How to run the EQ feedback conversation after the report

An L&D manager in Pune had forty emotional intelligence reports and one week. She booked forty conversations, twenty minutes each, back to back. By Thursday the pattern was visible from the corridor. Each manager came out saying a version of the same sentence: yes, I should work on that. Nobody wrote anything down. Nobody asked what "that" was.

The instrument was fine. The reports were fine. The conversations were the delivery mechanism, and they ran in the wrong direction: the reviewer read the data, explained it, suggested the action, and the manager agreed. Agreement is the cheapest response available to a person who has just been described.

Two jobs happen in the room

A feedback conversation has to produce two things: an interpretation the manager believes, and a commitment they will act on. Only one of those can be supplied from outside.

Give someone the interpretation and the moves left are accept or contest. Accept produces the polite sentence in the corridor. Contest produces an argument about the rater. Both end the meeting, and neither changes a Tuesday review.

Hattie and Timperley separate feedback about the task, the process, self-regulation, and the self as a person. Their judgement on the last is blunt: feedback directed at the self is the least effective of the four, because it carries almost no information about what to do and is filtered through an existing self-concept. They cite Kluger and DeNisi's meta-analysis putting praise at an effect size of 0.09, lower than no praise at all, at 0.34.

"You are low on influence" is self-level feedback. So is "you are great with people, but". Both describe a person, and nobody can test a statement about themselves the way they can test one about a meeting.

Why the manager has to speak first

Cohen and Sherman's review of self-affirmation research explains what happens when a person's sense of adequacy is threatened. Threats to self-integrity produce self-protective defences, and those defences interfere with taking in the information that did the threatening. A report saying your colleagues experience you as dismissive is a threat of that kind.

So the usual opening is backwards. Reassurance before the data ("you're one of our best managers") does not remove the threat; it makes the later criticism feel like a reversal. What lowers the defence is giving the manager something to do that is not defending. Reading their own data is that something.

Amrhein and colleagues videotaped 84 motivational interviews and coded the strength of clients' commitment language. Commitment strength in the session predicted drug use outcomes over twelve months, and the commitment around the client's own evaluation of a change plan carried most of it. The counsellor's enthusiasm was not the active ingredient.

An EQ report is not an addiction. Treat the parallel as a design hint: what predicts later behaviour is the strength of the sentence the person says about their own plan.

The order of work

Stage 1: prepare three things and no advice

Write three things on one page: the band sentence for each subscale you plan to discuss, with the norm group named; one behaviour note the manager wrote before seeing the report; one question you will open with.

Do not write a recommendation. The moment one exists on paper it gets delivered, and the manager's job collapses into agreeing with it.

Stage 2: say what this is not

Two minutes, and every sentence carries weight.

"This is a development read. It does not go into promotion, pay or a performance file. The band around each score is this wide, so small differences are noise. I have read it, you have not. I would rather hear your reading before I give mine."

That last sentence is the method: it transfers the interpretation without giving up the frame.

Stage 3: ask two questions and stop talking

"Which of these numbers matches something you already suspected about yourself?"

"Which one looks wrong, and what did the raters miss?"

Both are answerable, and neither requires the manager to defend a character. The second is more valuable: a manager who can give a specific reason a score is wrong is engaged with the data where change happens, even while disagreeing with it.

Then reflect back what you heard in their words, not yours. "So the low self-regulation score doesn't surprise you, but you think it's about the shutdown quarter, not how you normally work." An accurate summary is the cheapest evidence that you were listening rather than waiting.

Stage 4: go looking for where the report is wrong

This is the step teams skip, and it is where the finding usually sits.

Take the three behaviour notes the manager wrote before opening the report and mark each against the band: supported, contradicted or silent. Contradicted is not an instrument failure. It is usually the most useful row on the page: the report says he does not notice other people's state, the notes show he noticed and said nothing because the plant head always spoke first. That is a meeting protocol problem, and no amount of empathy training will touch it.

Ask: "If this number were right, what would we see on the floor that we are not seeing?" Then let the manager answer. They have more data than you do.

Stage 5: the manager writes the sentence

Hand over the pen. Literally, if the room allows it.

The sentence has a fixed shape: if this situation happens, then I will do this. Not "I will be more assertive". Not "I will listen better". A situation and an act, both observable by someone else in the room.

Then three additions: a date within four weeks when the situation next occurs; one named witness who will notice whether it happened; and a consequence the manager writes, usually a second attempt with a different tactic, not a penalty.

Read the sentence back. If the manager corrects your reading, that is a good sign. It means the sentence is theirs.

The sentences that end a conversation early

Most reviewers shut a conversation down without noticing.

The pattern in the right column is the same in every row. Ask about a situation, or hand back a decision. Do not describe the person.

A worked example

Numbers here are illustrative. The structure is not.

Eighteen team leaders, a scale with a standard deviation of 15 points and a reliability of 0.80. The standard error of measurement is 15 times the square root of 0.20, about 6.7 points, so a working band is roughly 13 points wide.

One leader scores 41 on influence, norm mean 50. Her band runs from about 35 to 48. Below the norm, and the score says nothing about why.

Before the report she wrote three situations. In each, she had a view that would have changed the outcome and did not say it: a machine issue unmentioned at handover, a production date she let pass in planning, a missing guard she reported afterwards instead of in the room.

Against the band: supported, supported, contradicted. The third shows she notices and acts, just not in front of the group.

Her plan: in the next plant review, in the first five minutes, I will raise the one issue I would otherwise mention afterwards, and I will name it as an issue rather than a question. Witness: the shift supervisor. Date: the review on the 24th. If it does not happen, raise it in the pre-meeting with the plant head instead.

Six months later her influence score was 46, inside her original band. The number that moved was different: issues raised by her team in plant reviews went from about two a month to nine.

The honest limit

A good conversation is a modest multiplier, not a cure. Lundahl and colleagues pooled 119 studies of motivational interviewing: average effect 0.28 against weak comparison groups, 0.09 against specific alternative treatments. Compare a well-run feedback conversation to another well-run conversation and the advantage shrinks to something you would not notice in a team of twenty.

The base is also clinical and educational: Amrhein's sample was 84 people in drug treatment, Hattie and Timperley's levels come from classrooms. Applying them to a manager reading a psychometric band is an analogy, and I would rather say so.

It cannot rescue an instrument the manager does not trust. If he thinks the questionnaire was nonsense, or that his raters had it in for him, no opening question survives the first two minutes. Diagnose that before the meeting and fix the rater pool, not the script.

It cannot substitute for performance management. If a manager is shouting at people, that is a conduct issue with a process and a timeline. Development data should not soften it or delay it.

And it costs time. Thirty minutes in the room, ten preparing, ten writing up: fifty minutes per manager. A programme that budgets twenty gets the old conversation.

One local variation. Where seniority is steep, "what do you think this score means" can land as a test with a right answer. The fix is for the reviewer to go first on something small and real: one thing they got wrong in their own last cycle. That is practice, not a finding.

What to do on Monday

Take the next five reports, not the whole cycle. For each, write the band sentence and one opening question, and no recommendation. Book thirty minutes. Run the five stages in order, and let the manager hold the pen at the end.

Then count two things: how many of the five left with a sentence they wrote themselves, and how many had a date in the diary. Below three out of five, the problem is the conversation rather than the instrument, and the first thing to change is your opening line.

If the reporting side is already handled and you want the conversation supported by the tool rather than by memory, the development report is built around bands and behaviours rather than percentiles. If you are still deciding whether the report should carry a score at all, the argument for bands and conditions over labels is worth reading first. And if the plan coming out of the conversation is "send him on a course", check whether emotional intelligence training moves anything measurable before you spend the budget.