Evaluating internal training: five indicators to track

1. Rate of application in the workplace

After an in­ter­val ap­pro­pri­ate to the role, gather sev­er­al forms of evid­ence—ob­ser­va­tion, a de­liv­er­able or man­ager feed­back—to ex­am­ine wheth­er the skill is being used. A single sub­ject­ive ques­tion cannot es­tab­lish a causal effect; com­pare co­horts cau­tiously.

2. Volume and nature of questions asked during training

Re­cur­ring ques­tions may in­dic­ate some­thing that needs cla­ri­fic­a­tion, but do not by them­selves prove that a module is poorly cal­ib­rated. In­ter­pret them along­side learn­ing out­comes and learner feed­back. The avail­ab­il­ity of this data de­pends on the tools and their set­tings: see the solu­tion for businesses and the features and data to confirm during a demonstration.

3. Time between training and the first related workplace task

For train­ing linked to an iden­ti­fi­able task, you can track the time until the first prac­tic­al ap­plic­a­tion. This also de­pends on op­por­tun­it­ies to prac­tise, the role and the con­text; a short­er in­ter­val does not by itself prove that the train­ing is better.

4. Full cost per learner who performs the expected task

Not just the cost of the train­ing itself: the total cost, in­clud­ing time spent, sup­port and any rework. This is the in­dic­at­or that speaks to fin­ance teams.

5. Content reuse rate

The reuse rate can shed light on how con­tent is used, but does not alone meas­ure its use­ful­ness or wheth­er its cost has been re­covered. Define a rel­ev­ant period and con­sider this in­dic­at­or along­side pro­duc­tion and main­ten­ance costs and the in­ten­ded out­comes.

Start with the decision to inform, not the dashboard

An in­dic­at­or is only useful if it can change a de­cision. Before col­lect­ing data, ask wheth­er the com­mis­sion­ing or­gan­isa­tion wants to retain a learn­ing path, im­prove an activ­ity or com­pare two de­liv­ery ap­proaches. Mon­it­or­ing in­ten­ded to im­prove an in­struc­tion does not re­quire the same setup as an in­vest­ment re­quest. Keep meas­ures that inform that de­cision and set aside those that trig­ger no iden­ti­fi­able action.

Write a meas­ure­ment agree­ment with the op­er­a­tion­al lead: the pop­u­la­tion con­cerned, ex­pec­ted task, ob­ser­va­tion period and con­di­tions needed for ap­plic­a­tion. Spe­cify who sup­plies the data and who can dis­cuss its qual­ity. If a pro­ced­ure changes during the period, record it; without that record, the dash­board may at­trib­ute an effect to train­ing that ac­tu­ally stems from a change in the work.

Pre­pare a defin­i­tion sheet for each se­lec­ted in­dic­at­or. It should spe­cify the nu­mer­at­or, de­nom­in­at­or, ex­clu­sions and treat­ment of un­known cases. “Learners who ap­plied the skill” means noth­ing unless you spe­cify which pop­u­la­tion they belong to and what op­por­tun­ity they had to act. Someone with no op­por­tun­ity to prac­tise must not si­lently be treated as someone who failed. Report these situ­ations sep­ar­ately so that the de­cision re­mains in­ter­pretable.

Build comparable costs and a defensible monetary valuation

Define the scope of costs before com­par­ing op­tions. Dis­tin­guish ini­tial design, de­liv­ery, par­ti­cipants’ time, sup­port and main­ten­ance. Re­using a re­source may reduce some pro­duc­tion ex­pendit­ure while still re­quir­ing cor­rec­tions. Spread­ing ini­tial costs across a hy­po­thet­ic­al future audi­ence would make a scen­ario look ar­ti­fi­cially at­tract­ive when its rol­lout has not been ap­proved.

Fin­an­cial ROI can be ex­pressed as at­trib­ut­able mon­et­ised be­ne­fits minus the in­cluded costs, di­vided by those costs. This for­mula does not solve the at­tri­bu­tion prob­lem. Doc­u­ment the origin of each be­ne­fit, the period and the as­sump­tions. A qual­ity im­prove­ment can matter without being hon­estly con­vert­ible into euros; in that case, report it along­side the cal­cu­la­tion rather than in­vent­ing a value.

Also dis­tin­guish freed-up ca­pa­city from budget sav­ings. Avoided rework may in­crease a team’s avail­ab­il­ity without re­du­cing an in­voice or payroll ex­pendit­ure. If you assign a value to that time, ex­plain the con­ven­tion used and what the time might be used for, without auto­mat­ic­ally present­ing it as cash saved. The fin­an­cial de­liv­er­able should in­clude a cau­tious scen­ario and the as­sump­tions that would change the de­cision, not an isol­ated figure presen­ted as cer­tain.

Treat missing data as information

De­scribe the col­lec­tion pro­cess before launch: when ob­ser­va­tion takes place, the in­struc­tions given to as­sessors and where the evid­ence is stored. Check that as­sessors in­ter­pret the suc­cess cri­ter­ia con­sist­ently when ap­ply­ing them to the same ex­ample. Oth­er­wise, a dif­fer­ence between teams may re­flect scor­ing rather than com­pet­ence. Retain meas­ure­ment caveats in the report for the com­mis­sion­ing or­gan­isa­tion.

Com­pare the pro­files of ob­served and un­ob­served par­ti­cipants where this in­form­a­tion is avail­able and ap­pro­pri­ate. Vo­lun­teers or em­ploy­ees with more avail­ab­il­ity may be overrep­res­en­ted. Do not re­place miss­ing data with an im­pli­cit suc­cess or gen­er­al­ise to absent par­ti­cipants without jus­ti­fic­a­tion. In­com­plete col­lec­tion may jus­ti­fy post­pon­ing a de­cision or making a lim­ited cor­rec­tion rather than pro­ceed­ing with a full rol­lout.

Application: making a decision despite uncertain attribution

The de­cision pack con­tains the defin­i­tion of rework, a cost record, avail­able ob­ser­va­tions and com­pet­ing ex­plan­a­tions. It may jus­ti­fy cor­rect­ing part of the learn­ing path or con­tinu­ing ob­ser­va­tion; it does not prove pos­it­ive ROI. If a com­par­is­on is sub­sequently or­gan­ised, the team doc­u­ments dif­fer­ences between pop­u­la­tions and work­ing con­di­tions. This ex­ample il­lus­trates a de­cision-making ap­proach without in­ven­ted amounts or claimed causal ef­fects.

Key takeaway

The avail­ab­il­ity of in­dic­at­ors de­pends on tools, col­lec­tion prac­tices and data pro­tec­tion rules. A dash­board is no sub­sti­tute for an eval­u­ation method; define the ob­ject­ives, period and limits of in­ter­pret­a­tion. See also consistent training across franchise networks and pricing.