Compare Research Output by Task Difficulty

A researcher who resolves five disputed company identities may contribute more useful work than someone who fills twenty straightforward company records. A raw completion leaderboard cannot tell you that. It combines assignment decisions, evidence availability, task scope, and execution into one number, then gives the number a person's name.

Before comparing people, determine whether they were doing comparable work. The worksheet below is intended to improve assignments and support, not to make employment decisions from a small operational sample.

Classify the task before judging its outcome

Use observable categories rather than invented difficulty points. For a company research queue, start with three types: confirmation from a clear authoritative source, interpretation across several consistent sources, and conflict resolution where identity or responsibility remains disputed. Keep the intended deliverable constant within each type.

Record the category at intake and allow a documented revision when an unexpected obstacle appears. A researcher should not need to fail before a task can be recognized as complex. Equally, a long duration alone should not automatically upgrade its category: the time might reflect waiting, interrupted work, or an unclear request.

The statistical idea behind this separation is familiar. NIST Engineering Statistics Handbook: Randomized block designs explains how other factors can affect a measured result and how comparable blocks can help control that variation in designed experiments. The operational worksheet here is not a randomized experiment or a validated employee assessment. Grouping tasks does not remove every difference between them.

A fictional comparison that changes the question

Imagine two researchers, Lina and Omar, working during the same reporting week. Every figure below is invented. Lina completes eighteen clear-source confirmations and two conflict cases. Omar completes six confirmations and six conflict cases. The headline totals, twenty versus twelve, invite a misleading conclusion.

Their worksheet instead contains these entries:

• Clear-source confirmation: Lina, eighteen completed, median twelve active minutes; Omar, six completed, median eleven active minutes.

• Conflict resolution: Lina, two completed, median forty-five active minutes; Omar, six completed, median forty-two active minutes.

• Review standard: both categories require an evidence-linked answer, limitations, and the next decision owner.

• Waiting: recorded separately because a pending account-owner response is not active research effort.

• Quality outcome: retained by case, including returns for correction and appropriate unresolved handoffs.

These invented observations show different task mixes. They do not establish that Omar is faster, because the samples are small and the individual conflicts may differ. The useful managerial question is whether the assignment system concentrates difficult work on one person while rewarding the other for volume.

Inspect cases before calculating an adjusted total

Read one ordinary case and one difficult case from each researcher. Check whether a completed conflict means genuinely resolved identity, a supported decision to keep entities separate, or merely a closed task. A defensible unresolved handoff can satisfy the agreed assignment; an unsupported confident answer cannot.

Add contextual notes for language requirements, source accessibility, interruptions, specialist assistance, and incomplete intake. Do not collect private personal explanations. The relevant question is what the work required and what resources were available, not speculation about the worker.

Avoid assigning an arbitrary multiplier such as one conflict equals four confirmations. That creates precision without evidence and encourages category inflation. If capacity planning needs an estimate, use observed ranges within categories, identify the period and sample, and retain the uncertainty. A planning estimate is not a universal productivity exchange rate.

Use the review to change the operating system

In the fictional example, the lead rotates a few suitable conflict assignments, offers joint review for difficult identity cases, and changes the dashboard to show task mix alongside accepted work. No retrospective score is imposed on either person. The next review asks whether delays and recurring defects changed within comparable categories.

If only one researcher handles a specialist category, report that there is no comparable peer group. Compare the process against its own requirements instead. Where evidence is too thin, the honest result is insufficient basis for comparison.

Keep the original assignment, revised category, accepted outcome, and explanation together. This makes the review challengeable and useful: colleagues can correct a misclassified task before an innocent operational number hardens into an unfair story about performance.

Continue with Measure Sales Navigator Prospecting with a Small, Honest Funnel or Compare Two Target Segments with a Fair Sales Experiment.