1
What is operational due diligence?
The phrase is used in two settings. At fund level, operational due diligence examines a manager rather than its investments: valuation policy, custody, segregation of duties, service providers and business continuity. Institutional investors treat it as a separate discipline from investment diligence, and a failure there can end an allocation on its own.
At company level, operational due diligence examines the operating business itself. That is the sense used here. The question is how work is done, by whom, at what cost, and with what margin for error.
2
Which steps does the client actually pay for?
A process map is only useful if it distinguishes between steps that produce something a client has agreed to buy and steps that exist to support them. Most businesses have never made the distinction on paper.
The exercise is straightforward and rarely comfortable. Take a representative job from enquiry to payment. List every step. For each, record who performs it, how long it takes, and whether it appears on an invoice. The proportion of hours that never reach an invoice is one of the more informative numbers a buyer can hold, and it is usually higher than management expects. Rework, chasing missing information and reconciling records between two systems tend to account for most of it.
3
Where does quality depend on a small number of people?
In many operating businesses the accuracy of the work rests on a handful of individuals who hold undocumented knowledge: how to price an unusual job, which supplier can be trusted on a short lead time, which client will accept a variation. Their judgement is genuine and the business is right to rely on it. The risk is that no one has recorded what it consists of.
The examination identifies these positions by tracing errors backwards. Where a step goes wrong only when a particular person is absent, that person is a dependency. Where a step is reviewed by the same individual regardless of volume, that review is a constraint on growth. Both belong in the report, with roles stated and names withheld.
4
Which tools were bought and never used?
Almost every business of reasonable size is paying for software that nobody opens. The licence count exceeds the active user count, and the actual work is done in a spreadsheet that sits outside the system of record.
This matters beyond the wasted subscription. Where the real record of work lives in files held by individuals, the data a buyer expects to inherit does not exist in usable form, and any plan assuming reporting from the system of record rests on an empty table. The gap between the system bought and the place the work happens indicates how a company is really managed.
5
How should automation potential be assessed rather than assumed?
Automation potential is easy to assert and slow to prove. A step is a plausible candidate when it is repeated at volume, defined well enough that two competent people would produce the same result, and judged against a standard that already exists. A step fails the test when the specification lives in someone’s head, when volume is low, or when an error would reach a client without anyone seeing it first.
For candidates that pass, the assessment is arithmetic before it is technical: volume multiplied by handling time, valued at the relevant wage, less the cost of running and reviewing the replacement. Where that calculation is not worth performing, the step is not worth automating, whatever the tool can do.
6
What belongs in the plan, and what does not?
Only three kinds of item survive into a plan we would sign. Those with a named owner inside the business. Those with a baseline number the company already reports, so that the change can be seen without new instrumentation. And those with a date by which the change should be visible.
Everything else is a pilot and should be described as one. Pilots are legitimate. They become a problem when they are counted in the value bridge.