What does a paid work sample predict about VA performance?

Question and scope

This review asks what a paid work sample can predict about a virtual assistant's future performance, and what it cannot. The scope is a single hiring decision for a defined task lane, not a general ranking of candidates or providers. Evidence comes from task-analysis guidance, selection-test principles, and the operational limits of a short exercise. The conclusion applies to the sample as designed, scored, and reviewed, not to a different exercise with the same name.

A useful answer separates three questions: can the candidate do the sampled task, can the candidate do it under the real constraints, and will the candidate keep doing it after the novelty fades. A paid sample can address the first two with reasonable confidence and the third only weakly. Buyers who treat one sample as a complete forecast will overread the result. The review therefore states a bounded claim and a follow-up test.

The review also distinguishes a work sample from an interview task. An unpaid exercise completed under time pressure measures willingness to perform for free as much as it measures skill. A paid sample with a realistic brief removes that distortion and gives the candidate a fair reason to invest effort. The payment is part of the method, not a courtesy.

What a sample can and cannot observe

A sample observes the candidate's choices on a defined input with a known standard. It can show whether the person reads the brief, finds the source record, follows the template, and asks a sensible question. It can also show how the person communicates a decision in writing. These are direct observations rather than inferences from an interview.

A sample cannot observe how the candidate handles a month of repetitive work, a difficult client, or a week when the tools fail. It cannot prove attendance, honesty under pressure, or long-term fit with a manager. Those qualities need references, a trial period, and a review cadence. The sample narrows the risk; it does not remove it.

The sample also reveals how the candidate handles a written brief with no one available to answer questions. That condition is common in asynchronous support, where the client is asleep and the assistant must decide whether to wait or proceed. A candidate who documents the assumption and proceeds carefully may be a better fit than one who stops entirely. Score that behavior explicitly rather than treating it as a communication style.

Designing a representative task

Start from a task analysis of the actual lane, including the inputs, tools, outputs, and decisions. O*NET's occupation profiles show how tasks, skills, and work context can be described in observable terms. Select the two or three tasks that consume the most time or carry the highest consequence. Build the sample around those tasks rather than around a generic writing prompt.

Use safe records that preserve the structure of the real work. A sample built on invented data can still test judgment if the thread, the attachment, and the conflicting instruction are realistic. Give every candidate the same brief, the same time limit, and the same reviewer. Record the pay rate and the expected duration so the exercise is fair and comparable.

Check the task against the client's actual tools before you send it. A sample that requires a project-management feature the client does not use measures the wrong skill. If the real work happens in email threads and spreadsheets, keep the sample there. The closer the fixture is to the daily interface, the more useful the prediction becomes.

Keep the fixture small enough that the reviewer can read every message before scoring. A sample with twenty threads may look realistic, but it produces fatigue and inconsistent review. Four or five well-chosen messages usually reveal the same judgment with less noise. The goal is a sharp observation, not a simulation of a full day.

Scoring dimensions and evidence

Score accuracy, completeness, source use, communication, and escalation judgment on one rubric. Accuracy covers names, dates, amounts, and the actual question. Completeness covers the requested output and the evidence attached to it. Source use asks whether the candidate opened the record rather than guessing from the message.

Communication covers tone, length, and whether the next step is clear. Escalation judgment covers whether the candidate stopped at a boundary and routed the decision with evidence. EEOC guidance on employment tests stresses job-relatedness and consistency, and the same logic applies to a paid sample. Score the evidence, not the impression left by the interview.

Record a short note for every score so a second reviewer can see why the number was given. A score without evidence is hard to defend and easy to inflate. Keep the notes in the same order as the rubric. When the scores are close, the notes are often more useful than the total.

Use a rubric that separates a wrong answer from an unanswered question. A candidate who asks for the missing approval may be more useful than one who guesses and happens to be right. Score the process as well as the output, especially for tasks where the correct answer depends on information the candidate cannot see. That distinction protects the review from rewarding luck.

Limits of a single sample

One sample is a small observation with a large standard error. The candidate may be nervous, the fixture may be easier than the real queue, or the reviewer may reward a style that the client dislikes. A single reviewer adds another source of variance. These limits argue for a clear claim, not for abandoning the sample.

The sample also measures the candidate's performance while being observed, which is not the same as independent work. A candidate who asks good questions during the sample may be more or less likely to ask them on a busy Tuesday. Treat the result as one input and state the uncertainty in the hiring note. Where the decision is close, add a second task or a short trial.

What to test after hiring

In the first two weeks, compare live output with the sample on the same dimensions. Look for the skills the sample scored highly and the ones it did not test. Record corrections by cause rather than by person. A pattern of similar corrections suggests the sample or the brief missed a requirement.

Add a short review at the end of the first month with the client and the assistant. Ask which tasks took longer than expected and which instructions caused the most questions. Use those answers to revise the sample for the next hire. The follow-up turns a one-time exercise into a feedback loop.

Keep a small prediction log with the sample score, the first-month review, and the outcome. After several hires, compare the predictions with the reviews to see which dimensions carried the most signal. The log does not need statistics to be useful; a repeated pattern is enough to change the exercise. Over time it becomes the team's own evidence about its hiring method.

Compare the sample score with the first month of live work on one page. If the sample predicted accuracy but not throughput, the exercise measured the wrong constraint. If it predicted neither, the brief or the scoring may be too generic. The comparison is more useful than a pass or fail label because it tells you what to change next.

Reporting the result honestly

Write a recommendation that names the task tested, the conditions, the score, and the main uncertainty. Separate observed evidence from interview impressions and reference comments. State what the sample did not cover, such as attendance or client-facing negotiation. A reader should be able to see why the recommendation follows from the evidence.

Keep the sample, rubric, and recommendation with the hiring record. If the hire succeeds or fails, revisit the prediction and note what the sample missed. This record improves future exercises and protects the decision from hindsight. The value of a paid sample comes from how it is designed, scored, and reviewed, not from the fact that it was paid.