Running robots is not like running ordinary software. When a program fails, a screen shows an error. When a robot fails, a cart stops in an aisle, an arm freezes mid-task or a sensor reads nonsense, and somebody has to walk over and deal with it. Because the consequences are physical, operations teams tend to focus on the next stoppage. The quarterly look-back is the opposite habit: a calm review, once a season, of what the last three months were trying to tell you. Here is how it works, in plain words.
What the look-back is
It is a ninety-minute meeting held once a quarter. The team reviews three lists: stoppages and faults, software or configuration releases, and changes to tools and equipment. They look for patterns, pick the single change most likely to help next quarter and write it down with an owner. That is the whole thing. No special software, no consultants, no slides.
The reason it works is that robot operations produce a lot of small events, and humans are bad at remembering small events. A look-back puts them on one page where patterns become obvious.
It helps to picture the meeting room. On one wall is the fault log printed out, on another the release notes, on a third the maintenance record. People walk along the walls with a marker, circling anything that happened within the same week. When two circles land next to each other, someone says “hang on, what changed then?” and the conversation starts. It is a low-tech method for a high-tech setting, and it works because it puts evidence in front of everyone at once.
The three lists, explained
Stoppages and faults. Anything that halted work or needed a person to step in. Include the ones fixed in two minutes by someone pressing a reset button. Those minor events add up and are often the most revealing.
Releases. Any change to the robots’ software, maps, routes, safety settings or task rules. Include who made it and when. If your fleet runs on a vendor’s software, include their updates too.
Tool and equipment changes. New sensors, replaced batteries, altered chargers, different floor markings, a swapped scheduling tool. Physical changes belong here as much as digital ones, because in robotics the line between the two is thin.
A worked example, no jargon
Imagine a small warehouse running a handful of autonomous carts. Over a quarter, the team logged fourteen stoppages. When they tagged each one with a plain cause, six came down to carts losing their position near one corner of the floor, four were battery-related, and four were assorted one-offs.
The release list showed a map update in week three. The equipment list showed that a set of reflective floor markers had been moved in the same week to make room for new shelving. Reading the lists side by side, the team realised that the position losses started right after that week. The map update and the moved markers had never been connected, because two different people made the changes for sound reasons.
The change they chose for next quarter was small: whenever markers or shelving move, someone checks the map the same day and runs a test route. It costs ten minutes and would have prevented six of fourteen stoppages. Understanding how position and perception work helps here; an explainer about how sensing parts are worked into factory robots gives useful background.
Why a look-back beats a faster reaction
Teams sometimes respond to repeat stoppages by trying to react faster: more staff on the floor, quicker alerts, shorter repair times. These are worthwhile, but they treat symptoms. A look-back asks whether the stoppages need to happen at all. In our example, faster reaction would have made the corner problem less painful without ever removing it.
Operators on the floor deserve a seat at the session. They see stoppages first-hand and often hold the explanation that no log captures: the cart that always hesitates near the loading bay, the aisle where lighting changes in the afternoon, the charger that is awkward to reach. A look-back that includes only managers and engineers loses this knowledge, and it is the cheapest, most accurate source the team has.
Keeping it safe and sensible
One note belongs here. A look-back is a review of operations, not a safety certification, and it does not replace the checks and procedures your site, your vendor and any applicable rules require. Treat any safety-related stoppage by its own proper process first, and then include it in the quarterly list as a pattern to examine. The look-back adds a layer of learning; it never replaces the layers that keep people safe.
A quick word about another trade
Review habits like this are common wherever work follows busy and quiet stretches. The website VehicleRentalSystem describes software for companies managing vehicle rentals, an industry that tracks its own cycles. The overlap with robot fleets is slight, though the habit of pausing to review is the same.
Running your first one
- Book ninety minutes in the last week of the quarter, with everyone who works on or near the robots.
- Collect the three lists beforehand from fault logs, release notes and maintenance records.
- Tag each stoppage with a plain cause, using everyday words.
- Read the lists against each other, looking for events that happened in the same week.
- Settle on one change, give it a person to own it and agree a test of success.
If your fleet involves automated guided vehicles, a handy reference for the wider context is an overview of driverless carts and AGV systems in storage facilities, which can help you decide where the themes in your own quarter fit.
Common mistakes
The first is skipping the small events. A look-back based only on dramatic failures misses the quiet pattern. The second is choosing too many changes. Robot operations are already complex, and every change carries its own risk, so a single well-chosen adjustment is wiser than a bundle. The third is forgetting to check last quarter’s change before choosing the next. Always begin by reading the previous page.
A fourth, subtler mistake is treating the exercise as a hunt for faulty equipment. Sometimes the equipment is faulty, but more often the cause sits in the gaps between changes: a map nobody updated, a charger schedule that no longer matched the shifts, a setting changed on one robot and not on the rest.
A final practical point concerns timing. Hold the session soon after the quarter ends, while the memory of events is fresh and before the next quarter’s urgent work arrives. Many operations teams pick a quiet shift or the day after a scheduled maintenance window, when the floor is calmer and nobody is half-listening for an alarm. Protect the slot the way you would protect a maintenance window, because it quietly pays for itself in fewer stoppages later.
Also decide in advance who will keep the page afterwards. In small teams this duty drifts, and by the next quarter nobody can find last time’s notes. Name a custodian, store the page beside the maintenance records and mention it in onboarding, so that the history survives staff changes and shift rotations without anyone needing to remember it.
What you get for your ninety minutes
Over a few quarters, the benefits are tangible. Repeat stoppages fall, because the main causes are tackled one by one. Handovers improve, because the pages explain what changed and why. New team members learn faster, because they can read a history of how the fleet behaved. And the work feels calmer, because the team has a deliberate way to decide what to fix first.
That is the plain-language version. Gather three lists, find the pattern, choose one change, check it next time. It is not exciting, and that is why it works.