Find who is missing waves and when drop-off happens, compare responders against non-responders on baseline, flag small-n subgroups, and grade your sample green, amber, or red — so you know whether the outcome data is still representative.
In short: Your outcome data is shaped by who stayed in the study. To handle attrition, identify who is missing waves and when they dropped off, compare responders against non-responders on their baseline values, flag any subgroup too small to trust, and grade the sample green, amber, or red before you report a single result.
Start in a dataset where every participant has a persistent ID across waves. This walkthrough uses DEMO-03 · Workforce Cohort — Vista Workforce Collaborative, multi-wave with stable IDs. Load your Decision Brief first so the attrition check is judged against the outcomes you'll actually report.
You are the Sopact Sense Assistant working over the DEMO-03 · Workforce Cohort dataset (clean data + persistent contact IDs). Load my Decision Brief (decision, audience, outcomes, indicators, evidence standard) first, then wait for my task.
The prompt finds the dropouts and tests whether they differ from those who stayed.
For [POPULATION], identify who is missing waves and when drop-off happens; compare responders vs non-responders on baseline. Flag small-n. Grade green/amber/red.
Five elements make it reliable: the dataset with persistent IDs, the missing waves analysis of who left and when, the responder vs non-responder comparison on baseline, the small-n flag for subgroups too thin to trust, and the grade G/A/R on representativeness.
Run on the Workforce Cohort dataset (DEMO 03) already loaded in Sopact Sense.
GRADE: green | Overall attrition | low and balanced; amber | Early drop | burst right after baseline; red | East subgroup | non-responders differ at baseline
The grade tells you whether the sample still represents the cohort. Green means attrition is low and the people who left look like those who stayed. Amber means a timing problem — an early drop right after baseline that thins later waves. Red means biased attrition — in the East subgroup, non-responders differ from responders at baseline, so any East result is skewed by who remained.
Fix the most biased subgroup with the smallest realistic change.
Take the lowest-graded element above and fix it using only what the program could realistically measure. Show the before → after grade and the single indicator/edit that moves it to green.
Produce a branded "missing & incomplete" report and a shareable link.
Create a 'missing & incomplete' report from this analysis in Sopact branding [or paste your website URL / brand guideline to apply your own]. List every element graded amber or red, what is missing, and the one input that fixes each. Lead with the decision this report informs.
Create a shareable link for this report and open it in a new tab.
Attrition is bias, not just missing rows. If the people who left differ from those who stayed, your outcome data is biased by who remained — the average improves simply because the strugglers stopped answering.
Always compare on baseline. The honest test is whether responders and non-responders looked different at the start. If they did, attrition is non-random and you must say so.
Flag small-n before you interpret it. A subgroup down to a handful of responders can't carry a claim. Mark it small-n and report the count, not just the percentage.
Catch early drop with a light touch. A one-question check-in between waves re-engages people before they vanish for good.
For the most biased subgroup, draft a one-question between-wave check-in that would reduce early drop-off.
Identify who is missing each wave and when the drop-off happens, then compare the people who left against those who stayed on their baseline values. If the two groups differ, attrition is non-random and biases your results — flag it, weight or caveat the affected subgroups, and report attrition rates alongside outcomes instead of burying them.
When participants who are struggling drop out faster than those who are thriving, the remaining sample skews positive and your average outcome rises for the wrong reason. Comparing responders and non-responders on baseline reveals whether that's happening, so you can correct for it rather than report an inflated gain.
There's no universal cutoff, but once a subgroup falls to a handful of responders, a single person's answer swings the result and confidence collapses. Flagging small-n and reporting the raw count keeps readers from over-interpreting a percentage built on too few people.
Open Sopact Sense, paste your program description, and put it to work.
Try in Sopact
