People Management
Did the Training Change Anyone's Work? Evaluate More Than Attendance and Satisfaction
Separate attendance, learner reactions, assessment performance and application at work. A filled internal-training evaluation design and two alternative fictional reports show how opportunity, denominators and changed work conditions affect the conclusion you can responsibly draw.
Evaluate training by separating participation, learner reaction, demonstrated learning and application at work. Choose a concrete work behavior, establish what you can compare before training and follow up after people have had a real opportunity to use the skill. Attendance and a positive survey can be useful records, but neither answers whether anyone's work changed.
This LandOffer.ai guide provides a filled evaluation design and two alternative fictional result reports. The data, learners and workplace records are synthetic. Actual local calculations check the example's counts and denominators only. No course was delivered, employee work inspected or training effect measured.

Decide what decision the evaluation must support
Begin with the action a training owner needs to take: revise a practice task, improve follow-up support, expand a pilot or stop assuming training will solve a workflow problem. If the aim is simply to declare success, it is easy to choose the most flattering number after seeing the results.
CDC's evaluation-planning guidance recommends defining the purpose, evaluation questions and feasible data-collection methods early. It distinguishes questions about delivery from questions about learning and its use afterward. These principles come from public-health training guidance; adapting them to a software team's internal learning is an editorial application, not validation of this article's example. Building an Evaluation Plan.
Write the expected behavior in terms someone can inspect. “Understand feature-flag retirement” is too broad. “Prepare a review note identifying an owner, current usage evidence, unresolved uncertainty and the next authorized step” gives an evaluator something more precise to look for. It does not authorize anyone to retire a real flag.
Keep the evaluation purpose distinct from individual performance management. Explain who can see records, what they will be used for and how long they are retained under the organization's actual policy. Do not surprise people by turning a development exercise into a ranking for pay or promotion.
Separate evidence that answers different questions
A completion record answers who attended or finished the assigned material. A reaction survey describes what respondents thought about it. A scenario exercise can show performance under assessment conditions. A later work sample can show whether the expected behavior appeared in a particular work context.
CDC distinguishes learning from learning transfer and recommends assessing both when possible. Its guidance explains that a posttest alone cannot establish change from the learner's starting point, and that delayed follow-up should reflect the chance to apply the skill. It also warns against treating satisfaction as evidence of training effectiveness. Measuring Effectiveness.
Before-and-after results still need interpretation. A repeated scenario may reward memory of its answer; different scenarios may differ in difficulty. A changed template or additional coaching can affect later work. A small internal evaluation can identify a useful next improvement even when it cannot isolate training as the sole cause.
Include an opportunity-to-apply check. Someone who has not encountered the relevant task is different from someone whose reviewed work omitted a required element. Counting both as failures turns scheduling and task allocation into an apparent learning deficit.
Worked asset one: a filled internal-training evaluation design
The fictional Cedar team plans a short session on writing feature-flag retirement review notes. Training owner Hester wants to decide whether to repeat the session unchanged or revise its practice and workplace support. Process owner Pascale approves the review-note criteria. Neither the course nor this evaluation grants production-change authority.
The eight synthetic participants are P1 through P8. Their identifiers are example labels, not anonymized real employees. Everyone completes the same defined assessment process; the later work review includes only people with an eligible work opportunity during the follow-up window.
The agreed design before delivery
| Evaluation element | Completed design decision |
|---|---|
| Work behavior | Review note states the responsible owner, current usage evidence, an unresolved question or explicit absence of one, and the next authorized step |
| Assessment rule | All four elements must be supported by the supplied case; a confident invented answer does not count |
| Before training | Each participant writes a note from scenario A; Hester scores against the four-element rule |
| Immediately afterward | Each writes a note from scenario B, designed for the same criteria but with different details; Pascale checks that the cases have comparable demands |
| Reaction | Optional short survey about relevance and delivery; report response denominator separately |
| Workplace baseline | For participants later having eligible work, review their most recent eligible pretraining note using the same criteria |
| Follow-up | After four weeks, identify who actually had an eligible task and review the first eligible note per person; record missing access or no opportunity separately |
| Review method | Hester scores; Pascale independently checks the initial four assessments and resolves criterion disagreements before the remaining review |
| Decision use | Revise the session or work support using observed gaps; do not use this pilot as an individual rating or proof of financial return |
These dates and methods are illustrative design choices, not a universal four-week rule. A team handling the task monthly may need longer; a team performing it daily may learn about barriers sooner. Choose a window that fits the task rather than one that conveniently produces a report before a budget meeting.
The review is bounded. Sixteen pre/post notes at four minutes each require 64 minutes of initial scoring. Five baseline and five follow-up work notes at six minutes each require another 60 minutes. The 124-minute total excludes Pascale's independent checks, reconciliation and any follow-up conversations. Those activities need separate capacity; the calculation is not an estimate of all evaluation effort.
What a scored response actually means
The authored posttraining scenario describes flag F-62, an identified service owner, a usage excerpt whose time range is missing and an unapproved retirement request. A passing note names the owner, identifies the excerpt and its missing time range, states that current usage cannot yet be concluded and asks the authorized reviewer to obtain bounded evidence before deciding the next step.
A note saying “No usage, safe to remove” fails the evidence and uncertainty criteria even if it is neatly written. A note identifying uncertainty can meet the rule without answering the underlying operational question. The assessment tests the specified review behavior, not whether a participant can produce a confident retirement decision.
CDC's quality standards connect assessments to learning objectives and include follow-up support. Its standards are a benchmark for CDC-developed or funded training; this article does not claim the fictional session meets all eight or qualifies for a CDC designation. Quality Training Standards.
Worked asset two: two honest reports from different possible outcomes
The next reports are mutually exclusive alternative continuations of the same design. They are not two real trials, pooled cohorts or evidence that a particular intervention works. Both use the same synthetic attendance, reaction and assessment counts so the workplace difference cannot be hidden behind those numbers.
All eight participants attend. Seven answer the optional reaction survey; six of those seven describe it positively. The result is six positive responses among seven respondents, with one nonresponse among eight attendees. It is not eight satisfied learners.
Four of eight meet all assessment criteria before training, and seven of eight meet them afterward. The same eight people have paired assessments, so the pre/post totals do not mix different respondents. This shows an improved result under the authored assessment conditions; it does not isolate training from practice, case differences or scoring effects.
Five people, P1–P5, have an eligible workplace task during follow-up. P6–P8 do not. Among the five paired work records, two pretraining notes meet all criteria. The alternatives diverge in the later work:
| Participant | Preassessment | Postassessment | Baseline work note | Follow-up A | Follow-up B |
|---|---|---|---|---|---|
| P1 | Meets | Meets | Meets | Meets | Meets |
| P2 | Does not meet | Meets | Does not meet | Meets | Does not meet |
| P3 | Meets | Meets | Meets | Meets | Meets |
| P4 | Does not meet | Meets | Does not meet | Does not meet | Does not meet |
| P5 | Meets | Meets | Does not meet | Meets | Does not meet |
| P6 | Meets | Meets | No paired work opportunity | Not assessed | Not assessed |
| P7 | Does not meet | Meets | No paired work opportunity | Not assessed | Not assessed |
| P8 | Does not meet | Does not meet | No paired work opportunity | Not assessed | Not assessed |
Report A: stronger reviewed work, limited attribution
Four of five eligible participants' reviewed follow-up notes meet the four-element rule, compared with two of the same five baseline notes. P4's later note still omits the evidence time range. The three participants without an eligible task are not classified as workplace failures.
This is consistent with application of the taught review behavior in the sampled work. However, a revised note template and routine peer coaching were also introduced during follow-up. With five paired work samples, no comparison group and one note per person, we cannot attribute the difference solely to training or claim it represents every future note.
Decision: retain the practice task for a further bounded pilot, clarify the evidence-time-range example and inspect whether the new template supports the same criteria. Do not announce an organization-wide error reduction or return on investment.
Four of five is 80%; two of five is 40%. The difference is 40 percentage points in this small authored set, not a 40% causal training effect. Lead with the five-person denominator and counts. This example has no calibrated uncertainty interval or statistical test and supports no population effect estimate.
Report B: assessment gains without improved sampled work
Two of five eligible participants' follow-up notes meet all criteria, the same count as their baseline work. P2, P4 and P5 omit a current evidence time range. The pre/post exercise result remains four of eight versus seven of eight; it does not establish transfer to work.
In this alternative continuation, the workplace template has no place to record the evidence period, and reviewers have continued accepting notes without it. Interviews in the authored case identify that mismatch as a candidate barrier, not proof that it explains every omission. Three participants still had no eligible task and remain unassessed for transfer.
Decision: Pascale owns a template-and-review-guidance revision before another follow-up. Hester will revise the practice example only if its demands differ from the intended work. Do not repeat the same session for everyone merely because the workplace count did not improve.
This second report does not establish that training was useless or that the learners failed. It identifies a gap between assessment performance and the sampled work, then assigns an action to the person who controls part of the environment. Further review is needed to learn whether the proposed support change helps.

Check missing evidence before making a program decision
Keep response and opportunity denominators visible. If only the most confident learners volunteer a work sample, their results may not describe the whole audience. If a manager chooses only the best later note, a comparison with an ordinary baseline note is uneven. Use a stated selection rule and report deviations. Here, the first eligible note per person is the rule; replacing it with a later corrected note would require a separate measure.
Record changed conditions alongside outcomes: new instructions, different tasks, easier cases, extra coaching or access restrictions. A favorable result can justify continued investigation without supporting a claim that training caused the business outcome. A disappointing result can reveal a work-system barrier rather than a need for more slides.
Protect the people whose work is reviewed. Use approved access, collect only what the evaluation needs and share program findings at an appropriate level. A small team can be identifiable even with names removed. Follow the organization's retention and review rules rather than inventing confidentiality guarantees.
Match the conclusion to the evidence you have
An evaluation can finish with a useful limited conclusion: the exercise improved, application remains uncertain, a particular work sample changed or a support barrier needs attention. A report that names those limits is more actionable than a single satisfaction score labeled “impact.”
Use the LandOffer.ai workplace guides alongside your actual learning and data-governance processes. Before expanding a training, ask what behavior was observed, who had a chance to apply it and what else changed. Then choose the next improvement and its owner. Preserve missing opportunity and alternative explanations in the decision record rather than dropping them from the summary.
Sources and Further Reading
- CDC: Evaluate Training, Measuring Effectiveness. Dated October 28, 2024. Official training-development guidance; checked October 27, 2026.
- CDC: Evaluate Training, Building an Evaluation Plan. Dated May 21, 2026. Official public-health guidance; checked October 27, 2026.
- CDC: Quality Training Standards. Published May 15, 2024; updated and reviewed October 1, 2026. Standards' scope is CDC training; checked October 27, 2026.


