases.
Prices exclude VAT. Monthly plans cancel any time. Founding-year rates: locked for twelve months for anyone who signs up in 2026.
VII · Method & security
English judgments come from The National Archives’ Find Case Law service and are re-used in line with its nine principles for computational analysis. Czech and Slovak decisions come from the courts’ own open databases. Here is what that means in practice.
We read judgments the courts have chosen to publish. No scraped filings, no leaked documents, and nothing about any person beyond what a court has already put on the record. Our proprietary Czech datasets are lawfully obtained and held to the same rule.
Find Case Law does not hold every judgment. Every figure we show carries the number of decisions behind it and a confidence interval, so a pattern in 12 cases never reads like a rule.
Courts revise and withdraw judgments. We re-check the published record on a schedule, drop withdrawn decisions and re-run any figure they touched.
A claim, an outcome or a quotation enters the database only if it matches the original text character for character. Fuzzy matches are rejected. Every line links to the judgment on the court’s own page.
Where a court anonymised a party, so do we. We do not profile parties, witnesses or counsel, and we do not combine records in ways that could identify someone a court chose to protect.
Judgment text and the data drawn from it sit behind authentication. Nothing is exposed to search engines, and automated crawling of the service is blocked and monitored.
How a cohort is defined, what counts as a win, and how a quote is verified is written down and available to every subscriber. Ask the connector how a number was built and it tells you.
A solicitor reviews cohort definitions and a sample of outputs each quarter, and a subscriber can challenge any figure. Changes to the method are versioned and dated.
We report how the published record reads. We do not rank judges, label their impartiality, or predict what any individual will do. The independence of the judiciary is the premise of the product, not a variable in it.
Freedia is a preparation accelerator, not a decision-maker. Outputs are descriptive statistics about an incomplete published record and are not legal advice or a prediction of any outcome. Nothing on this service is a comment on the impartiality of any judge or tribunal. Personal data in judgments is processed in accordance with UK GDPR and the Data Protection Act 2018.
Live now · UK firms, chambers & funders
Tell us which plan and which assistant. You get a private connector address and key for Claude or ChatGPT, or your first briefing, within one working day.
Requests go straight to David. Expect a personal reply within one working day. Prefer email? Write to svoboda@advokaciesvoboda.eu.