Twenty-two items, sixty minutes, one candidate. She finishes eleven, writes a decision on each, and leaves eleven untouched. The assessor marks eleven, finds nine correct, and gives her the second-highest score in the cohort. The candidate at the next desk finishes seven: three decided, four delegated with a name and a date, one binned with a reason. He scores mid-table and the hiring manager asks why he was slow.
Nobody in that room did anything wrong. The answer key counted items handled, and item counting is what an in-tray answer key usually does. But the exercise is called a prioritisation exercise, and a mark scheme that pays per item handled measures throughput. The two are not the same, and the gap between them is where good candidates disappear.
Why the answer key drifts towards volume
The mechanics push you there. An in-tray is fifteen to thirty pieces of paper, memos and telephone slips, and the typical time limit is an hour, according to the practice material most candidates train on. Write an answer key by listing every item with a correct priority and action beside it, and the natural way to turn that list into a score is to count the lines the candidate got right. The key is a grid of item-level truths, so the score becomes a coverage percentage.
There is a second reason, less visible. Absent work leaves no trace on the script. A candidate who correctly decides not to handle an item produces a blank space, and a blank space looks identical to a candidate who ran out of time. Ask an assessor to mark what is in front of them and they mark what is written, not what was wisely left alone. The exercise then quietly rewards the person who wrote the most words, and words are cheaper than judgement.
The British Psychological Society’s guidelines for assessment and development centres are explicit about observation and about independence. Assessors work through Observe, Record, Classify and Evaluate, they complete their evaluations and report forms before the integration session, and each competency is meant to be measured more than once, in different exercises. Nothing in that guidance tells you how to score an in-tray script. That is your design decision, and if you do not make it deliberately, the item count makes it for you.
The three-way triage rule
Before the day, write one rule that every item must pass through. Three outcomes, each with a test a new assessor can apply without asking you.
| Bucket | The test | What the script must show |
|---|---|---|
| Act now | Only this role can decide it, the consequence lands inside a day, and the decision is hard to reverse | A decision, the reason, and who is told |
| Delegate | Someone else owns the outcome, but this role owns the standard | A named person, a deliverable, a date, and the limit of their authority |
| Defer or bin | No consequence this week, or the item is waiting on a fact nobody has yet | The trigger that brings it back, or the reason it dies |
Everything else is escalation, a variant of Act: you are deciding to move it up, and the script has to say what you hand over and what you keep.
The tests matter more than the labels. An hour in a quiet room, two hundred kilometres from the plant, is not a Tuesday morning, and the “only this role” test separates real urgency from the urgency an item’s author wrote into the memo. Give the candidate the facts they need to apply it: who reports to whom, the approval limits, who is on leave. Without that you are scoring their imagination.
A well-built tray holds at least three items that score full marks only if deferred or binned. If every item deserves action, the tray is a to-do list and the triage test has nothing to bite on.
What a correct delegation looks like on paper
Most scripts lose marks here, and most keys fail to award them. “Forwarded to Ramesh for action” is not a delegation, it is a hand-off, and hand-offs are how work disappears in real organisations.
A scored delegation carries four fields.
- A named person with capacity. A name, not a function. “The HR team” is nobody. If the item requires a person who is travelling, the script has to notice that.
- A deliverable, not an activity. “Prepare the revised dispatch schedule showing the two options and the cost of each” can be checked on Thursday. “Look into the dispatch issue” cannot.
- A date tied to the item’s own deadline, not a default of Friday. If the client meeting is Wednesday, the date is Tuesday.
- The authority limit. What the person may decide alone, and the threshold at which it comes back. This is the field that separates a manager from a router.
Score each field separately, one mark each, and score zero when the delegate is a peer or a superior with no reporting line to the candidate. Delegating upward is a real skill, but it is a different one, and it belongs in the escalation column where you can see it.
Change the key in four places
Score the decision, not the count. Three columns per item: bucket correct (two marks), reason cited from the paper (one mark), and handling quality (up to four marks, weighted by bucket). An item left blank is scored on whether the blank was the right answer, which the assessor can only judge against the keyed bucket.
Fix the marks available in each bucket. Decide in advance that the tray holds, say, forty marks of Act, thirty of Delegate and twenty of Defer. A candidate who acts on everything can reach at most the Act ceiling and loses the other fifty. This is the single structural change that makes triage unavoidable rather than optional.
Remove the completion bonus. No marks for volume, no penalty for untouched items, no tie-break on items handled. If you need one, use the quality of the two highest Act decisions.
Write the wrong answers down. For each item, record the plausible wrong bucket and why it is wrong. “Delegate to finance” on the safety incident is the answer most candidates give, and an assessor who has not seen it written will credit it as sensible.
Run the wash-up off the triage sheet
The assessor’s sheet should look like the candidate’s: item number, bucket assigned, bucket keyed, reason quoted, delegation fields ticked. That gives the integration session something to argue about that is not a feeling.
Two mechanics keep it honest. Each assessor completes their own sheet, including the rating, before the session opens, as the BPS guidelines require. And the discussion runs in candidate order, so a strong first impression cannot set the frame for everyone who follows.
Then apply one question per candidate: which single item, handled differently, would have changed the outcome most? If the assessors disagree, the tray has an ambiguous key, which is a design finding you can fix before the next cohort. The same discipline that keeps a group exercise from becoming an airtime test applies here: record what was done, not how it felt.
For a centre where the integration session has already turned into a negotiation over scores, fix the scoring drift first. A better answer key read aloud in an uncalibrated room just gives the loudest assessor better material.
An illustrative run
A development centre for twenty-eight first-time managers, four sittings of seven, a twenty-two item tray, sixty minutes, two assessors per sitting. Numbers are constructed for illustration.
On the old key, marks per item handled, the top three in each sitting were almost always the three who wrote the most. Two candidates who delegated accurately but handled only eight items placed in the bottom third and were told they needed to improve time management.
On the new key, with bucket ceilings and the four delegation fields, six candidates move up by more than five places. The largest single move is a candidate who handled nine items, deferred five with a trigger written against each, and delegated three with all four fields complete. Her Act decisions were right on all four of the items she took. Under the old key she was a slow writer. Under the new one she is the only candidate in the cohort who ran a full triage.
The useful output was not the reshuffled list. It was that eight of the twenty-two items had a keyed bucket that two assessors read differently, which is a tray problem rather than a candidate problem.
What this cannot do
An in-tray script is written evidence of a decision, not the decision. A candidate can write a fluent delegation to a colleague who does not exist and lose nothing, because nothing tests whether the hand-off happened. Pair the in-tray with a role-play where the delegate pushes back, or with a verbal briefing where the assessor plays the person receiving the work.
The exercise also has a known validity problem that no scoring rule fixes. Reanalysis of published assessment centre data by Lance and colleagues found that variance between exercises dominates variance between dimensions, which means a rating from one exercise tells you more about that exercise than about the person’s stable capability. An electronic in-basket study by Becker, Meiring and van der Westhuizen, with 89 participants, found support for only one of the intended dimensions, with multicollinearity between most of the others. Read plainly: the in-tray is a work sample and a decent one, not a window into a trait. Do not let it alone carry a decision on judgement.
The last limit is local. The buckets encode your approval limits and reporting lines. A tray built for a plant head in Pune will mis-score a candidate for a sales role in Gurgaon, and reusing one key across roles is how a centre measures familiarity with a business it does not run.
Monday
Take your last in-tray answer key and rewrite it with three columns beside each item: the keyed bucket, the plausible wrong bucket, and the four delegation fields if the item is a delegation. Then re-score the most recent cohort’s scripts blind against it, before looking at their original ratings. Count how many candidates move more than three places. If it is more than two or three, your centre has been buying a prioritisation shortlist from a writing race. Our assessment centre design work starts from the answer key, and if the dimensions being rated need trimming first, begin with cutting the framework down to what managers use.
Sources
https://www.assessmentday.com/in-tray-exercise.htm https://www.klug-md.de/downloads/BPS_PTC_AC_Guidelines.pdf https://www.cogn-iq.org/blog/assessment-centers-validity https://sajip.co.za/index.php/sajip/article/view/1582/2554
