Benchmarking board independence before the data arrives
Governance codes set the yardsticks for independence, tenure and committees. This draft shows how to build the comparison, and why it cannot yet carry any figures.
A board that wants to know whether its independence, tenure and committee make-up are sound needs a yardstick. The usual yardstick is a national corporate governance code, and most directors meet it through a summary written by someone else. That habit is the first thing to break. This piece sets out how to build a benchmark that holds up when a shareholder or regulator reads it. It also says plainly where the draft stops. The comparative facts this brief depends on were not in the material we were given, so this draft contains no figures, and no code provision is quoted from memory.
Before you start
Four things must exist before anyone writes a comparison. Without them, the benchmark will describe what the board believes about itself, not what the code asks of it.
- A dated register of every director, with appointment date, committee seats and any relationship with the company, its management or its largest shareholders.
- The governance code for each market in which the company is listed or regulated, in the version in force on the day of the review.
- The OECD Corporate Governance Factbook, which this brief was meant to rest on as its comparative source. It was not among the material supplied.
- A named owner. In most boards this is the company secretary, working with the chair of the nominations committee.
Write the register before you read the code. Otherwise the code sets the questions and the register merely answers them, which is the wrong way round. A register written first forces the board to say what it knows. A register written second tends to repeat what the board hopes is true.
Three things to benchmark
Independence is the dimension most boards think they have settled and many have not. Each code defines an independent director in its own words, and the definitions do not always count the same things. A useful test is to list every relationship the code names and check each director against that list, one line at a time, rather than relying on a single yes or no. A director who is independent in the board's own judgement may still fail a specific test in the code. Where that happens, the board should explain the gap in writing, not resolve it quietly.
Tenure is a question of both rule and judgement. A code may set a point after which a director's independence must be justified, and a board may set a limit of its own. Either way, the benchmark needs the actual date of each appointment, not an estimate. In our experience, boards count tenure from different starting points: the day a director joined this board, the day they joined a predecessor body, or the day of a first re-election. Each choice gives a different answer. Pick one convention, write it down and apply it to every director, including the chair.
Committee expectations are where codes tend to be most specific and where boards are least careful. The question is not only whether the audit, remuneration and nominations committees exist. It is who sits on them, and whether those members meet the independence standard the code applies to that particular committee. The answer can differ from the board-level test, and it can differ between codes. Read the committee provisions separately from the board provisions, and do not assume that one code's test carries over to another.
A sensible test for each dimension is the same. Could the company show, on one dated page, how each director meets each requirement? Could a new director check that page in an afternoon and reach the same conclusion? If not, the document is a description, not a benchmark, and it will not survive a question from an investor who has read the code.
Where benchmarks go wrong
Most weak benchmarks fail in the same few places. In our experience the problem is seldom the code itself. It is the paperwork built around it, and the habits that let the paperwork drift from the facts.
- Comparing a summary of a code with the code itself, so that a qualification in the text is lost at the first step.
- Mixing conventions for tenure, so that one director is counted from joining this board and another from joining a predecessor body.
- Recording committee membership as it stood at the last meeting rather than on the date of the review, which hides a recent change of member.
- Treating a director's own declaration of independence as the answer, when the real test is the code's list of relationships.
The last failure is the most common. A board that collects independence declarations once a year and never checks them against events in between will find that its benchmark describes the previous year. A new appointment, a new business relationship or a change in a director's other roles can alter the answer within months. The register should be updated when something changes, not when the annual review comes round.
What the sources do not yet show
The three sources checked for this piece cover other subjects. The European Securities and Markets Authority's current items cover post-trade settlement, risk monitoring and its work programme. The Financial Stability Board's items cover cross-border payments, stablecoins, resolution and private credit. The World Bank's financial sector pages cover small business finance, financial stability and its assessment programme. None of them addresses board independence, tenure or committee composition, so none can supply a figure for this brief.
The figures the brief needs sit in the OECD Corporate Governance Factbook and in the national codes themselves. Until a reviewer has read those documents, any percentage of independent directors, any average tenure or any committee norm would be an invention. This draft therefore states none, and it does not describe what any code requires beyond the general method set out above.
How to present it to the board
Put the register first, the definitions second and the comparison last. The chair should see the gaps before the directors are asked to vote on anything that depends on them. Show each gap with its cause: a missing date, a disputed relationship, or a committee seat that no longer matches the register. A gap with a named cause and a named owner is a manageable item. A vague gap that appears in a footnote is the kind that reaches an investor's letter.
What it means for a leader
Benchmarking is a question of evidence, not of ambition. A board that can show its register, its definitions and its committee memberships on one dated page is in a stronger position than a board that cites a score it cannot reproduce. The first task is to make the record reliable. Comparison with other companies comes after that, and it is only worth doing when the inputs are sound.
The most useful exercise is a stress test. Ask the nominations committee which directors would still count as independent under the strictest code the company faces, and which would not. If the answer is unclear, the register is not yet good enough. If the answer names a director, the chair should discuss it with that director before the question is raised by someone outside the boardroom.
Keep the benchmark separate from the board evaluation. One measures compliance with a set of rules. The other measures how the board works together and whether it asks the right questions. Blurring the two tends to make both less useful, and it invites a debate about definitions when the real question is about conduct. A good benchmark gives the evaluation a clean starting point, and that is its main value.
Limits of the data
This draft has three limits. It contains no figures, because the core comparative source was not supplied. It quotes no code provisions, because none were supplied and we have not written provision wording from memory. And it says nothing about any particular board, so it cannot place any company against the benchmark it describes.
Before publication, a reviewer should read the OECD Corporate Governance Factbook and the code for each market covered. Any figure added should carry its publisher and year, and every definition in this piece should be checked against the text of the code it concerns. Until that is done, treat this article as a method for building a benchmark, not as a finding about how boards compare.