Check Use on the Job
When To Use
The sponsor asks "did anyone actually do anything with it?" and the forms cannot say. The last-day answers were warm, the programme ended weeks ago, and nobody has looked at the work since. Run this to answer one question: where is the new behaviour showing up at work, where is it not, and what is in the way?
When Not To Use
If the programme ended days ago, use has had no time to show: run Write the Feedback Form now and come back. If the budget holder wants a money figure, run Estimate the Training ROI; answers about use do not add up to a return. A manager noting obstacles week by week uses the barrier log from Brief the Managers.
Inputs
- The behaviours people should be using by now, and the leading indicators, from your evaluation plan if you have one
- When the programme ended, how many people took part, and how the follow-up can reach them
- The existing reports that hold the group-level work measures, and your minimum group size for reporting
If you have none of this, I start from the programme title and what it was meant to change at work, and mark the output as a first draft.
Approach
This is Kirkpatrick Level 3 (behaviour on the job) and the transfer tiers of LTEM, both from their originators' public pages. Results are read by group; nobody is selected by how they perform. Self-report alone is weak and work measures alone are mute, so the two are read side by side, and volunteers add the detail neither can give. The failure it prevents: one enthusiast's story reaches the sponsor as proof, while the approval step that stopped everyone else from trying never appears on anyone's list.
Workflow
- Ask three questions: which behaviours should be visible at work by now, how long after the programme do you want to ask, and who must never see raw answers (usually line managers)?
- Set the timing: a delay of [weeks] that you choose. Too early and use cannot show; too late and people no longer remember which course it was.
- Draft the anonymous follow-up, three to five items: where have you used [behaviour], what happened, what got in the way, what would help. Add a separate route to volunteer for a conversation, kept apart from the answers so that volunteering never unmasks an answer.
- List the group-level work measures to read beside it: the leading indicators from the evaluation plan, taken from reports that already exist. Never per person. Note what else moved those measures in the period.
- Write the conversation guide, for volunteers only, after written consent (the purpose, an anonymous write-up, the right to stop or withdraw). Same order each time: what did you do, what followed, what could someone see, what helped, what blocked. You hold these conversations. Volunteers come forward; nobody is picked as a success or a failure.
- Once the answers are pasted in, sort them into a barrier list: the barrier, how often it appeared, who could remove it. Most barriers turn out to be conditions (time, tools, a process step), and the list says so.
- Write the limits into the output: self-report, a small volunteer sample, what it cannot show. Suppress every group under the minimum.
Output Format
# Use on the Job Check
Programme: [name] | Ended: [date] | Asked after: [weeks, set by you] | Responses: [count] of [group size] | Minimum group reported: [number]
## Anonymous follow-up
1. Where have you used [behaviour] since the programme? [options, including "not yet"]
2. What happened when you did? [open]
3. What got in the way? [open]
4. What would help? [open]
Volunteer route: [separate link or contact, never stored with the answers]
## Group measures to read beside it
| Leading indicator | Existing report | Before | After | What else changed in the period |
|---|---|---|---|---|
| [indicator] | [report name] | [your figure] | [your figure] | [tool, workload, process] |
## Conversations with volunteers
Consent in writing ([purpose], [anonymous write-up], [can stop or withdraw]), then: what did you do / what followed / what could someone see / what helped / what blocked
## Barrier list
| Barrier | Where it showed (behaviour, setting) | How often it appeared | Who could remove it |
|---|---|---|---|
## Limits of the evidence
- [self-report] / [volunteers: count, not representative] / [what this cannot show] / [groups suppressed]
## Decision
[Sponsor name] decides by [date] which barriers to remove and which role owns each one.
Done When
- The follow-up has three to five items, and the volunteer route sits apart from the answers
- Every work measure is a group figure from an existing report, and the delay is your own choice
- Each barrier names who could remove it, the limits are written in, and no group under the minimum is reported
Quality Bar
- A conversation happens only with a volunteer who has consented in writing. Consent and data protection points: check with a qualified adviser.
- Nothing reaches a manager that could identify a respondent; details that would are cut, not softened
- Where the behaviour is absent, look for the condition first: the answer is usually in the work, not in the people
- Claude drafts the follow-up and sorts the answers; you hold the conversations, results are read by group, and nobody is singled out by how they perform
Next
Run ldg-estimate-the-training-roi (Estimate the Training ROI) when a money figure is asked for.
About the makers
This pack is made by 90n (90n.ai): AI-native ways of working and tools for Product, Marketing and HR teams, built by ex-McKinsey founders. We map where time and money leak, redesign the work around AI and stay until your people run it themselves, and we run our own company on Claude. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry).
1---2name: ldg-check-use-on-the-job3description: Drafts the check on use some weeks after a programme, with a short anonymous follow-up, the group-level work measures to read beside it, a consent-first conversation guide for volunteers, and a barrier list that names who could remove each barrier. Use for "run ldg-check-use-on-the-job", "training follow-up survey", "level 3 evaluation", "behaviour change after training", "did the training work", "30 day follow-up", "training transfer survey", "did anyone actually use it", part of The Complete Claude Guide for L&D Pack by 90n.4metadata:5 source: https://skillmd.com/skills/polar-bear-org/ldg-check-use-on-the-job6---78# Check Use on the Job910## When To Use11The sponsor asks "did anyone actually do anything with it?" and the forms cannot say. The last-day answers were warm, the programme ended weeks ago, and nobody has looked at the work since. Run this to answer one question: where is the new behaviour showing up at work, where is it not, and what is in the way?1213## When Not To Use14If the programme ended days ago, use has had no time to show: run Write the Feedback Form now and come back. If the budget holder wants a money figure, run Estimate the Training ROI; answers about use do not add up to a return. A manager noting obstacles week by week uses the barrier log from Brief the Managers.1516## Inputs17- The behaviours people should be using by now, and the leading indicators, from your evaluation plan if you have one18- When the programme ended, how many people took part, and how the follow-up can reach them19- The existing reports that hold the group-level work measures, and your minimum group size for reporting20If you have none of this, I start from the programme title and what it was meant to change at work, and mark the output as a first draft.2122## Approach23This is Kirkpatrick Level 3 (behaviour on the job) and the transfer tiers of LTEM, both from their originators' public pages. Results are read by group; nobody is selected by how they perform. Self-report alone is weak and work measures alone are mute, so the two are read side by side, and volunteers add the detail neither can give. The failure it prevents: one enthusiast's story reaches the sponsor as proof, while the approval step that stopped everyone else from trying never appears on anyone's list.2425## Workflow261. Ask three questions: which behaviours should be visible at work by now, how long after the programme do you want to ask, and who must never see raw answers (usually line managers)?272. Set the timing: a delay of [weeks] that you choose. Too early and use cannot show; too late and people no longer remember which course it was.283. Draft the anonymous follow-up, three to five items: where have you used [behaviour], what happened, what got in the way, what would help. Add a separate route to volunteer for a conversation, kept apart from the answers so that volunteering never unmasks an answer.294. List the group-level work measures to read beside it: the leading indicators from the evaluation plan, taken from reports that already exist. Never per person. Note what else moved those measures in the period.305. Write the conversation guide, for volunteers only, after written consent (the purpose, an anonymous write-up, the right to stop or withdraw). Same order each time: what did you do, what followed, what could someone see, what helped, what blocked. You hold these conversations. Volunteers come forward; nobody is picked as a success or a failure.316. Once the answers are pasted in, sort them into a barrier list: the barrier, how often it appeared, who could remove it. Most barriers turn out to be conditions (time, tools, a process step), and the list says so.327. Write the limits into the output: self-report, a small volunteer sample, what it cannot show. Suppress every group under the minimum.3334## Output Format35```markdown36# Use on the Job Check37Programme: [name] | Ended: [date] | Asked after: [weeks, set by you] | Responses: [count] of [group size] | Minimum group reported: [number]38## Anonymous follow-up391. Where have you used [behaviour] since the programme? [options, including "not yet"]402. What happened when you did? [open]413. What got in the way? [open]424. What would help? [open]43Volunteer route: [separate link or contact, never stored with the answers]44## Group measures to read beside it45| Leading indicator | Existing report | Before | After | What else changed in the period |46|---|---|---|---|---|47| [indicator] | [report name] | [your figure] | [your figure] | [tool, workload, process] |48## Conversations with volunteers49Consent in writing ([purpose], [anonymous write-up], [can stop or withdraw]), then: what did you do / what followed / what could someone see / what helped / what blocked50## Barrier list51| Barrier | Where it showed (behaviour, setting) | How often it appeared | Who could remove it |52|---|---|---|---|53## Limits of the evidence54- [self-report] / [volunteers: count, not representative] / [what this cannot show] / [groups suppressed]55## Decision56[Sponsor name] decides by [date] which barriers to remove and which role owns each one.57```5859## Done When60- The follow-up has three to five items, and the volunteer route sits apart from the answers61- Every work measure is a group figure from an existing report, and the delay is your own choice62- Each barrier names who could remove it, the limits are written in, and no group under the minimum is reported6364## Quality Bar65- A conversation happens only with a volunteer who has consented in writing. Consent and data protection points: check with a qualified adviser.66- Nothing reaches a manager that could identify a respondent; details that would are cut, not softened67- Where the behaviour is absent, look for the condition first: the answer is usually in the work, not in the people68- Claude drafts the follow-up and sorts the answers; you hold the conversations, results are read by group, and nobody is singled out by how they perform6970## Next71Run ldg-estimate-the-training-roi (Estimate the Training ROI) when a money figure is asked for.7273## About the makers7475This pack is made by 90n ([90n.ai](https://90n.ai)): AI-native ways of working and tools for Product, Marketing and HR teams, built by ex-McKinsey founders. We map where time and money leak, redesign the work around AI and stay until your people run it themselves, and we run our own company on Claude. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry).