FCA authorised · FRN 724952 0117 325 0027Quote & buy →
Data and analytics
If you sell statistical advice or analysis as a consultant, freelancer or firm, yes: you need professional indemnity insurance. Clients make high-stakes decisions on your numbers, such as whether a trial worked, how to price a product or whether a dataset is safe to release. When a model is mis-specified or an analysis is wrong, the loss is financial and can dwarf your fee. Public liability does not cover it. If you work only as an employee, your employer usually carries the risk instead.
Part of: Professional indemnity for consultants
In short
Consultant statisticians are judged against what a competent statistician would have done, with reference points such as the Royal Statistical Society’s Code of Conduct, the UK Statistics Authority’s Code of Practice for Statistics (edition 3.0) and, for clinical work, the ICH E9 guideline on statistical principles for clinical trials. The main PI risks are flawed analyses behind a failed trial or regulatory submission, mis-specified models that lead a client to misprice, and supposedly anonymous outputs that let people be re-identified. Expert witnesses owe an overriding duty to the court under CPR Part 35 and, since Jones v Kaney, can be sued by their own clients for negligence.
Last reviewed 5 October 2026 by the Apex professional indemnity team.
Statistical work turns data into decisions. A sample size calculation decides how many patients a trial recruits. A pricing model decides what thousands of customers pay. A weighting scheme decides what a published survey says. A disclosure control decides whether a table can go online. If the method is wrong, the client acts on a false picture, and the cost of putting it right lands on whoever produced the numbers.
Public liability (PL) covers accidental injury and damage to property, such as a laptop bag that trips a visitor at a client’s office. It has nothing to say about a model that underprices risk or an analysis that sinks a regulatory submission. Those are financial losses caused by professional work, which is what professional indemnity (PI) insurance is for.
The trade is broad: trial statisticians working for sponsors and contract research organisations, modellers and data scientists in financial services, survey and market research statisticians, consultants to public bodies, and expert witnesses. Your policy should describe every strand you take on, because a wording written for “data analysis” may not obviously include expert evidence or trial work.
These scenarios are illustrative. They show the kind of allegation statisticians face; they are not real claims.
None of these turn on bad luck. In each case the client says a competent statistician would have chosen, checked or explained the method differently.
No regulator licenses statisticians, but several codes describe what competent practice looks like, and a claim will usually be argued against them.
| Reference point | What it says | Why it matters to you |
|---|---|---|
| Royal Statistical Society Code of Conduct (revised November 2025) | Applies to all RSS Fellows. Work with due care and diligence; present methods and results accurately and without misrepresentation; offer work only within your competence; disclose conflicts; set out the risks if your judgement is overruled. When using AI, check models are trained on representative data, assess bias and validate results. Public interest duties, including privacy, take precedence over obligations to clients. | The clearest statement of what the profession expects, and a benchmark an expert on the other side will use. |
| Chartered Statistician (CStat) | The RSS’s highest professional award. Applicants must first be Fellows, normally have at least five years working in statistics, pass a peer-reviewed assessment and commit to continuing professional development and revalidation. | Clients and insurers treat it as evidence of competence. |
| Code of Practice for Statistics, edition 3.0 (UK Statistics Authority) | Sets the standards for producers of official statistics, built on Trustworthiness, Quality and Value. Its principles include being rigorous, being open about quality, including uncertainty and bias, and managing data responsibly. Organisations outside government can apply it voluntarily. | Public sector clients will expect your outputs to meet it. |
| ICH E9: Statistical principles for clinical trials | Assumes that responsibility for all statistical work in a clinical trial lies with an appropriately qualified and experienced statistician. For trials supporting a marketing application, the design and the main features of the planned analysis should be set out in a protocol before the trial begins. | Trial sponsors will measure your protocol and analysis plan against it. |
| Civil Procedure Rules, rule 35.3 | An expert’s duty to help the court on matters within their expertise overrides any obligation to whoever instructed or paid them. | Your report must be independent, even though your client may later sue you over it. |
| Jones v Kaney [2011] UKSC 13 | The Supreme Court abolished expert witnesses’ immunity from suit for breach of duty in relation to their participation in legal proceedings. Absolute privilege against defamation claims remains. | An expert can be sued by their own client for a negligent report. |
Re-identification is the claim that sets statistical work apart from most consultancy, because the harm reaches people who never hired you. If outputs you prepared allow someone to be identified, the data was personal data all along, and your client faces complaints, a regulator’s questions and possibly compensation claims from the people affected.
PI can respond to a client’s claim that your anonymisation was negligent, if the policy includes data protection liability. Breach response costs on your own systems belong to a cyber policy, and PI wordings commonly exclude fines and penalties.
| Usually covered by PI | Often excluded or limited | Needs a different policy |
|---|---|---|
| Negligent study design, sample size calculations and analysis plans | Promises that a trial, model or forecast will achieve a result | Accidental injury or damage at a client’s premises (public liability) |
| Errors in models, code and calculations that cause a client financial loss | Fines and penalties | A cyber attack on your own systems, and the breach response (cyber) |
| Disclosure of personal data through inadequate anonymisation, where data protection liability is included | Bodily injury arising from clinical or medical work | Harm to trial participants (the sponsor’s clinical trials insurance) |
| Expert witness work, including claims by your own client | Uncapped indemnities and other liability you accept only by contract | People you employ (employers’ liability) |
| Defence costs, including the cost of an independent statistical review | Problems you knew about before the policy began | Computers and software (office or equipment cover) |
Cover is always subject to the insurer’s acceptance and the policy terms. If AI or machine learning forms part of your work, describe it plainly when you buy or renew, so there is no argument later about whether it fell within your declared activities.
Your limit is usually set by your clients. Pharmaceutical sponsors, contract research organisations, banks, insurers and public bodies routinely put a PI requirement in their terms, often with a minimum figure and sometimes on an each and every claim basis. Size it by the decision your work supports, not the fee: the cost of repeating a trial, a year’s premium written on a flawed pricing model, or the sum at stake in a dispute where you are the expert.
Statistical claims have a long tail. A trial can take years to report, a pricing model can run for several renewal cycles before the losses show, and an expert witness claim usually follows the end of the case. PI is claims-made, so the policy that pays is the one in force when the claim arrives, not the one you held when you did the work. Keep cover continuous, check the retroactive date when you move insurer, and arrange run-off if you retire, close your consultancy or take an employed role. See run-off cover explained.
A complete proposal gets better terms than a bare one, and a broker can only present what you tell us. Have these ready:
Speak to a broker
PI for statisticians, placed by a named broker
Start the online proposal and save it as you go, or leave your number and a named broker will call you back, usually the same working day.
Apex Insurance Brokers is an independent insurance broker based in Bristol, established in 2009 and authorised and regulated by the Financial Conduct Authority. We are not tied to one insurer: we work with over 30 markets, including Lloyd’s syndicates through wholesale brokers, and every client has a named broker who handles the placement, mid-term changes, certificates for clients and the renewal.
Yes, if you provide statistical advice or analysis to clients. A flawed trial design, a mis-specified model or a weak anonymisation can cause losses far larger than your fee, and public liability does not cover financial loss. PI covers your legal liability for negligent work, subject to the policy terms.
No law requires statisticians to hold PI, and the Royal Statistical Society’s Code of Conduct does not mention insurance. In practice the requirement comes from clients: sponsors, contract research organisations, financial firms and public bodies commonly make PI a condition of the contract, often with a minimum limit.
Yes, if the client can show the model fell below the standard of a competent statistician, for example through untested assumptions or poor validation. Record your assumptions, validation results and limitations, get sign-off on key choices, and set out the risks in writing if the client overrules you.
Yes. In Jones v Kaney the Supreme Court abolished expert witnesses’ immunity from claims by their own clients for negligence connected with legal proceedings. Your overriding duty under CPR rule 35.3 is still to the court. Make sure your PI expressly includes expert witness work.
It can, if your policy includes data protection liability, subject to the terms. The client will argue that inadequate anonymisation made the data personal data. The ICO recommends testing identifiability with a motivated intruder test, and PI wordings commonly exclude fines and penalties.
It can, provided the work falls within the business description on your policy. Tell your insurer how you use these methods. The RSS Code of Conduct expects Fellows using AI to check training data is representative, assess bias, validate results and communicate findings transparently, and a claim will test whether you did.
Apex arranges professional indemnity insurance for statisticians across the UK. Tell us about your work and we’ll find cover that fits. Or call 0117 325 0027.
Start your PI proposal Call 0117 325 0027Apex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information, not advice on your individual circumstances. Cover is always subject to the insurer’s acceptance and the policy terms, and this page does not guarantee that cover will be available or on what terms.
