The Business Case, Honestly

Process mining business cases are built on published multipliers. Those are the weakest part of most of them.

Savings that are real

Waiting time removed. Usually the largest component of cycle time and the most reducible.

Rework avoided, once its origin is traced rather than its detection point.

Handover reduction, where each handover carries a queue.

Manual steps automated, where task mining identified the sequence.

Compliance findings — payments made twice, approvals skipped, segregation of duties breached — which are recoveries and risk reductions rather than efficiency claims.

Each is measurable in your own log before and after, which is what makes it defensible.

Figures to avoid

Percentage improvements from vendor case studies with no accessible method.

Savings assuming every variant is eliminated, which none are.

Extrapolation from a pilot on the cleanest process to the whole estate.

Any number you cannot trace to a method you could describe. One unsupportable figure checked by a finance director discounts every subsequent one.

Sizing from your own log

Decompose cycle time: working time, waiting time, rework time.

Take the largest non-value component and estimate a conservative reduction.

Multiply by case volume, which the log gives you exactly.

Convert to cost at your own rates, or to cycle time if that is the commercial driver.

Show the assumption, so the assumption gets challenged rather than the conclusion.

The compliance case separately

Duplicate payments, skipped approvals and breached segregation are findings with a direct value and they frequently pay for the project in the first analysis.

They are also the easiest to verify, since each is a specific case number someone can check.

Present them separately from the efficiency case, because they are certain and the efficiency case is an estimate, and mixing them makes both look like estimates.

What to commit to

Not a percentage.

A specific change in a specific measure, with a date and the same query afterwards.

"Median time between approval request and decision from six days to under two within two quarters, measured from the same event log."

Checkable, which is uncomfortable and is what makes the second phase fundable.

Costing it honestly

Licensing, by case volume or by user, and the model matters at scale.

Extraction effort, which is your own people and is the largest first-year cost.

Data quality remediation.

Ownership, permanently.

Consultation, where task mining is involved.

A programme that spends its budget before producing a finding has no evidence for year two, which is an argument for phasing rather than for a platform-first purchase.

The compliance findings that pay for it

Certain, specific and usually found in the first analysis.

Duplicate payments, identifiable as two payment events against one obligation.

Approvals skipped on cases above a threshold.

Segregation breaches, where one person performed two activities that must be separated.

Controls applied after the event they control.

Cases processed by an unauthorised role.

Each is a case number someone can verify, which is why these land where efficiency estimates argue. Present them first and separately.

Sizing the waiting reduction

The largest efficiency claim available and the one to build carefully.

Decompose cycle time into working, waiting and rework.

Take the waiting at the top three transitions.

Estimate a reduction from the reason category: availability problems reduce substantially and cheaply, capacity problems do not.

Apply to volume, which the log gives exactly.

State the assumption, so it is the assumption that gets challenged.

Claim conservatively. A delivered small number funds phase two; an undelivered large one does not.

Sizing it from your own log

A defensible number, built from data you already extracted.

Take total case duration and split it into working time and waiting time.

Take the rework loops and count the repeated activities.

Take the deviating variants and count the cases in them.

Cost the largest of the three at your own rates, conservatively.

State the assumption, so it is the assumption that gets challenged rather than the conclusion.

Do not quote vendor percentages. One unsupportable figure checked by a finance director discounts every subsequent number the programme produces.