Plan the Evaluation
When To Use
Leadership sees a high completion rate and moves on, and nobody agreed before the build what should change. Run this while the design is still open, so the sponsor says what counts as proof first and the course is built to produce it.
When Not To Use
This is the plan, not the instruments. For the end-of-session form itself, run Write the Feedback Form. If the programme has already run and you need evidence of use at work, run Check Use on the Job; for a money figure after the fact, run Estimate the Training ROI.
Inputs
- The agreed objectives and the on-the-job actions behind them
- The sponsor's reason for the programme, and the business measure they already watch (a report or a named indicator, at group level, with no individual data)
- The practice and checks planned so far, and who manages the learners
If you have none of this, I start from the programme title and the sponsor's one-line reason, and mark the output as a first draft.
Approach
The method is the Kirkpatrick four levels (reaction, learning, behaviour, results), planned from Level 4 down as the model's originators describe on their public page, with leading indicators and required drivers. Planning top down is the whole point: the result fixes the behaviours, the behaviours fix what the course must practise, and the form comes last. The failure this prevents: three months after launch the sponsor asks what changed, and the only numbers anyone kept are attendance and a satisfaction average, because the measure that mattered had no owner and no first reading.
Workflow
- Ask three questions: what result does the sponsor expect to move, which existing report shows it, and who owns that report?
- Level 4 first. Write the targeted result in the sponsor's own words and measure, tied to that report. Baseline and target are your figures or [placeholders]. Then pick two or three leading indicators that should move within a timeframe you set, because results lag and attention does not wait.
- Level 3. Name a few critical behaviours, each observable on the job. Per behaviour, list the required drivers in four kinds (reinforce, monitor, encourage, reward), each with an owner, usually the manager. A behaviour with no driver is a hope; I say so in the plan.
- Level 2. Map each behaviour to the course's own practice and check, plus one confidence and one commitment item. A behaviour with no practice or check is flagged to the designer now, while it costs nothing to fix.
- Level 1. Relevance to the job first, then engagement. The plan names the two questions that matter; the instrument is built in Write the Feedback Form.
- For every measure, fill in the data source, the owner, the date of the first reading and the minimum group size below which nothing is reported.
- List the outside factors (tools, workload, process changes, staffing) that could move the result either way, so a flat number is not blamed on the training or on people, and a good one is not claimed whole.
Output Format
# Evaluation Plan
Programme: [name] | Sponsor: [name, role] | Agreed before build on: [date]
## Level 4 · The result we agreed
| Result in the sponsor's words | Measure and existing report | Baseline | Target | Owner | Review date |
|---|---|---|---|---|---|
| [result] | [indicator, report] | [your figure] | [you set] | [role] | [date] |
## Leading indicators
| Early sign | Data source | Owner | First reading | Expected within |
|---|---|---|---|---|
| [short-term sign] | [source] | [role] | [date] | [timeframe you set] |
## Level 3 · Behaviours and required drivers
| Critical behaviour | Reinforce | Monitor (group level) | Encourage | Reward | Driver owner |
|---|---|---|---|---|---|
| [observable action] | [reminder, job aid] | [what is looked at] | [manager conversation] | [recognition] | [role] |
## Level 2 and Level 1 · What the course itself shows
| Behaviour | Practice and check in the course | Confidence and commitment items | Gap for the designer |
|---|---|---|---|
| [behaviour] | [check] | [items] | [none / no practice yet] |
- Level 1: relevance question [question]; engagement question [question]. Minimum group size for any reported figure: [size you set]
- Outside factors to read beside the result: [tool, workload, process change]
## Decision
[Sponsor name] agrees the Level 4 result, the leading indicators and the driver owners by [date], before the design document is written.
Done When
- The Level 4 result is in the sponsor's words and tied to a report that already exists
- Every critical behaviour has at least one required driver with an owner, and any behaviour with no practice or check in the course is flagged
- Every measure has a data source, an owner, a date and a minimum group size; baselines and targets are your figures or placeholders
Quality Bar
- Planned top down, with a few behaviours someone will really observe; a plan that starts from the form is a Level 1 plan with extra rows
- The plan never claims the training alone will move the result; the outside factors are named
- Data protection points on any measure are listed for you to check with a qualified adviser
- The sponsor agrees what counts as proof before anything is built; behaviour and learning data are read for the group, never used against a person
Next
Run ldg-write-the-design-document (Write the Design Document) so the design carries the agreed measures.
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-plan-the-evaluation3description: Plans the evaluation of a programme before it is built, with the Level 4 result in the sponsor's words, leading indicators, Level 3 behaviours with required drivers and owners, Level 2 and Level 1 checks, and a data source, owner and date per measure. Use for "run ldg-plan-the-evaluation", "kirkpatrick model", "kirkpatrick model of evaluation", "evaluation plan", "how will we know the training worked", "measure training impact", "leading indicators for training", "agree the measures before we build", part of The Complete Claude Guide for L&D Pack by 90n.4metadata:5 source: https://skillmd.com/skills/polar-bear-org/ldg-plan-the-evaluation6---78# Plan the Evaluation910## When To Use11Leadership sees a high completion rate and moves on, and nobody agreed before the build what should change. Run this while the design is still open, so the sponsor says what counts as proof first and the course is built to produce it.1213## When Not To Use14This is the plan, not the instruments. For the end-of-session form itself, run Write the Feedback Form. If the programme has already run and you need evidence of use at work, run Check Use on the Job; for a money figure after the fact, run Estimate the Training ROI.1516## Inputs17- The agreed objectives and the on-the-job actions behind them18- The sponsor's reason for the programme, and the business measure they already watch (a report or a named indicator, at group level, with no individual data)19- The practice and checks planned so far, and who manages the learners20If you have none of this, I start from the programme title and the sponsor's one-line reason, and mark the output as a first draft.2122## Approach23The method is the Kirkpatrick four levels (reaction, learning, behaviour, results), planned from Level 4 down as the model's originators describe on their public page, with leading indicators and required drivers. Planning top down is the whole point: the result fixes the behaviours, the behaviours fix what the course must practise, and the form comes last. The failure this prevents: three months after launch the sponsor asks what changed, and the only numbers anyone kept are attendance and a satisfaction average, because the measure that mattered had no owner and no first reading.2425## Workflow261. Ask three questions: what result does the sponsor expect to move, which existing report shows it, and who owns that report?272. Level 4 first. Write the targeted result in the sponsor's own words and measure, tied to that report. Baseline and target are your figures or [placeholders]. Then pick two or three leading indicators that should move within a timeframe you set, because results lag and attention does not wait.283. Level 3. Name a few critical behaviours, each observable on the job. Per behaviour, list the required drivers in four kinds (reinforce, monitor, encourage, reward), each with an owner, usually the manager. A behaviour with no driver is a hope; I say so in the plan.294. Level 2. Map each behaviour to the course's own practice and check, plus one confidence and one commitment item. A behaviour with no practice or check is flagged to the designer now, while it costs nothing to fix.305. Level 1. Relevance to the job first, then engagement. The plan names the two questions that matter; the instrument is built in Write the Feedback Form.316. For every measure, fill in the data source, the owner, the date of the first reading and the minimum group size below which nothing is reported.327. List the outside factors (tools, workload, process changes, staffing) that could move the result either way, so a flat number is not blamed on the training or on people, and a good one is not claimed whole.3334## Output Format35```markdown36# Evaluation Plan37Programme: [name] | Sponsor: [name, role] | Agreed before build on: [date]38## Level 4 · The result we agreed39| Result in the sponsor's words | Measure and existing report | Baseline | Target | Owner | Review date |40|---|---|---|---|---|---|41| [result] | [indicator, report] | [your figure] | [you set] | [role] | [date] |42## Leading indicators43| Early sign | Data source | Owner | First reading | Expected within |44|---|---|---|---|---|45| [short-term sign] | [source] | [role] | [date] | [timeframe you set] |46## Level 3 · Behaviours and required drivers47| Critical behaviour | Reinforce | Monitor (group level) | Encourage | Reward | Driver owner |48|---|---|---|---|---|---|49| [observable action] | [reminder, job aid] | [what is looked at] | [manager conversation] | [recognition] | [role] |50## Level 2 and Level 1 · What the course itself shows51| Behaviour | Practice and check in the course | Confidence and commitment items | Gap for the designer |52|---|---|---|---|53| [behaviour] | [check] | [items] | [none / no practice yet] |54- Level 1: relevance question [question]; engagement question [question]. Minimum group size for any reported figure: [size you set]55- Outside factors to read beside the result: [tool, workload, process change]56## Decision57[Sponsor name] agrees the Level 4 result, the leading indicators and the driver owners by [date], before the design document is written.58```5960## Done When61- The Level 4 result is in the sponsor's words and tied to a report that already exists62- Every critical behaviour has at least one required driver with an owner, and any behaviour with no practice or check in the course is flagged63- Every measure has a data source, an owner, a date and a minimum group size; baselines and targets are your figures or placeholders6465## Quality Bar66- Planned top down, with a few behaviours someone will really observe; a plan that starts from the form is a Level 1 plan with extra rows67- The plan never claims the training alone will move the result; the outside factors are named68- Data protection points on any measure are listed for you to check with a qualified adviser69- The sponsor agrees what counts as proof before anything is built; behaviour and learning data are read for the group, never used against a person7071## Next72Run ldg-write-the-design-document (Write the Design Document) so the design carries the agreed measures.7374## About the makers7576This 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).