The signal layer behind 1,000 reps

01The context

I joined Cisco's global sales analytics team in 2019. The brief was modest: build dashboards. The actual problem wasn't.

1,000+ account executives across 47 countries were operating on stale data. Reports compiled overnight. Pricing decisions made on yesterday's signal. Analyst capacity was the official bottleneck — every week brought new requests for "just one more report" and the queue grew faster than the team could clear it.

The team's response, historically, had been to hire more analysts and build more reports. I diagnosed it differently.

02The diagnosis

The bottleneck wasn't the analysts. It was the compile step.

Reps weren't underperforming — they were operating blind. Analysts weren't underdelivering — they were stuck in 12-hour report-compile loops, where the actual analytical work was a fraction of the elapsed time. Building more reports would compound the problem. Every new dashboard added another compile, another fragile pipeline, another four hours of weekly maintenance.

The right move was eliminating the compile step entirely. Replace the manual report layer with real-time, self-serve infrastructure. Move analysts up the value stack — from compiling to interpreting. Move reps up too — from reading week-old PDFs to querying live data themselves.

The fix wasn't more reports. It was deleting the report.

03The build

Sole engineer-PM on the platform side, in coordination with data engineering and ML teams. Shipped over four quarters.

1. Real-time data integration Aggregated CRM, ERP, and external sources into a unified hub. 2.5M sales records processed daily, 99.7% uptime SLA. Reps queried live data instead of waiting for the next report cycle.
2. Self-serve dashboards Drag-and-drop customization. Account executives configured their own views without an IT request or an analyst ticket. Time-to-insight collapsed from hours to minutes.
3. Predictive models on deal attributes XGBoost classifiers trained on 24 deal features. Flagged $23M of at-risk pipeline (87% accuracy validated against actual outcomes) and surfaced $8.5M in cross-sell opportunities — sales closed 34% within 90 days.
4. Pricing strategy from 3,200 deals Synthesized win-loss patterns across 18 product lines into a tiered pricing recommendation. Presented to VP Product and VP Sales. Restructure launched Q3 2020.

04The result

Validated by Finance via independent time-motion study. Adopted by 1,000+ AEs globally within two quarters of full rollout.

Three features Sales had requested were reprioritized off the back of the deal-attribute analysis, accelerating time-to-market by one quarter.

05The pattern

This wasn't on the roadmap. The brief was "build more dashboards." I diagnosed the bottleneck — compile latency, not analyst capacity — and architected the solution that the brief, taken literally, would have made worse.

That's the operator move I'd run again, twice: at Mesh-AI, the brief was "scale the consulting motion" and the actual problem was that consulting wasn't the product yet. At OM, the brief — to the extent there is one — is "score jobs against my profile" and the actual problem is that the signal layer underneath every GTM team is broken.

The credential isn't building dashboards or pipelines or scoring engines. It's diagnosing what management, the team, or the market is asking for incorrectly — and shipping the thing that actually solves the underlying problem instead.