Is your sales rep underperforming, or just under-resourced?
United Rentals runs its sales org as Territory → District → Region. I dug into ~862K rows of zip-level sales data to separate rep performance from opportunity — and cleaned up a genuinely messy dataset along the way.
The question
United Rentals sells construction and industrial equipment through a sales hierarchy: reps own territories, territories roll into districts, districts into regions. The problem is that raw performance numbers alone can't tell you whether a rep is doing badly — or was simply handed a territory with nothing to sell into. I approached it in layers: descriptive analysis first, then performance metrics (revenue per active account, activity-to-revenue ratios), comparative analysis across regions, and workload analysis against forecast opportunity.
Cleaning the data was the actual analysis
About 8.66% of rows were missing rep-level identifiers — and it wasn't random. Grouping the gaps by region showed they were concentrated almost entirely in Northeast Canada, Midwest, and Pacific West, which pointed to a data-collection gap in those regions rather than noise.
Before dropping those rows, I checked what it would cost: rows with missing rep data accounted for only 2.03% of total Rental Revenue TTM. Losing 8.66% of rows to protect the integrity of every rep-level metric was worth a 2% revenue trade-off — a decision made on evidence, not a default "just drop the nulls."
What the outliers were hiding
Outlier analysis on rental revenue and active-account counts turned up the same pattern again: 41–59% of outliers across nearly every revenue metric traced back to Western and Northeast Canada. That's not just a cleaning footnote — it's a signal that Canada may behave structurally differently from the US book of business (pricing, market maturity, or both), which changes how you'd even compare a Canadian rep to a US one.
Findings that shaped the follow-up work
- Rental revenue is heavily right-skewed — the 25th, 50th, and 75th percentiles were all zero, meaning a large share of territories generated no revenue at all in the window analyzed.
- Active-account counts varied widely by rep, pointing to real workload imbalance rather than uniform territory sizing.
- Call-hour logging was near-zero for most reps but very high for a few — worth separating "not calling" from "not logging calls" before judging effort.
- Findings were built into a Tableau dashboard and presented to stakeholders, informing a restructuring of data capture across two sales regions.