Published September 23, 2026
Published September 23, 2026.
Productivity is one of the easiest things to measure badly. Counting keystrokes, screenshots, or idle minutes feels objective, but it measures presence rather than results, and it damages trust. A better approach measures whether the work the business needs is being completed to standard, on time, and without creating rework. That kind of measurement improves the system instead of policing the person.
This guide explains which signals are worth tracking, how to set them, and how to review them without turning check-ins into surveillance.
Start with outcomes, not activity
Activity metrics answer "was the person busy?" Outcome metrics answer "did the work get done well?" A support assistant who handles forty tickets but leaves three angry customers unresolved is not productive in any useful sense. Begin with the result the role exists to produce, then choose the smallest set of indicators that reveal it.
For most assistant roles, four dimensions are enough:
| Dimension | Question it answers | Example signal |
|---|---|---|
| Completeness | Was the work finished? | Queue cleared to agreed state |
| Quality | Was it done to standard? | Correction rate on sampled items |
| Timeliness | Did it meet the window? | Items completed within target time |
| Reliability | Was it predictable? | Commitments met without chasing |
Notice that none of these require watching the person work. They rely on the output.
Choose a baseline before a target
Targets set without a baseline are guesses. For the first two to three weeks, simply record the current numbers. Then set a modest target and review it. A baseline also protects the assistant: if the process is the constraint, the data shows it, and the fix is a better system rather than more pressure.
Baselines are especially useful for quality. If one in ten sampled items needs correction, that is a fact you can improve with a clearer brief or an example, not a reason for blame.
Use sampling instead of reviewing everything
You do not need to inspect every completed item. A light sampling plan gives you credible evidence about quality without pulling all the work back into review. Sample more heavily in the first month and reduce the rate as quality holds. If the sample shows a problem, widen the review temporarily until it is fixed, then relax again.
Make the review a conversation about the system
A good review asks three questions: what is working, what is blocked, and what would make the work easier? Keep it short and regular. The weekly review rhythm works well for most roles. When a metric drops, look first at the inputs: was the brief clear, was access available, did a dependency fail? The KPIs guide covers how to set the measures themselves.
This framing matters because it changes what the assistant does with a bad result. If the review is a blame session, the assistant hides problems. If it is a system review, the assistant surfaces them early.
Avoid the surveillance trap
Monitoring software is popular and usually counterproductive. It creates anxiety, encourages activity that looks productive rather than work that is useful, and consumes trust that is hard to rebuild. If you cannot tell whether work is happening without watching a screen, the real problem is that the work is not scoped with clear outputs and a visible queue. Fix the scope, then measure results.
There are narrow exceptions, such as regulated environments that require audit logging. Even then, use the logs for compliance, not for hour-by-hour supervision, and tell the assistant what is recorded and why.
Combine numbers with judgment
Metrics are evidence, not verdicts. A quiet week might reflect a holiday, a dependency failure, or a quieter queue. Read the numbers alongside context before drawing conclusions. Use data to ask better questions, then let the conversation and the work itself inform the decision.
Retire metrics that stop being useful
Every measure has a cost. If a metric no longer changes a decision, stop collecting it. A short list that the team actually uses is more valuable than a dashboard nobody reads. Review the metric set each quarter and remove anything that has become ritual.
Common mistakes to avoid
The first mistake is measuring activity instead of outcomes. The second is setting targets with no baseline. The third is reviewing everything instead of sampling. The fourth is using monitoring software as a substitute for clear scope. The fifth is treating a metric as the whole truth. Avoid these and the measurement becomes useful rather than corrosive.
Your next step
Write one sentence describing the outcome the role exists to produce. Then choose one signal for each of the four dimensions and record a baseline for two weeks. Hold the first review as a system conversation. If you want help setting measures that fit the role, book a free consultation.
Frequently asked questions
Should I track a virtual assistant's hours?
Track agreed windows and delivery, not idle time. If you pay hourly, you need accurate time records, but those are for payment, not for judging value.
What is the single best metric?
There is no universal one. For most roles, on-time completion with a low correction rate is a strong pair. Add a capacity or backlog signal when volume varies.
How often should I review performance?
Weekly for a short operational check and monthly or quarterly for a deeper review. Keep the regular check brief and move real decisions to the scheduled review.
Sources and editorial limits
This article follows SBA guidance on managing employees and draws on general performance-management practice described by the U.S. Office of Personnel Management for setting outcome-based expectations. It is general guidance, not a measurement guarantee.