--- id: v10-check-whether-a-change-worked slug: check-whether-a-change-worked sequence: 11 title: "Check whether a change worked" url: https://etiscope.com/support/guide/check-whether-a-change-worked/ video_state: draft video_duration_seconds: 160.811 video_url: https://media-etiscope.jointaction.cloud/lessons/check-whether-a-change-worked/60655de439d76ce1174025b788cfbd7f26ad0cd65009cd79e65f3c078796b4ec/video.mp4 captions_url: https://media-etiscope.jointaction.cloud/lessons/check-whether-a-change-worked/60655de439d76ce1174025b788cfbd7f26ad0cd65009cd79e65f3c078796b4ec/captions.vtt --- # 11. Check whether a change worked One assessment tells you where the risk is. Two tell you whether something changed. Here you'll learn the two ways to get that second assessment, and the one boundary that keeps both honest. **Before this lesson:** 10. Create and track risk controls (https://etiscope.com/support/guide/create-and-track-risk-controls/) ## What you'll be able to do The learner can run both comparisons and knows the one boundary — changing entered values models the same captured movement; a change in how the work is done needs a new capture. - [choose-the-comparison] Choose between re-assessing after a control and modelling first, based on the question being asked. - [reassess-after-a-control] Re-assess after a control with Repeat from the original assessment, and compare the two side by side. - [model-with-duplicate] Model a proposed change with no new capture by duplicating an assessment, editing its entered values and saving to re-score it. - [read-effectiveness-chart] Read the risk control effectiveness chart as the spread of the assessments that exist on the control's work task. - [respect-the-boundary] State the boundary: what-if models exposure; a change in movement needs a new capture. ## Video - MP4: https://media-etiscope.jointaction.cloud/lessons/check-whether-a-change-worked/60655de439d76ce1174025b788cfbd7f26ad0cd65009cd79e65f3c078796b4ec/video.mp4 - Captions (WebVTT): https://media-etiscope.jointaction.cloud/lessons/check-whether-a-change-worked/60655de439d76ce1174025b788cfbd7f26ad0cd65009cd79e65f3c078796b4ec/captions.vtt **Chapters:** 0:00 Two ways to compare · 0:20 Re-assess after a control · 1:06 What-if modelling · 1:52 The effectiveness chart · 2:16 The one boundary ## The lesson ### Two ways to compare So far you've produced one number about one task. That's enough to decide where to act, but not enough to say whether acting worked. There are two ways to get a second assessment, and which one you use depends on the question you're asking. **Re-assess after** — The change is a control that has been built, installed or rostered. Assess the task again once it's genuinely in place. **Model first** — The change hasn't been made yet, and might not be. Model it against the capture you already have, before anyone spends money. Neither is better than the other. Modelling is the cheaper and more constrained of the two; re-assessment is the one that can show a change in how the work is actually done. ### Re-assess after a control When a control from [Create and track risk controls](/support/guide/create-and-track-risk-controls/) is implemented, assess the task again — **the same task, the same way you assessed it before**: same work task, same boundaries, a representative sample of the same work. Create that follow-up assessment in the app with **Repeat**: on the work task's screen, open the original, come back, and choose **Repeat** in the top bar. It carries over the work task, person, location, work area, shift, tags and notes, but not the job — set that again, and all that's left is to capture. As well as saving time, **Repeat** helps make sure the follow-up assesses the same thing as the original. Then open the two assessments and compare the scores and the graphs. **What you can say is what you saw:** the same task, before and after, and where the movement changed. Two assessments of one task are evidence about that task. Describe them that way, rather than as a controlled trial, and a good result will stand up to scrutiny later. > **Definition** — Repeat copies the last assessment you opened, except its job > **Definition** — Shown here: two assessments of one task, to demonstrate the screen ### What-if modelling The second comparison needs no new capture at all, and you do it in Etiscope Capture on the iPad or iPhone rather than in the dashboard. Suppose the proposal is fewer repeats per shift: a rotation rather than a rebuild. You'd like to know whether it's worth making the case before you make it. 1. In the app's assessment list, swipe the assessment's row and choose **Duplicate**. 2. Open the duplicate, then **Report actions** → **Add extra information**. 3. Change the value the proposal would change — repeats, duration, perceived load, exposure — and save. The duplicate re-scores immediately. The original is untouched, and the difference between the two scores is your model of the change. This is the technique behind a business case: a number attached to a proposal, produced immediately, from a capture you already own. > **Definition** — A duplicate shares the original's video and sensor data > **On screen** — Duplicate · Add extra information · Save ### The effectiveness chart Back in the dashboard, the control's **Stats** tab draws risk control effectiveness: the spread of scores for every assessment on the control's work task, as a box-and-whisker. As before-and-after assessments accumulate, this is where the population answer lives — not one task's story, but the distribution across everything captured on that work task. **Read it for what it shows:** the spread of the assessments that exist. It doesn't show that the control caused the shift, whatever the chart's title suggests. Open the records from the **Assessments** tab when you want the stories behind the spread. > **Definition** — Each box: the spread of scores for assessments linked to one control ### The one boundary One boundary keeps both comparisons honest, and it's the most useful thing to take away from this lesson. Editing values re-scores **the same captured movement**. So what-if models changes to exposure and organisation: durations, repeats, loads. Those are real changes and they move the score legitimately. It can't model a change in posture or technique, because the movement in the data hasn't changed. Editing a duplicate can't show the worker bending less, so a smaller number produced that way wouldn't mean anything. If the change alters how the body moves — new equipment, a redesigned workstation, a different way of doing the job — capture it again. **Model the paperwork; re-capture the movement.** ## Mistakes that matter - Modelling a technique change with what-if. The captured movement is unchanged, so a number produced that way doesn't describe the change. - Comparing captures of effectively different tasks. Different boundaries, a different sample or a different work task make the before and after hard to compare, however similar the titles look. - Reading the effectiveness chart as proof because of what it's called. It shows the spread of the assessments that exist. ## Not covered here - Capture mechanics — normalising, recording, checking — see lesson 5, Capture an assessment (https://etiscope.com/support/guide/capture-an-assessment/) - The extra information sheet this technique edits — see lesson 6, Complete the record: Extra Information (https://etiscope.com/support/guide/complete-the-record/) - Creating the control being checked — see lesson 10, Create and track risk controls (https://etiscope.com/support/guide/create-and-track-risk-controls/) **Next lesson:** 12. Monitor the program in the dashboard — https://etiscope.com/support/guide/monitor-the-programme/