WEEKLY PERFORMANCE INSIGHTS
A locally-run AI-assisted review tool and weekly PDF scorecard system that inverted the quarterly review cycle — shifting reps from passive recipients of verdicts to active mid-quarter coaches of their own performance.
Two problems. One shared cause.
The first month of every quarter was already spent.
The global scorecard dataset arrived with 300–400 records spanning all roles. Before a single Performance Review could be written: isolate the team’s records, complete exclusion review for each one, then write the reviews, then book the delivery meetings. Two to three weeks from scorecard release to delivery. The entire first month of the new quarter consumed closing out the previous one.
Feedback on Q1 arrived in late April.
A case that represented a clear coaching opportunity in early January was buried under three months of subsequent work by the time the review landed. The specifics — the customer, the troubleshooting sequence, the decision point — were no longer recoverable. The coaching moment existed. The learning opportunity did not survive the delay.
The entire performance feedback cycle — review, exclusion, write-up, delivery, and rep visibility — was compressed into a single event at the end of every quarter.
Match the methodology before you compress the cadence.
Before the system could be built, its scoring methodology had to be understood from the inside out. The official scorecard is generated from a set of data sources that pull from the CRM. To reconstruct that score at a team level — accurately and on a weekly cadence — required reverse-engineering how those data sources were constructed: what filters were applied, what records were included or excluded by default, what criteria determined whether a record was in scope.
The goal was not to approximate the official score. It was to match it exactly — so that the weekly number a rep received was a reliable predictor of where their quarterly score would land, not a parallel metric that told a different story.
Four data streams. One weekly cadence. One PDF per rep.
Four parallel Saved Searches, each scoped tightly to frontline reps on the team, each calibrated to avoid the CRM’s result-set size limits, each delivering its output as a weekly CSV. Every week, four CSVs arrive. Together they contain the raw material for a complete picture of the team’s performance from the case level up. A locally-run tool processes the four streams, surfaces records for exclusion review in manageable weekly increments, and generates an individualized PDF scorecard for each rep. The PDF includes scores per metric alongside the underlying records — exclusion decisions visible, manager notes embedded, coaching guidance traveling with the data.
THE FEEDBACK LOOP
Scrub through a fictional rep’s quarter week by week. Watch the trend lines build. See the coaching notes appear.
What this system carried forward.
A review cycle is only as fast as its slowest task.
The formal Performance Review was not slow because writing it was hard. It was slow because exclusion review had to happen first — deferred until the quarter was already over. The system did not make any individual task faster. It redistributed them across the quarter so that none were standing in the line when the scorecard arrived.
Match the methodology before you compress the cadence.
The value of a weekly score depends entirely on whether it predicts the official quarterly score. If the exclusion logic approximated rather than replicated, the weekly number would become noise rather than signal. The investment in reverse-engineering the official dataset’s construction was not preliminary work. It was the work.
Freed capacity finds its own highest use.
The time recovered from the exclusion backlog was not directed into more administrative work — it became available for development conversations that had previously been too infrequent to be useful. Individual Development Plan conversations moved from every six months to every quarter. The system created the conditions; the investment followed naturally.