1. The meeting where the dashboard stops working
The scene is common enough that it has become a genre in L&D writing. A leadership programme has run. It cost a serious amount of money. Someone from finance asks what it returned, and the answer that comes back is 94% completion, 4.2 out of 5 satisfaction, and a slide of quotes from participants who found it valuable.
None of those numbers is false. They are simply not answers to the question that was asked. Completion tells you the programme ran. Satisfaction tells you people did not resent it. Neither is a claim about whether anyone leads differently now, and the person asking knows that.
The scale of the gap is well documented. In DDI's Global Leadership Forecast, roughly four in five HR leaders name behaviour change as the most valuable measure of a leadership programme — and fewer than one in five feel confident they can actually track it. That 60-point gap is the whole problem in a single statistic. It is not that organisations are measuring the wrong thing on purpose. It is that they know what matters and cannot see it.
2. Three different things called “impact”
Much of the confusion in these conversations comes from a word doing three jobs at once. Kirkpatrick's levels, Phillips' ROI methodology and their descendants all make the same underlying distinction, so it is worth stating plainly rather than rehearsing the frameworks:
- Learning — did the content land? Quiz scores, knowledge checks, module completions. Easy to measure, weakly related to anything that matters.
- Capability and behaviour — do these people now make better decisions in real situations? This is what almost everyone actually means by impact, and it is the layer that is routinely missing.
- Business results — did retention, productivity or engagement move? Objective, and attributionally treacherous: a dozen other things moved during the same twelve months.
The frameworks are fine. The problem is that they tell you to “collect baseline data” and then say nothing about how to baseline the middle layer — capability — objectively and at scale. That instruction is where most evaluation plans quietly die.
3. The one thing you cannot do later
Almost everything in a measurement plan can be added late. You can bolt on a survey, pull business metrics retrospectively, run interviews after the fact, rebuild the analysis six months on. There is exactly one element that cannot be recovered once the programme has started, and it is the baseline.
After the fact, the starting point can only be reconstructed — from participants who now remember themselves as worse than they were, from managers whose recollection is coloured by knowing the programme happened, from metrics that had their own trajectory. This is why every serious guide to coaching and training ROI published in the last two years arrives at the same instruction: measure before you intervene, or accept that your number is an estimate. Finance teams have learned to tell the difference.
Baseline first. Develop. Measure again. The order is not a preference — it is the only order in which the evidence exists.
4. What to baseline with, honestly compared
Five instruments are realistically available for baselining leadership capability. Each has a failure mode, and knowing them is more useful than a vendor comparison chart where one column happens to win everything.
| Instrument | Objectivity | Repeatable? | Cost |
|---|---|---|---|
Self-report questionnaires Cheap, fast, and gameable. Measures self-perception, which shifts after a programme partly because the programme taught people the vocabulary to describe themselves differently. A pre/post gain here is genuinely ambiguous. | Low | High | Low |
360-degree feedback Rich and credible with participants. Weak as a pre/post delta instrument: rater panels turn over, rater standards drift, and relationship dynamics contaminate the signal in ways you cannot see in the output. | Medium | Low | Medium |
Assessment centres The behavioural gold standard, and effectively unrepeatable at cohort scale. Cost and scheduling mean almost nobody runs one twice on the same population purely to measure change. | High | Low | Very high |
Business KPIs Objective, and attributionally hopeless on their own. Everything else in the business moved during those twelve months too. Useful as a corroborating layer, not as the primary evidence of a development effect. | High | High | Low |
Situational judgement assessment Measures decisions rather than perceptions, scores algorithmically with no human rater, and can be repeated with a parallel form. Carries a practice-effect caveat that must be managed with form design and interval length. This is the instrument innerly.me provides. | High | High | Low–medium |
The pattern is not subtle. The instruments with the strongest objectivity are the ones you cannot afford to run twice, and the ones you can repeat cheaply are the ones a motivated participant can move at will. A situational judgement assessment sits in the gap: objective because it scores decisions algorithmically rather than asking people to rate themselves, and repeatable because delivering a parallel form to a cohort costs a fraction of an assessment centre.
Its caveat is practice effects, and it should be stated rather than buried: we cover the SJT methodology and its limits in detail here.
5. The baseline → develop → re-measure playbook
Decide what should move — in writing
Name the competencies the programme is meant to change and what a meaningful movement would look like. Evaluations fail from vague intent far more often than from bad instruments. If nobody can say in advance what success looks like, no measurement will produce it afterwards.
Baseline the cohort
Before the first session. Objective instrument, whole cohort, results held as the reference point. If you are running a comparison group, baseline them in the same window with the same instrument.
Run the programme, and record the dosage
Track who actually received what. Half of every cohort attends half of everything, and a delta read against assumed exposure rather than actual exposure will mislead you about what worked.
Re-measure with a parallel form
Same instrument, same population, different scenarios. Report the delta per competency — and report the competencies that did not move, prominently. An evaluation that only ever finds success is not an evaluation, and experienced executives read it as marketing.
Present the comparison, own the limits
One slide: where the cohort started, where it ended, on the same instrument, over a stated window. Then, out loud, the attribution caveat. Owning the limits is what makes the rest of it credible — the alternative is a number that dies in the first round of questions.
6. Four ways this goes wrong
Measuring at 30 days
You will get a lovely number and it will mean nothing. At one month, participants are still enthusiastic and still remember the content. Behaviour has not had time to embed or to fail. Wait.
Changing the instrument mid-programme
A 360 at the start and a self-report survey at the end is not a pre/post design; it is two unrelated measurements with a hopeful arrow drawn between them.
Reverse-engineering the ROI figure
Starting from the number leadership wants to hear and working backwards. It is more common than anyone admits, it is obvious to anyone numerate, and it poisons the credibility of every future L&D evaluation in that organisation.
Reporting only what improved
If three competencies moved and one did not, the one that did not is the most interesting finding you have — it tells you what the programme cannot do. Hiding it is how measurement becomes marketing.
7. Where innerly.me fits
innerly.me is the capability layer of this chain, and only that layer. Participants work through realistic workplace scenarios; a deterministic algorithm scores their decisions against 17 behavioural indicators across four competencies — Adaptability, Effective Communication, Execution Ability, and Systemic Thinking. No human rater is involved, so the same person assessed twice is compared against the same standard both times.
That gives you the pre/post comparison a 360 struggles to produce and an assessment centre cannot afford to. It does not give you a revenue figure, and we will not invent one for you. What it gives you is the link in the evidence chain that is currently missing entirely for most organisations — which happens to be the link every framework tells you to establish and none of them tells you how to build.
Frequently asked questions
How soon after training should you re-measure?
Not at 30 days. Measuring immediately after a programme ends captures enthusiasm and recall rather than embedded behaviour, which is why so many L&D evaluations produce flattering numbers that nobody in the business believes. The practical window for leadership behaviour change is 6 to 12 months after the intervention. Coaching engagements can support an interim read at around 90 days because the cadence is shorter, but the 90-day figure should be reported as an early indicator, not as the result.
Can you measure leadership development without a control group?
Yes, but you should be honest about what you lose. Without a comparison group you cannot separate the effect of the programme from everything else that happened to those people over the same 12 months — a reorganisation, a new manager, ordinary experience. A pre/post design on a single cohort demonstrates that capability changed; it does not prove the programme caused the change. In most organisations that is an acceptable trade, provided the claim is worded accordingly. If the programme is expensive enough to be politically contested, invest in a comparison cohort: an equivalent group, measured at the same two points, who receive the development later.
Are 360-degree reviews enough to show behaviour change?
360s are valuable, but they are a difficult pre/post instrument. Three problems compound across cycles: rater panels change, so you are partly measuring different people's opinions; raters recalibrate their own standards over time, so a stable score can hide real change and vice versa; and relationships colour ratings in ways that are invisible in the data. A 360 tells you how a leader is experienced by those around them, which is genuinely important. It is a weaker basis than a standardised, algorithmically scored instrument for the specific claim "this capability moved by this much."
What is the difference between measuring learning and measuring capability?
Measuring learning asks whether the content was absorbed — quiz scores, module completions, knowledge checks. Measuring capability asks whether the person now makes better decisions in realistic situations. The gap between the two is where most leadership development quietly fails: people can pass every knowledge check on delegation and still not delegate. Capability measurement puts the participant in a situation and observes the choice, which is why situational judgement assessments and simulations sit closer to behaviour than any questionnaire does.
How do situational judgement tests avoid test-retest inflation?
They reduce it rather than eliminate it, and any vendor claiming otherwise is overselling. Three mechanisms do the work. First, an SJT has no memorisable answer key — scenarios are judgement dilemmas and the scoring weights are not disclosed to participants, so there is little to rehearse. Second, the re-measure should use a parallel form: the same competency structure and difficulty calibration with different scenarios. Third, a 6–12 month interval places the second sitting well beyond the window in which recall meaningfully inflates scores. For high-stakes evaluations, a comparison cohort remains the only way to quantify what practice effect is left.
What should we measure if we cannot connect training to revenue?
Measure capability, and say plainly that capability is what you measured. The mistake most L&D teams make under pressure is reaching for a revenue number they cannot defend, which then collapses under the first serious question and damages the credibility of everything else in the deck. A defensible chain has three links: capability moved (which you can measure directly), behaviour changed on the job (observable via manager and 360 input), and business metrics moved (which you can report alongside, with an honest attribution caveat). Owning the limits of link three is what makes links one and two believable.
Sources
- DDI — Global Leadership Forecast, on measuring the results of leadership development (behaviour-change vs tracking-confidence gap).
- McKinsey — on the share of executives who strongly agree leadership development achieves its intended outcomes.
- Harvard Business Impact — guidance on defining measurable behaviour outcomes and establishing measurement before and after a programme.
- International Coaching Federation (ICF) — pragmatic approach to measuring the ROI of coaching, including baseline establishment.
- US Office of Personnel Management (OPM) — on the validity of situational judgement tests as a selection and assessment method.