I ran a 21-day occupancy-driven cleaning pilot at one of our mid-sized office sites to test a simple hypothesis: can cheap occupancy sensors give us reliable enough data to cut night shifts by 30% without affecting cleanliness or client satisfaction? The answer was yes — with careful design, clear KPIs, and a practical decision framework. Below I share the exact blueprint I used so you can run the same pilot for your sites.
Why a 21-day pilot?
I chose 21 days because it's long enough to capture weekday/weekend patterns, a couple of atypical days (meetings, events), and staff learning curves. Shorter pilots gave noisy results; longer pilots delayed decisions. The 3-week window balances speed and statistical usefulness for occupancy trends.
Goals and success metrics
From the start I set three measurable goals:
Key performance indicators (KPIs):
Sensor selection — cheap, reliable and easy to deploy
I tested a combination of low-cost sensors that are widely available and simple to manage. My priorities were: non-invasive install, battery life, data accessibility, and GDPR-friendly operation.
I avoided cameras entirely to keep the pilot simple and privacy-friendly.
Deployment plan and zones
Divide the site into functional zones. For our pilot we used:
Sensor placement rules I followed:
Data collection and simple architecture
Keep the architecture minimal. For the pilot I used battery PIRs connecting to an inexpensive Zigbee gateway (CC2531/ConBee II) linked to a Raspberry Pi that logged all events into a simple PostgreSQL database. Alternative: many PIRs have Bluetooth/Wi‑Fi options and can push data to cloud dashboards (but that increases running costs).
Data model: timestamp, sensor_id, event_type (motion/no-motion/contact open/close), battery, zone_id.
Sampling rules:
Decision rules to reduce night shifts
I converted occupancy into operational rules. The rules must be simple so shift supervisors can apply them without ambiguity.
These thresholds are conservative — they keep hygiene-critical areas protected while targeting low-use desk zones where night cleaning yields little benefit.
Communication and stakeholder buy-in
Before switching any shifts, I briefed three groups:
Transparency matters. I shared dashboards and a simple FAQ so staff knew occupancy data was anonymous (no cameras, no personal device tracking in the dashboard). A one-week feedback window was kept active.
Sample 21-day timeline
| Day | Activity |
|---|---|
| 1–3 | Install sensors, verify data flow, baseline night shifts |
| 4–10 | Collect occupancy, refine buckets, start QA checks |
| 11 | Apply decision rules to propose reduction list — review with client & supervisors |
| 12–21 | Implement reduced night shifts for selected zones; monitor QA & complaints; reinstate if necessary |
Quality assurance and fallback
I kept a strict QA process: cleaning supervisor visits and scores a checklist twice weekly for switched-off zones. Scores had to match baseline within a 10% margin. A single justified complaint or a QA drop below threshold triggered immediate reinstatement of night cleaning for 7 days and a root-cause review.
Costs and expected savings
Example cost breakdown for our trial site:
| Item | Cost (GBP) |
|---|---|
| 10 PIR sensors (Aqara) | £120 |
| Zigbee gateway + Raspberry Pi | £80 |
| Installation & config (1 day) | £200 |
| Misc (cables, mounts) | £50 |
| Total one-off | £450 |
Night shift saving: if one night operative = £90/night (incl. NI), and the site ran 5 night shifts/week, cutting 30% equates to 1.5 fewer shifts/week = ≈£135/week savings → payback under 4 weeks. Even with conservative numbers, ROI looked compelling.
Results I observed
Over the 21 days we reduced night shifts by 33% for targeted zones with no measurable drop in QA scores and zero justified complaints. The occupancy data revealed predictable patterns: certain desk blocks were virtually unused after 4pm, while small meeting rooms spiked briefly and benefited from targeted morning cleans. The sensors were cheap and robust; battery changes were not an issue in the pilot timeframe.
Two practical lessons:
Risks and mitigations
Key risks I planned for:
If you’d like, I can share the scripts I used to aggregate sensor events into 15-minute buckets and the simple dashboard layout we used for the client. Running a low-cost occupancy pilot is one of the fastest ways I’ve seen to reduce unnecessary labour costs while keeping workplaces clean and users happy.
