Skip to content

Part 2 of 7 · Safety incident logger series ~5 min read

How a report takes forty seconds

There is a straightforward relationship between the number of fields on a safety form and the number of reports a business receives, and it is not a gentle one. This post is about getting to three.

Key takeaways

  • Three inputs: what happened, a photo if there is one, and the anonymity choice.
  • Location, time, site and reporter are inferred, and all are editable.
  • The anonymity choice is offered first, because it changes what people write.
  • There is no category dropdown. Categorising is done afterwards from the description.
  • A report can be sent with no photo and no location. Neither is required.

Three inputs

How a safety report is made in forty secondsA vertical chain of five steps entered by a box labelled Tap the saved link on the home screen. Step one asks name or anonymous, always first; choosing anonymous exits to a state where no follow-up is possible, and the form says so. Step two asks what happened, in one box, in the reporter's own words. Step three offers a photo, optional and one tap. Step four infers the rest: the time, the place and the site, using site boundaries from the sites sheet. Step five is Sent, in about forty seconds. A note says anonymity is asked before the description because it changes what people write.AWS ACCOUNTTap the saved linkhome screenName or anonymous?asked first, alwaysAnonymousno follow-up possible,and it says sochosenWhat happened?one box, your wordsA photo?optional, one tapInfer the resttime, place, siteSite boundariesfrom the sites sheetSentabout forty secondsAnonymity is asked before the description, because it changes what people write.
Fig 1. The whole reporting flow. Asking about anonymity first is a small ordering decision that materially changes the content of the reports you receive.
  • App integration
  • Machine learning
  • Security & identity
  • Front-end & mobile

Why anonymity comes first

If somebody types a description and is then asked whether to attach their name, they have already written it as a named person. Asking first means an anonymous report is written as an anonymous report, which is frequently a different and more useful document — particularly when what somebody wants to say involves how a job is usually done rather than a one-off.

The trade-off is stated in one line where the choice is made: “Anonymous means we cannot ask you anything about it.” That is genuine and it is the whole cost, and most people who choose anonymity have understood it and chosen it anyway.

What is inferred

FieldInferred fromEditable?
TimeNowYes — “this happened yesterday” is common
LocationThe phone, if permittedYes, and it can be typed instead
SiteWhich site boundary the location falls inYes, and it is a short pick list
ReporterThe signed staff linkOnly by choosing anonymous
CategoryThe description, afterwardsBy whoever triages, not the reporter
SeverityThe description, afterwardsBy whoever triages, not the reporter

The last two rows are the important ones. Asking a reporter to categorise an incident makes them think about how the business classifies things instead of about what happened, and asking them to rate severity invites them to talk themselves out of reporting: “it was only a near miss” is how a near miss goes unreported.

No dropdowns

The strongest temptation in this design is a category list, because it makes everything downstream easier. It is also the single most effective way to reduce your report count, for a reason that is easy to miss: a person standing in a yard with a phone who cannot see a category that fits will assume the system is not for this and stop.

How the number of form fields affects the number of safety reportsA horizontal row of five boxes. Eleven fields: about four reports a year. Six fields: maybe twenty. Three inputs: hundreds. Categorised after: by one person. Same analysis: with far more data. A note says categorising afterwards costs one person a few minutes a week and multiplies the input.FIELDS VERSUS REPORTS11 fields~4 reports a year6 fieldsmaybe 203 inputshundredsCategorised afterby one personSame analysisfar more dataCategorising afterwards costs one person a few minutes a week and multiplies the input.
Fig 2. The trade the whole design is built on. Everything a dropdown would have captured can be recovered from the description afterwards, and the description only exists if the form was short enough to fill in.

The photo

Optional, and it does more work than any field would. A photograph of a stack of pallets carries the height, the position, the surroundings and the lighting, none of which anybody would have typed. It also makes the report concrete for whoever reads it, which measurably changes how quickly an action gets created.

It uploads with a presigned PUT straight to S3, so a phone on a yard connection is not holding anything open, and the report is submitted whether or not the upload has finished. A report that fails because a photo did not upload is a report lost.

Next: how triage routes it.

All posts