SOMAT is the software Harlow Stewart uses for safety audits and surveys with airlines, and we're now adapting it for other organisations whose practice has to stand up to scrutiny, from housing providers to financial services.
Observers and assessors record what happens during ordinary work, an independent evaluator reviews each record, and the results are then compared with outcomes you already track, such as complaints or readmissions.
A new filling line was commissioned two months before the audit.
The supplier's self-assessment still describes cleaning validation for the old line.
The auditor can't confirm cleaning validation for the line now making the product.
A corrective action is raised, and a related deviation is recorded at goods-in the following month.
Evaluator: "Answers matched what we found everywhere else on site. Raised so that self-assessments are refreshed after any change to a line."
The customer mentions that their partner died in the spring. The agent moves on to the arrears balance.
A payment plan is proposed without an income and expenditure check.
The customer agrees to a plan above what they said they could afford.
A complaint is received and the plan is broken in its first month.
Evaluator: "The vulnerability prompt was on screen at the time, so I've raised this for coaching, and also because the prompt may need redesigning."
Black mould on the bedroom ceiling above the cot, first reported by the tenant three weeks earlier.
The mould is noted, but no likely cause is recorded and the extractor fan isn't checked.
A mould wash is booked, with no repair to the source of the damp.
The tenant reports the mould again, and the case becomes a formal complaint.
Evaluator: "The fan had no power, so the cause was there to be found. Raised for the inspection checklist, which doesn't ask about ventilation."
The client's mother says she is struggling to cope at home. The conversation returns to medication.
Family intervention isn't discussed, although the care plan is due for review at this visit.
The family is left without support, with the next review set for twelve weeks' time.
The crisis team is called out after a relapse at home.
Evaluator: "The care plan prompt was there. Raised for supervision, and to ask whether twelve weeks is the right default between reviews."
Heavy traffic on the route, with a school crossing patrol ahead.
The car keeps its speed into a junction before the nearside is confirmed clear.
A pedestrian steps out, and the driver has to brake hard and swerve.
No collision. The near miss is logged and the driver is debriefed after the call.
Evaluator: "A good recovery from a position that shouldn't have arisen. Raised for the junction commentary in refresher training."
The class arrives six minutes late, straight from a fire drill.
Equipment is handed out one bench at a time, without the usual monitors.
Two benches are off task while they wait, and the practical is cut short.
The end-of-unit assessment shows the class behind on practical skills.
Evaluator: "Routines held up well until the late start. Raised as a whole-school question about how lessons restart after a drill."
The young person says they may be moving to their nan's in another town. The mentor moves on to the next activity.
The possible move isn't passed to the partner's coordinator, so no handover is planned.
The mentoring relationship ends abruptly when the move happens.
At the six-month check, the young person isn't in education, employment or training.
Evaluator: "This wasn't down to the mentor alone, as nobody was clear on the partner's handover process. Raised with the partner."
A late runway change from air traffic control, with a tailwind on the new runway.
The approach briefing isn't updated for the new runway.
The aircraft is high and fast at 1,000 feet.
The crew goes around and lands safely on the second approach.
Evaluator: "The go-around was the right call. Raised as an example of a late change leaving too little time to brief."
SOMAT brings together work that is often spread across several spreadsheets and systems, from designing a questionnaire or observation scheme to analysing the results. Keeping it in one place means that when an analyst finds a pattern, they can open the sessions behind it and see exactly what was recorded and by whom.
Questionnaires and self-assessments, scored, with documents or photographs attached to the answers as evidence. A survey that changes is replaced by a new version, so earlier results still refer to the questions that were asked at the time.
An observer records what happened as a sequence of events, for example a cue from a customer, the agent's response and how the call ended. Each scheme sets which events may follow which, which keeps observers' records consistent with one another.
An evaluator reviews each session and can record their own answer beside the original, so both are kept and you can see where they differ. Anything that needs attention straight away can be flagged.
Dashboards and Excel exports compare what was recorded with the outcomes you measure. You can click through from a chart to the sessions behind it, and an analyst checks each pattern before it's reported as a finding.
Each finding is reported with the number of cases behind it, whether it held in each area or team when they were looked at separately, and a note from the analyst who checked it.
We're careful with the wording. Most findings show that two things tend to go together, which is a good reason to look more closely or to try a change, although on its own it can't show that one causes the other, and we say so on the finding itself. If later sessions stop supporting a finding, we withdraw it.
Repeat reports within six months of the inspection. The difference appeared in each of the three areas when they were analysed separately. The two were found together, which on its own can't show that one causes the other.
Analyst's note: supports making "likely cause" a required, coded field from next quarter.
Outside aviation, SOMAT is at the pilot stage. We've built example versions for seven sectors so that you can see how it might look in yours, using fictional organisations and made-up data, and the regulations named on each are there to show the kind of question that sector is being asked. A working version would be designed with a pilot partner, around their own questions and the outcome measures they already report.
SOMAT grew out of aviation. On a line observation audit, a trained observer sits on the flight deck during a normal flight and notes the threats and errors that came up and how the crew dealt with them, using a framework called threat and error management (TEM). Harlow Stewart runs these audits for airlines, and SOMAT is the software it records them in.
Much of that structure transfers to other work where people have to deal with the unexpected. The table shows how the steps of an observation on the flight deck correspond to those on a customer call.
| Step | On the flight deck | On a customer call |
|---|---|---|
| Opens | Threat | Customer cue |
| Handled | Managed or not | Heard or missed |
| Slip | Crew error | Agent error |
| Ends | Undesired aircraft state | Unsuitable outcome |
Harlow Stewart brings the aviation safety expertise. It runs observation audits and safety surveys for airlines, and SOMAT records an observation in the same way that Harlow Stewart's observers carry one out.
Elan designs, builds and supports the software. We began rebuilding SOMAT in September 2024, and if you get in touch about a pilot, you'll be talking to the people who built it.
How an observation method from airline safety can be used on customer calls and housing inspections, and the limits to bear in mind when doing so.
What a SOMAT finding contains, what it is careful not to claim, and why a finding can be withdrawn after it has been published.
We're looking for a small number of organisations in each sector to run a six-month pilot with us. You'd help shape the version you use, and if you carry on afterwards, you keep a founding-partner rate.