Cleaning robotsGoal: put robots to work10 weeks
Cleaning robots that get fixed before the customer calls
A young company rents out commercial floor-cleaning robots to offices and shopping centres in Poland and Germany. The manufacturer’s app showed each robot separately, so most problems were discovered when a customer called.
- Client
- Cleaning-robot rental company (robots as a service), PL & DE
- Their team
- Founder & CEO (owner), 2 field technicians, account manager
- Our team
- Engagement lead, 2 engineers
- Timeline
- 10 weeks to production
86% → 95%
robot uptime in contracted cleaning hours
−60%
technician trips to customer sites
84%
of stoppages caught before the customer calls
5 → 2 days
to start at a new site
The business goal
Grow from 45 to 150 robots without tripling the technician team — and give customers proof of cleaning they can renew contracts on.
Where it was
Robots got stuck on cables, ran low on battery mid-route or lost their map after furniture was moved. Technicians learned about it from angry customers and drove to the site — often for a one-minute fix.
45 robots
checked one by one in the maker’s app
Customer calls
robot stuck or not cleaning
Technician drives
1.4 visits per robot a month
What we did
- 1Outcome brief · 1 weekTechnicians listed which errors matter and which can be fixed remotely; uptime baseline measured from robot logs.
- 2Pilot · 4 weeksA fleet dashboard on the manufacturer’s cloud API — status, errors, maps, battery — with alerts in Slack, at 5 sites. 84% of stoppages caught first — continue.
- 3Production · 4 weeksAll sites. Remote actions: restart, re-map, change cleaning missions. Monthly customer reports with cleaned area and coverage maps.
- 4Run & improve · ongoingNew robot models added; a checklist and map templates that cut new-site setup from 5 days to 2.
What did not work
Predicting brush and battery failures from robot logs: there is not enough history yet. We now store all telemetry and will try again at 150 robots.
Who did what
The client’s team
Technicians defined the alerts and remote fixes; the account manager designed the customer report.
Our team
Integration with the manufacturer’s cloud, the dashboard, alerts and reports.
Together
A weekly look at every stoppage the alerts missed.
Before and after
Under the hood
Integration with the robot manufacturer’s cloud API (status, error codes, maps, missions); time-series store for telemetry; alert rules with Slack and SMS; remote actions through the API; report generator with coverage maps; an adapter layer ready for a second robot brand.
Our customers now get a report before they think to ask — and our technicians drive out only when a robot really needs hands.