A soft skills assessment grid lists the skills to observe, one observable criterion for each, and levels described in terms of behaviour. It is also called an interpersonal skills evaluation grid. It replaces a general impression (“he communicates well”) with facts that two assessors would record in the same way.
Below you will find two complete templates to copy. The first is generic, covering six skills. The second applies to a conversation with an unhappy customer. After that come the method for building your own, the biases to avoid and how to use the grid in role play.
Skills grid: why observable criteria rather than impressions
“Good people skills”, “comfortable speaking”, “doesn't listen”. These judgements do not say what the person actually did. They vary from one assessor to the next and give the learner nothing to work on.
An observable criterion describes an action or a phrase that can be seen or heard. “Rephrases the request before answering” can be checked. “Is empathetic” cannot. The test is simple: two people watching the same scene should tick the same box.
This requirement applies to all soft skills, whose definition and full list you will find here. A skills grid is only worth something if each line points to a specific behaviour.
Soft skills assessment grid: the generic 4-level template
This template suits most workplace conversations. The scale follows the same logic throughout. At level 1, the behaviour is absent. At level 2, it is inconsistent. At level 3, it is mastered. At level 4, it holds up under pressure.
- Active listening. Observable criterion: lets the other person finish and builds on what they said.
- Level 1: interrupts, answers beside the point.
- Level 2: lets them speak, but prepares a reply and misses part of what was said.
- Level 3: lets them finish, asks open questions, picks up the other person's words.
- Level 4: also picks up on what is left unsaid (hesitation, change of tone) and checks it with a question.
- Rephrasing. Observable criterion: restates the request in their own words and has it confirmed.
- Level 1: does not rephrase.
- Level 2: repeats the other person's words without checking they have understood.
- Level 3: summarises the facts and asks for confirmation (“If I've understood correctly, ... is that right?”).
- Level 4: rephrases the facts and the feelings, then corrects the summary if the other person qualifies it.
- Managing emotions. Observable criterion: keeps a steady tone and pace when tension rises.
- Level 1: raises their voice, justifies themselves or shuts down.
- Level 2: stays polite, but speeds up and gives curt answers.
- Level 3: keeps a calm tone and names the other person's emotion without judging it.
- Level 4: stays steady in the face of a personal attack and brings the conversation back to the subject.
- Assertiveness. Observable criterion: states a position or a refusal clearly and respectfully.
- Level 1: gives in or imposes, without explaining.
- Level 2: states their position in a roundabout way or apologises several times.
- Level 3: says what they can and cannot do, and why.
- Level 4: holds their position after several attempts to push back and offers an alternative.
- Clarity of message. Observable criterion: puts forward one idea at a time, in words the other person understands.
- Level 1: confused message, jargon, several subjects mixed together.
- Level 2: understandable message, but long and with no apparent order.
- Level 3: structured message, short sentences, an example to back it up.
- Level 4: adapts their vocabulary to the other person and checks understanding.
- Handling objections. Observable criterion: welcomes the objection, asks for details, then responds.
- Level 1: ignores the objection or contradicts it immediately.
- Level 2: responds with a prepared argument, without asking for details.
- Level 3: asks for details about the objection, then responds to the point raised.
- Level 4: deals with the objection, checks it has been resolved and moves on to the next step.
To score, record the highest level for which all the behaviours were observed. Add an “evidence” column: the phrase heard or the action seen that justifies the level.
Skills grid, applied example: a conversation with an unhappy customer
An applied grid follows the flow of the situation. Each skill matches a moment in the conversation, which makes live observation easier.
- Receiving the complaint. Observable criterion: lets the customer explain their problem without interrupting.
- Level 1: interrupts or disputes the facts within the first few seconds.
- Level 2: listens, but starts justifying before the customer has finished.
- Level 3: lets them finish and acknowledges the inconvenience (“I understand this delay has put you in a difficult position”).
- Level 4: lets them finish, acknowledges the inconvenience and asks about the consequences for the customer.
- Staying calm. Observable criterion: tone and pace stay steady in the face of criticism.
- Level 1: answers in the same tone as the customer.
- Level 2: stays courteous, but hides behind procedure.
- Level 3: keeps a calm tone and does not take the criticism personally.
- Level 4: stays calm in the face of a threat (“I'm going to cancel my contract”) and refocuses on resolving the issue.
- Rephrasing the problem. Observable criterion: summarises the situation and gets the customer's agreement.
- Level 1: jumps to the solution without summarising.
- Level 2: gives an incomplete summary, the customer has to repeat themselves.
- Level 3: summarises the facts and gets a “yes” from the customer.
- Level 4: summarises the facts, the impact and the customer's main expectation.
- Proposing a solution and closing. Observable criterion: states an action, a timeframe and who is responsible.
- Level 1: no proposal, or a promise that cannot be kept.
- Level 2: proposes a solution, with no timeframe or next step.
- Level 3: proposes a solution with a timeframe and checks it suits the customer.
- Level 4: offers two options, lets the customer choose and recaps the commitments.
The same logic applies to a corrective conversation with an employee. The lines become: state the facts, listen to the other person's version, set out the expectation, agree a commitment and a follow-up.
Building your own interpersonal skills evaluation grid in 5 steps
- Start from a real situation. Choose a conversation your teams genuinely have: a complaint, difficult news, a negotiation. The grid serves a specific training objective.
- Keep 4 to 6 skills. Beyond that, the observer can no longer keep up. Keep the ones that change the outcome of the conversation.
- Write one observable criterion per skill. Start with an action verb: rephrases, announces, proposes, checks. Remove the adjectives.
- Describe 4 levels in terms of behaviour. An even number avoids the safe middle score. Each level describes what the person does, not what they lack.
- Test with two assessors. Observe the same scene separately, then compare. Any difference points to wording that needs rewriting.
Assessing interpersonal skills: who assesses, when, and with which biases
Who assesses
- The learner (self-assessment). Useful for raising awareness. Compare it with an outside view, as people often judge themselves poorly.
- A peer. They know the job and learn by observing. Give them two lines to follow, not the whole grid.
- The trainer. They ensure consistency across learners and link the observation to the progress plan.
- A tool. A recording or an automated analysis lets you review the scene and find the exact moment.
Combining at least two viewpoints makes the score more reliable. The gap between self-assessment and observation is often the best starting point for the debrief.
When
- Before training: an initial baseline, on a first role play.
- During: after each practice session, on just two or three lines, to guide progress.
- After: at the end of the programme, then a few weeks later, with the same grid and a comparable scenario.
Keeping the same grid from start to finish makes it possible to run a before-and-after assessment that proves progress.
Common assessment biases
- Halo effect: one striking quality (ease, for example) pulls all the scores up.
- Leniency or severity: the assessor scores everyone high, or everyone low.
- Central tendency: scores cluster in the middle to avoid making a call.
- Recency effect: the end of the conversation weighs more than the beginning.
- Contrast effect: the learner is compared with the previous one rather than with the grid.
Three safeguards limit these biases: note the evidence before the level, assess one line at a time, compare your scores with a second assessor.
Assessing soft skills during a role play and its debrief
The grid comes into its own in role play. A questionnaire measures what the learner knows. A role play built on a soft skills scenario shows what they do when the other person pushes back.
- Before: show the grid to the learner. They need to know what they will be observed on.
- During: the observer notes phrases and actions, with the moment they occur. They do not assign a level yet.
- After: the learner assesses themselves first. The trainer then shares their observations, line by line, with evidence.
Finish with one or two lines to work on, then replay the scene. Reflexes take hold through repetition, not by reading the grid.
With Face Up, the learner practises by voice and video with an AI avatar that plays the difficult counterpart. The debrief shows which objectives were met or not, strengths, areas for improvement and time-stamped key moments.
Keeping evidence of learning
In French vocational training, the national quality framework behind the Qualiopi certification (the quality label French training providers need to access public and mutualised funding) requires providers to assess whether learners have met the objectives. A dated grid, completed before and after, with the observed evidence, meets this requirement for skills that are hard to test in writing.
Keep the blank grid, the completed grids and the progress summary. For more detail, see the guide to soft skills assessment and Qualiopi, and these methods for evaluating and measuring soft skills.
Frequently asked questions
How many levels should a soft skills assessment grid have?
Four levels work for most purposes. There are enough to show progress and few enough to stay distinct. An even number also forces the assessor to make a call.
Should you give a numerical score?
The level is enough to drive progress. A numerical score is mainly useful for comparing before and after. Either way, the observed evidence matters more than the number.
Can the same grid be used for every job?
The generic grid serves as a common base. The criteria should then be rewritten using the words and situations of the job. A salesperson and a nurse manager do not face the same objections.
How long does it take to fill in the grid?
With 4 to 6 lines, a few minutes after the conversation are enough, as long as the evidence was noted during the scene. A grid that takes too long to fill in ends up not being used.
From grid to practice
Copy one of the two templates, adapt the criteria to a real situation your teams face and test it on a first role play. You will quickly see which lines need rewriting.
To see how a grid translates into a scenario and a debrief with an AI avatar, you can book a Face Up demo.
