18%
Overtime spend reduced
Recorded across three enterprise accounts within eight weeks of dashboard adoption.
← Back to ProjectsFeatured Case Study // Recorded Outcomes
Data
Workforce Analytics Dashboard: Harri
A custom ETL layer turned raw Harri exports into daily KPI dashboards for labor cost and turnover.
3
Enterprise accounts
The measured reduction was observed across three hospitality operations.
Daily
KPI refresh cadence
Labor cost, turnover, and shift-fill views update from automated exports each day.
Operating Context
Project Brief
Raw workforce exports became a daily decision layer. The system normalized labor data, calculated a small set of operating KPIs, and surfaced overtime and turnover patterns early enough for managers to act.
- Role
- CRM operations analyst and analytics builder
- Scope
- Automated ETL, KPI modeling, and workforce-performance dashboards
- Timeframe
- Eight-week measured outcome window
01 // SITUATION & CHALLENGE
The Operational Bottleneck
Hospitality managers using the Harri platform had raw shift and labor data but no way to spot trends: overtime abuse, high-turnover roles, and scheduling inefficiencies were invisible.
02 // ACTION & ARCHITECTURE
Engineering Strategy & Solution
Built a custom analytics layer on top of Harri's data exports: automated ETL pipeline, KPI dashboards for labor cost, turnover rate, and shift fill time, updated daily.
03 // VERIFIED RESULT
Delivered Operational Impact
3 enterprise accounts reduced overtime spend by 18% within 8 weeks of using the dashboards. Turnover hotspots identified and actioned for the first time.
Architecture
System Flow
- 01
Ingest
Scheduled Harri exports enter an automated processing pipeline.
- 02
Normalize
Python and SQL clean the fields and align account-level workforce definitions.
- 03
Model
The KPI layer calculates labor cost, turnover rate, overtime, and shift-fill time.
- 04
Decide
Power BI views surface daily exceptions and trends for operating managers.
Delivery Conditions
Constraints
- Source exports needed consistent definitions before account-level comparisons were meaningful.
- Managers needed a small operating view, not another dense data dump.
- Refreshes had to run predictably without rebuilding the dashboard by hand.
Judgment
Engineering Decisions
Separate transformation from presentation
Cleaning and KPI logic live upstream, keeping dashboard views consistent as source exports change.
Prioritize operational exceptions
The reporting layer highlights overtime, turnover, and fill-time problems instead of asking managers to inspect every row.
Refresh daily, not continuously
A daily cadence matched the management decision cycle while keeping the pipeline simpler and more reliable.
Supporting System Visuals
More on IG →
Technologies & Protocols Used
PythonSQLPower BIDatadogExcel Automation
Similar Operational Problem?