A legal practice: 1,900 lost billable hours recovered annually
A commercial litigation set in the City was reconstructing billable hours from memory and spreadsheets at month-end. The partners suspected they were under-billing; they did not know by how much.
The reading
Fourteen fee earners, three partners, a practice manager who had held the firm together since 2011. The set billed by the six-minute unit, as is customary, but the capture of those units was anything but customary. Each fee earner kept their own spreadsheet, with their own conventions, submitted to the practice manager on the last working day of the month.
The partners' private suspicion, confirmed in our first conversation, was that fee earners were under-recording because reconstructing a month from memory is a lossy process. The practice manager estimated the shortfall at ten per cent. We thought it was higher.
The firm's case management system had, in fact, an excellent docket: every document opened, every email sent, every call logged. The docket and the billing simply did not speak.
The docket already knew
The interesting observation was that the firm already had a near-perfect record of what each fee earner had done. The case management system logged document access with second-level precision. The telephone system logged calls. The email system logged sent items. None of this flowed into the billing spreadsheets.
Fee earners were, in effect, being asked to reconstruct, from memory, data the firm already held with precision.
- 14 fee earners, 14 different time-recording conventions
- Month-end reconstruction averaging 4.5 hours per fee earner
- Estimated under-recording of 10-15%
- Docket data complete but siloed from billing
- No partner visibility of WIP mid-month
What we built
We wrote a reconciliation layer that read the docket, the telephone log, and the sent-items folder, and proposed billable units against each matter. The fee earner reviewed the proposal, adjusted where necessary (a long document review was not always the same as a long document open), and approved. The whole process took around twenty minutes per week instead of 4.5 hours per month.
The technical choice we are most proud of: we did not try to be clever about inferring the nature of the work. We presented the raw signals (document opened, duration, email sent, call made) and let the fee earner characterise them. The system handled the arithmetic; the fee earner handled the judgement.
The stack was deliberately conservative: a .NET service reading from the existing SQL Server case management database, a lightweight web front-end for review, and a direct write to the billing module.
The numbers at ninety days
| What | Before | After |
|---|---|---|
| Fee earner time on time-recording | 4.5 hr/month | 1.3 hr/month |
| Billable units captured per fee earner | baseline | +14% |
| WIP visibility for partners | Month-end | Daily |
| Disputed time entries per quarter | 23 | 4 |
| Annualised recovered billables | - | ~1,900 hours |
The system that was already there
The fee earners were not under-recording out of carelessness. They were under-recording because the firm was asking them to do, by hand, a thing the firm's own systems had already done electronically.
The one-week training test here was almost trivial: a new fee earner joined in month four and recorded time correctly on day one. The system did the remembering, which is what systems are for.