Start from a real task in the queue

A work sample should come from the queue the new hire will actually handle, not from a puzzle that only tests general intelligence. Pick two or three ordinary items that a competent specialist would finish in under an hour. Remove client names, account numbers, and anything that would expose a customer. The sample should look like the work, including the tools and the messy inputs.

Choose items that have a visible finish line, such as a cleaned record, a drafted reply, or a reconciled list. Avoid tasks whose quality depends on a policy the candidate cannot know yet. If the real task needs a policy decision, split the sample so the candidate prepares the evidence and the reviewer makes the decision. That split also shows whether the candidate understands the boundary of the role.

Give the candidate the same tools and constraints

A sample completed in a different tool tells you less than one completed in the tool the team uses. Provide a sandbox account, the same templates, and the same reference documents the specialist would have on day one. State the time limit and the expected output before the candidate starts. If the tool cannot be shared, say so and adjust the sample rather than pretending the gap does not matter.

Include one constraint that forces a judgment, such as an incomplete source record or a request that falls outside the written rule. The candidate should be able to mark the item blocked and explain what is missing. Review both the finished output and the note about what could not be completed. A candidate who hides a gap is a bigger risk than one who names it clearly.

Score with a rubric written before the review

Write the rubric before you see any candidate's work, because a rubric written afterward tends to describe the person you already prefer. Use four or five dimensions such as accuracy, source discipline, clarity, judgment, and speed. Define what a strong, acceptable, and weak response looks like for each dimension. Keep the definitions short enough that two reviewers can apply them the same way.

Score the sample without the candidate's name attached when the process allows it. Ask a second reviewer to score one sample independently and compare the results. A large gap between reviewers is a signal that the rubric needs a clearer example, not that the candidate is confusing. Keep the scored samples so later performance can be compared with the original evidence.

Time-box the sample and pay for it

A work sample that takes a full day is unpaid work, and it will discourage strong candidates who have other options. Keep the sample to a stated limit, usually sixty to ninety minutes, and pay the candidate at the role's hourly rate. Tell the candidate the limit in advance and stop the exercise when the time is up. The quality of a time-boxed sample is more comparable across candidates.

If the role requires a longer exercise, break it into a paid trial shift with a clear scope and a named reviewer. Do not ask for free production work, client deliverables, or a complete project plan. The sample exists to predict performance, not to extract value from the applicant. A fair process also protects the client from work that was never reviewed or approved.

Separate craft from communication

Two candidates can produce similar outputs and still differ in how they explain their work. Score the written note that accompanies the sample, including how the candidate describes assumptions, sources, and open questions. Look for plain language and specific references rather than a polished narrative. The note is often the best evidence of how the person will communicate during a real handoff.

Ask one follow-up question about a choice the candidate made, and record the answer. A candidate who can explain the tradeoff without defensiveness is easier to coach. A candidate who cannot describe the source of a number may still be guessing in production. Keep the follow-up short and use the same question for every candidate in the same role.

Check how the candidate handles ambiguity

Real queues contain requests that are incomplete, contradictory, or outside the written rule. Build one ambiguous item into the sample and see whether the candidate asks a precise question or invents an answer. The best response usually names the missing field, the decision owner, and the effect on the deliverable. A vague request for clarification is weaker than a specific one.

Do not penalize a candidate for stopping at the boundary you defined. Penalize a candidate who guesses at a payment, access, or customer-promise decision. The sample should reward the behavior you want on the queue, including the willingness to leave an item pending. Record that behavior in the rubric so the hiring decision uses the same standard as the review.

Keep the sample out of live client work

Use synthetic or fully redacted data for the sample, even when the candidate will eventually work with real accounts. A hiring exercise is not a controlled production environment, and access cannot be audited the same way. If real data is unavoidable, limit it to a small approved extract and remove identifiers before sharing. Record who approved the use of that extract.

Do not send the candidate's output to a client or use it to fill a live queue. The sample is evidence for a hiring decision, not a deliverable. Delete or archive the sample according to the retention rule for hiring records. Tell the candidate how the sample will be stored and how long it will be kept.

Compare the sample with the first ninety days

A work sample is useful only if it predicts something the team can observe later. After ninety days, compare the original rubric scores with the new hire's quality, rework, and escalation behavior. Look for dimensions that predicted well and dimensions that did not. Do not use the comparison to judge the person retroactively; use it to improve the sample.

If the sample did not predict performance, ask whether the task was too artificial, the rubric was too vague, or the first ninety days measured something different. Change one part of the process at a time and keep the earlier scores for comparison. A hiring process that never checks its own predictions is only a preference test. Keep the comparison in the hiring file so a future reviewer can see how the rubric was tested.

If the records and decision rights are already defined but need a steady operator, workforce operations support can turn approved examples, decision lines, and review rules into a draft your manager can check. The manager still owns exceptions and final approval.

For the next step, compare this workflow with Map Philippine holidays into a client service calendar and Handle a client complaint about an outsourced specialist. For access and oversight, use the authoritative planning reference alongside your own policy.