Build your requirement template from your last five deals, not from a library

The usual way a firm gets a diligence checklist is that somebody copies the last deal, or downloads one. Both produce a list that is roughly right and impossible to argue with, because nobody remembers why any particular line is on it.

There is a better source sitting in your own closed deals, and getting it out takes an afternoon.

The extraction

Take your last five closed deals of one type. For each requirement on each deal, record three things: was it asked for, was it supplied, and was it waived.

Then look at four patterns.

Asked every time, supplied every time. This is the spine of the template. It is also usually shorter than people expect — often twelve to twenty items carry most of the actual work.

Waived every time. This is not five judgement calls. It is an item that does not belong on the template for this deal type, and it has been costing you a small amount of counterparty goodwill on every deal by being asked for and then not mattered.

Added by hand every time. The reverse, and the more valuable of the two. An item the deal team wrote in on five deals running belongs on the template. Somebody has been retyping it for a year.

Asked, and then never chased. The most interesting category. These are items nobody actually needs but nobody is willing to remove, and their real cost is not the asking — it is that they train the team to treat the whole list as advisory.

What a template should carry per item

Not just a label. Four properties, each of which prevents a specific problem later.

The kind of thing it is. A document, a written answer, a system connection, a signed attestation, a check you run yourself, or internal work owed by your own team. These behave differently enough that collapsing them into "things we need" is what makes lists unusable.

Who owes it. By role, not by name — the borrower, a guarantor, the applicant, or an internal team. On a multi-party deal this is what stops a guarantor being asked for the operating company's management accounts, which is how a counterparty decides the list is generic.

Whether it blocks. If everything is mandatory then nothing is. Mark the items that genuinely stop the file being IC-ready and let the rest be optional, honestly.

How long the evidence stays good. Management accounts have a shelf life. A registry check has a shelf life. Recording it means an item accepted in week two can be flagged as stale in week nine rather than discovered at committee.

Internal items belong on it

The single most common omission.

"Legal to review the facility agreement" is a real piece of work that must happen before the file is complete, and on most templates it does not appear, because templates are lists of things the counterparty owes.

The consequence is that internal work cannot be late — there is nothing for it to be late against — and the delay it causes is invisible. Putting it on the list costs one line and makes roughly half your cycle time measurable for the first time.

Keep it small enough to be believed

The temptation with any template is accretion. Every deal that goes slightly wrong adds a line, nothing is ever removed, and after two years the list is the union of every deal the firm has ever run.

A ninety-item list where each item has an owner and a reason is workable. A thirty-item list where twelve are aspirational is not, because the team learns that the list overstates, and once they have learned that they apply the discount to all of it — including the twelve items that genuinely block.

Review on a cadence, and treat removal as a normal act rather than an admission.

The loop

The whole mechanism is: waivers tell you what to remove, hand-added items tell you what to add, and both require a system that records what was decided rather than only what was finally true.

That is the entire method. It is not sophisticated, and almost nobody does it, because the raw material — what actually got waived and added, across deals — usually exists only in five separate email threads.