Download the scorecard (Excel, scores itself) (XLSX · 12 KB) · the scorecard to print (PDF) (PDF · 29 KB)
A trial is the best chance you will get to see a system on your own site, with your own people, before you commit to it. It is also easy to come out of one with a feeling rather than a result. A demo is built around the best clip; a trial should be built around a normal week.
This guide sets out how to plan the trial, what to score, and how to read the result. It works for a single system, for two or three side by side, and for comparing a new supplier with the one you have now. Still choosing a type of system? See forklift–pedestrian detection: cameras, tags or on-vehicle systems.
The same area, the same cameras, the same dates. If one system watches the loading bay on the day shift and another watches the yard at night, you are comparing two different sites. Choose one area where people and vehicles really do meet, and keep it the same for every system.
What counts as an event. Write the definitions down: what a near miss is (for example, a person within a set distance of a moving vehicle), which zones are restricted, which PPE applies where, and what speed limit applies in each area. Give every supplier the same list. If the definitions differ, the counts can't be compared.
Who reviews events, and how often. Name the person, and decide whether they review daily or weekly. Review time is one of the things you are testing, so keep it consistent.
How long. Long enough to include nights, weekends if you work them, and the busiest hours. We'd suggest at least two weeks of events.
What you will check against. Missed events are the hardest thing to measure. Decide in advance how you will look for them: near-miss reports, supervisor walks, or a person watching a sample of footage from the same cameras.
These are the rows in the scorecard. Each is scored from 1 to 5 for each system, and you set a weight for each row to match what matters most on your site.
| What to score | How to check it |
|---|---|
| Time from camera access to first events | The date access was given, and the date the first usable events arrived. |
| Events per week, by type and zone | From each system's reports. Useful context, not a score on its own (see below). |
| False events | A person checks a sample of events and marks the ones that weren't what the definition says. |
| Missed events | Compare with near-miss reports, supervisor walks or a sample of footage watched by a person. |
| Clip quality | Does each clip show enough to understand what happened and act on it? |
| Review time | Minutes a person spends reviewing each week. |
| Reports | Can a manager use them without the supplier on the call? |
| Privacy settings | Masking of areas you don't need, face blurring, retention periods that fit your site's policy. |
| Support and changes | How quickly a new zone or detection is set up when you ask. |
More events is not automatically better. A system that reports twice as many events may be seeing more, or may be reporting more false events. Look at the false events per day and the missed events together.
Look at what happened next. The point of a record is action. For each system, count how many events led to something being done: a conversation on shift, a toolbox talk, a change to a route or a rule. A smaller number of events that the team acts on is worth more than a long list nobody opens.
Weight it for your site. A port may weight missed events and clip quality highest; a site with a small safety team may weight review time and reports. Set the weights before the trial starts, so the result isn't shaped after the fact.
A distribution centre with 6 cameras over its yard and loading bays trials two systems, A and B, for two weeks on the same cameras. The site's definitions: a near miss is a person within 3 metres of a moving forklift or HGV; the yard crossing and bay edges are restricted zones; hi-vis is needed everywhere outside the office.
The site weights missed events and false events at 3, review time at 2, and the rest at 1. The scorecard multiplies each score by its weight and gives each system a weighted score out of 5. In this invented case A comes out ahead, mainly on false and missed events, even though B was quicker to review. The team then checks the result against one question: which system's events led to changes on site?
If you already have a system, run the new one on the same cameras while the current one keeps running, and score both on the same rows. Agree the definitions before the trial, so the comparison is fair to both. The switching guide covers what else to check: what you can take with you, contract dates and the people side.
Most sites start with Safety Mapping, a free two-week trial on one camera over the area you worry about most. The camera can be ours, on its own 4G with no access to your network, or one of yours once access is agreed (we check the camera's view first, and if it doesn't show the area clearly, we start on ours). After two weeks you get a written report and a review call, and decide whether to continue. You're welcome to score us with this scorecard. See Safety Mapping.
For the wider questions to ask any supplier, see the buyer's guides for UK sites and US sites.
Long enough to include nights, weekends if you work them, and your busiest hours. We'd suggest at least two weeks of events. Count from the first events, not from the day the camera went up, so setup time doesn't shorten the trial.
Enough to cover one area where people and vehicles really meet, and the same cameras for every system you compare. One well-placed camera over a busy crossing or loading bay can tell you more than several cameras on quiet areas.
The person who would review them if you went ahead, usually someone in the safety or operations team. Keep the same person and the same routine for every system, because review time is one of the things you are measuring.
Go back to the clip and the written definition. If the clip shows the event met the definition, the system that missed it scores lower on missed events. If it didn't, the system that flagged it scores lower on false events. If the definition itself was unclear, fix it and apply the fixed version to both.
The fairest test of any system is one area, two weeks, on the same camera, and a written report you can hold it to. Safety Mapping: a free two-week trial on one camera.
Last checked November 2026