A smart-vending operator learned about broken machines from angry location owners, not from the machines themselves. We turned raw telemetry into deduplicated, diagnosed service tickets that slot into the nearest technician's route — cutting machine downtime 44%.
Typical results
46 hrs
Average time back online
88%
First-visit fix rate
85%
Alert noise filtered out
Figures reflect typical results for this kind of automation.
The challenge
With 900 machines in the field, the telemetry system fired over 1,200 alerts a month — duplicates, blips and real failures mixed together. Dispatchers triaged by gut feeling, technicians drove out without knowing which part to bring, and machines sat dark for more than three days while sales and location relationships bled.
Before and after
Time back online per failure
−44%First-visit fix rate
+44%Wasted technician trips
−80%The automation
We built a pipeline that deduplicates every alert and diagnoses it against a library of known error patterns — a power blip that resolves itself never becomes a ticket. Genuine failures open a service ticket automatically, pre-filled with the machine's location, access notes and the parts most likely needed, then slot into the nearest qualified technician's route. A fleet dashboard tracks uptime and repeat offenders for the operations team.
How it works
- 01
Every alert deduplicated
Telemetry anomalies and offline pings from 900 machines are collapsed into single incidents — no more five tickets for one dead compressor.
- 02
Diagnosed before dispatch
Each incident is matched against known error patterns; self-resolving blips are watched, real failures move forward.
- 03
Tickets that brief the tech
Genuine failures open a ticket with location, access notes and the parts most likely needed already attached.
- 04
Routed to the nearest tech
The ticket slots into the closest qualified technician's route, weighted by machine revenue and downtime.
- 05
Uptime tracked fleet-wide
A live dashboard shows downtime, repeat offenders and fix rates, feeding maintenance planning.
Starts from
- Machine telemetry & offline alerts
- Location & access notes
Delivers
- Ticket with likely parts
- Slotted into nearest route
- Fleet uptime dashboard
The impact
Machine downtime dropped 44%, with average time back online falling from 82 hours to 46. First-visit fix rate climbed from 61% to 88% because technicians arrive with the right parts, wasted trips fell from 45 a month to nine, and 85% of alert noise never reaches a dispatcher.
Questions about this automation
How does it stop duplicate alerts from becoming duplicate tickets?
Alerts from the same machine within a set time window are grouped into one incident before anything else happens. The incident is then compared with a library of known error patterns, and short blips such as a brief power loss are watched for a set period. Only failures that persist go on to become service tickets.
How does the ticket know which parts the technician needs?
Each known error pattern is mapped to the parts most often used to fix it, based on past repair records. When a ticket opens, the AI attaches that likely parts list along with the machine's location and access notes, so the technician can load the van before leaving instead of making a second trip.
How are tickets assigned to technicians?
The ticket goes to the nearest technician qualified for that machine type and is slotted into their route. Priority is weighted by the machine's revenue and how long it has been down, so high-earning sites are fixed first. A dashboard tracks downtime and repeat failures. It is a multi-step build, usually rolled out in stages, starting with alert deduplication.