Measure time saved without ignoring failed cases

Where the extra work starts

Include every attempted case and its correction time so the claimed saving reflects the batch your team actually receives.

Invoice rekeying is a bounded task you can test today. Use the same invoices to compare your manual method with ParseMill, including source checks, corrections and export. The monthly plan costs $49 USD for 500 PDF processing pages and one user; judge it against the work it actually removes for you.

Define eligibility before running the pilot. Cases excluded by an agreed input boundary can be listed separately, but cases should not become “out of scope” merely because the result was inconvenient.

A practical way to get it right

Create one case record with linked attempt rows. Record active effort, result and any manual recovery. Distinguish a successful first attempt, a successful retry, an abandoned automated attempt with manual completion and an unresolved case. Preserve the reason for each failure.

Calculate total effort across all eligible cases, including failed attempts. Divide by completed cases only with an explicit label and disclosure of unresolved work; otherwise the figure can hide effort whose outcome has not yet arrived. Report first-pass success separately from final completion.

Compare with the current process under the same completion standard and case mix. Include review and correction in both. A tool may be useful on a narrower input class, but that narrower scope needs a new clearly stated trial rather than retrospective removal of failures.

Microsoft: Excel functions by category supports counts by multiple conditions. Count case outcomes separately from attempt rows so retries do not inflate the success denominator.

What this looks like on a real task

Synthetic example. The people, references and figures below are invented to demonstrate the method.

Ten eligible cases consumed 83 minutes so far. Nine completed cases consumed 74 minutes; the unresolved case's nine minutes remain visible and may increase. First-pass acceptance is six of ten cases. Do not report 30 divided by six as the overall pilot effort, because it describes only the easiest successful subset.

Catch the cases that create rework

A retry is another attempt, not another independent completed case. If one case has three attempts, count its outcome once while including all effort. If staff abandon recording after failure, reconstruct only what reliable evidence supports and mark the missing timing.

Do not describe unresolved cases as zero-time failures. Their recorded effort is partial, and their final completion cost is unknown until recovery or an authorized stop decision occurs.

Microsoft: Calculate an average demonstrates weighting by counts. Use the actual outcome-group sizes when summarizing effort, and retain unresolved effort outside a misleading completed-case average.

Decide whether the change is worth keeping

Reconcile the eligible case list, attempt log and final outcomes. Explain the largest failure causes and the input boundaries they suggest. A useful decision may be to improve a review step or narrow the pilot. Report the all-attempts effort before discussing projected capacity, and keep unsupported cases visible in any later summary.

The workflow audit checklist can help record the observed friction before considering a change.