Most of the trouble in an expert instruction is already there on the day it arrives. A question that is really a question of law. A bundle with the contract missing. A deadline set by the directions rather than by the work. A party you advised four years ago and had stopped thinking about. None of it improves for being left, and all of it is cheap to raise in the first week and expensive to raise in the sixth.

The review takes an instruction apart before you commit to it. You write out each question you are being asked to answer and say what kind of question it is — within your field, partly within it, a matter of law, a matter of primary fact for the witnesses, or too loosely framed to answer at all. You mark off what you have been sent against what you have not. You answer eight questions about prior connections. Then you set the volume of documents against the time you actually have between now and the deadline.

What comes out is a capacity assessment with the arithmetic shown, a dated record of the checks you made, and a numbered letter to those instructing you setting out exactly what you need. The hours you have planned go straight to the Fee Estimation Tool, so you price the work you have just scoped instead of starting again.

• Every question classified, with paragraph 2.3 of PD 35 in mind where one falls outside your field
• Sixteen categories of document marked held, outstanding or not needed
• An independence screen covering prior connections at the point of acceptance
• A capacity assessment against the deadline, with every allowance shown so you can argue with it
• A numbered list of clarifications, in a form you can send

It does not decide whether you can accept. That is your judgement, and sometimes your professional body’s. What it does is make sure the questions get asked while the answers can still change the instruction.

In 1993 Cresswell J set out what a court expects of an expert witness, in seven short propositions. They have been cited ever since, and CPR Part 35 codified the heart of them: your duty is to help the court, and it overrides your duty to whoever instructed you and pays you.

Almost nobody fails those duties deliberately. They fail them gradually. The assumption counsel supplied and you never labelled as an assumption. The awkward fact that did not make it into the report. The opinion you formed on thin data and never marked as provisional. The view that shifted after you read the other expert’s report, and that nobody was told had shifted.

The audit walks the duties across the stages of an instruction rather than treating them as a single box to tick at the end. Independence is tested when you accept. Disclosure of facts and assumptions is tested when you write. Whether your opinion has changed is tested after reports are exchanged, and again after the joint discussion — which is precisely where duty six is most often breached, and where it does the most damage under cross-examination.

• All seven duties, each in plain language with the point it is getting at
• Worked through by stage: acceptance, report, Part 35 questions, joint discussion, trial
• Flags anything you could not answer cleanly, so nothing is left half-considered
• A dated record that you ran the audit, to keep on the file

It does not certify that you are independent. No tool could. What it produces is a structured record of the questions you asked yourself and when — which is a far better answer under cross-examination than trying to remember.

More on the seven duties and where experts trip up →

The structure of a compliant expert report does not change much from case to case, but it does change between types of proceedings — and that is where reports get into difficulty. A criminal report under Part 19 of the Criminal Procedure Rules needs sections a civil report does not. Children proceedings need a confirmation civil proceedings never ask for. The statement of truth differs in each.

Pick your field and the type of proceedings, and the builder assembles the headings, the declaration and the correct statement of truth, with short guidance notes under each heading that you delete as you write.

The technical sections change with your field. Engineering gives site inspection, applicable standards, testing and cause of failure. Digital forensics gives continuity of evidence, tools and hash verification, and the limits of attributing activity to a person rather than a device.

• Civil, single joint expert, criminal, family, children proceedings and tribunal
• Field-specific sections for ten disciplines
• Optional sections for chronology, glossary, joint discussions and appendices
• Downloads as a Word document with A4 margins, ready to write into

Reports get into difficulty in predictable ways. The opinion that appears in the conclusions but never in the body. The range of opinion in the field that is never acknowledged, which is one of the most common criticisms made of expert reports. The instruction that shaped the whole analysis but is never set out. The assumption supplied by the solicitor that is never labelled as an assumption.

The checker is a last pass over the finished draft. Work through it before the report is served and you catch the gaps while they are still cheap to fix — rather than in a Part 35 question, or in cross-examination.

It complements the Structure Builder rather than repeating it. That one gives you the skeleton before you write. This one tests the finished document.

• Completeness and internal consistency of the report
• Whether the substance of instructions and all material facts are set out
• Whether the range of opinion is addressed and reasons given
• Whether the declaration and statement of truth are present and correct

Instructing solicitors ask for your publications when they are deciding whether to instruct you at all, and again when your CV goes into the bundle. Most experts rebuild that list from memory and a folder of PDFs, under time pressure, every time.

The tracker replaces that with a single record you add to as you publish. The form adapts to what you are recording, so a book does not ask for a volume number and a chapter asks for the book title and publisher.

Three controls then shape the output: chronological order, citation style, and whether to include everything or peer-reviewed work only. The bibliography rebuilds itself every time you change anything — for your CV, your report appendix, or an instructing solicitor who asked this morning.

• Journal articles, books, chapters, conference papers, reports and more
• Author–date or numbered style; newest first, oldest first, or grouped by type
• A peer-reviewed-only filter, which is often all that is wanted
• Copy, print, or export a backup file to carry between computers