Recruitment Dashboard: What to Track and What to Ignore

A recruitment dashboard should help someone decide where to look next. It is not a wall of numbers, a second candidate tracker, or a score that claims the hiring process is healthy. Its job is narrower: present a small set of reliable workflow signals, connect each signal to its source, and make the responsible next action visible.

That distinction matters in candidate screening. A stage count can show movement without showing whether reviewers used the approved criteria. A clean funnel can hide missing evidence. An alert can reveal that a record is waiting, but not whether the wait is intentional. Build the dashboard so a recruiter can move from the summary to the underlying role, criterion, and candidate record before changing the process.

Current recruiting-system documentation illustrates the difference between a dashboard and its supporting reports. Greenhouse describes report dashboards as providing at-a-glance information about sourcing, pipeline health, goals, jobs, candidates, and recruiting processes; its widgets link to fuller reports. Workable's Report Center overview distinguishes current status reports from historical reports and documents the date field and update timing for each report. The practical lesson is simple: the summary points to the investigation. It is not the investigation itself.

Start with the decision, not the chart

Before choosing a visual, write the operating question and the person expected to act. “Which active roles need recruiter attention today?” is a dashboard question. “How many candidates exist?” is only a count until the audience, time frame, and next action are defined.

For each proposed view, record six items:

• Question: what the viewer needs to understand.
• Audience: the recruiter, hiring manager, or operations owner using the answer.
• Source: the authoritative role, candidate, scorecard, or workflow record.
• Time frame: the period or point in time represented.
• Drill-down: the source records the viewer can inspect.
• Action owner: the person responsible when the signal needs attention.

If one of those fields is blank, the view is likely decorative. Remove it or narrow the question until the team knows what it means and what happens next.

Keep current state, activity, and cohorts separate

Recruiting data changes meaning depending on how candidates are grouped and dated. Greenhouse's guide to cohort, activity, and snapshot reports distinguishes three useful perspectives: a cohort follows a defined group of candidates, activity records what happened during a period, and a snapshot captures the state at a point in time.

Use the same distinction in your cohort, activity, and snapshot views:

• Snapshot view: which roles and candidates were active at the defined check time, where they sat, and who owned the next step.
• Activity view: which reviews, stage movements, scorecard submissions, or exceptions occurred during the selected period.
• Cohort view: what happened to candidates who entered the process under the same role version or during the same intake period.

Do not place these numbers side by side as if they share a denominator. A current pipeline includes people who arrived at different times. An activity count describes events, not unique candidates. A cohort can mature after the reporting period ends. Label the view and time basis wherever the number appears.

Build five evidence-ready dashboard sections

1. Open-role attention

Show each active role's current version, recruiter owner, hiring-manager owner, next workflow checkpoint, and unresolved operating exception. This is not a league table of recruiters. It is a routing view that reveals where a role lacks a clear next action or a required input.

2. Candidate flow

Show current stage counts, movement during the chosen period, and time in stage only when the team has defined what starts and ends that stage. Greenhouse's pipeline history documentation lists applicants per stage, pass rate, and average days in stage, and it notes that candidates who bypass a stage may still affect stage totals. That caveat is a useful reminder to document how your own system treats skipped stages before interpreting the funnel.

3. Screening evidence readiness

Track whether the resume-review record is complete enough for the next person to inspect. Useful signals include required criteria with an evidence state, source links that resolve, material unknowns carried forward, and human recommendations tied to the approved role version. The recruiting metrics guide explains how to define evidence completeness and connect a signal to a workflow correction.

Keep “not established in the resume” separate from “does not meet.” A dashboard should surface missing evidence without converting uncertainty into a negative candidate judgment.

4. Review readiness

Show whether the next reviewer has the inputs required to work: current criteria, source evidence, submitted scorecards, named open questions, and a specific review request. A completion indicator should describe the record, not the person. “Scorecard source example missing” is actionable; “interviewer quality low” is an unsupported verdict.

5. Exceptions

Create a visible lane for unreadable files, conflicting role versions, missing owners, source mismatches, incomplete reviewer notes, and AI-assisted output awaiting verification. Every exception needs a reason, owner, next action, and reviewed update. The recruiting operations guide shows how to route exceptions without changing the candidate's evaluation.

Write a definition beside every signal

A trusted dashboard is built from explicit definitions. For every count or rate, document its population, numerator, denominator, date field, source record, excluded states, update event, and accountable owner. If the definition changes, preserve the effective date so earlier and later views are not silently blended.

Definitions should be available from the dashboard, even if the detailed note lives elsewhere. “Screening complete” might mean that every required criterion has an evidence state, every material uncertainty is named, and a recruiter has recorded a human recommendation. Another team may use a different completion rule. What matters is that the rule is visible, stable, and applied consistently within the view.

Reconcile a sample of dashboard values with the source records before relying on the view. Check a role with active candidates, one with a skipped stage, one with an exception, and one whose criteria changed. If the summary cannot be reproduced from the source, fix the definition or data path before adding more widgets.

Make every summary inspectable

A dashboard should let a skeptical reviewer move from an aggregate signal to the records that produced it. A stage total should open the relevant candidates. An evidence-readiness signal should reveal the incomplete criteria. A reviewer-completion signal should point to the expected scorecards and source notes.

This drill-down protects against two common errors: treating a clean visual as proof that the underlying records are sound, and asking a recruiter to reconstruct context from memory. The recruitment tracker guide describes the working fields and source links the dashboard should summarize rather than duplicate.

Give each audience a focused view

Recruiters, hiring managers, and recruiting-operations owners do not need the same screen.

• Recruiter view: active roles, candidates ready for review, unresolved evidence, exceptions, and owned next actions.
• Hiring-manager view: candidates ready for a defined review, the role criteria in force, supporting sources, open questions, and the requested decision.
• Operations view: missing owners, definition drift, repeated exceptions, source mismatches, and signals that need a workflow review.

Use shared definitions across views, but change the level of detail and requested action. One overloaded dashboard forces each audience to decode information that belongs to someone else.

Use AI to prepare signals for human review

A team may test AI for grouping verified exception reasons, drafting a plain-language summary of a change, flagging empty fields, or proposing which source records deserve inspection. Test each task on representative records and require the responsible person to compare the output with the source.

Do not allow a model to create missing evidence, infer capability from a dashboard pattern, change a candidate stage, or decide who advances. Generated summaries and alerts remain provisional until a recruiter or hiring manager verifies them. The dashboard should make that review state visible.

Recruitment dashboard launch checklist

• Can every view be tied to a named operating question?
• Is the intended audience visible?
• Are current state, activity, and cohorts labeled separately?
• Does every signal have a documented source and time basis?
• Can the viewer drill down to the underlying role or candidate record?
• Are status, evidence, and human recommendation kept distinct?
• Does each exception have an owner and next action?
• Have dashboard values been reconciled with representative source records?
• Are AI-assisted summaries visibly awaiting or recording human verification?
• Does every candidate decision remain with a qualified person?

Build the shortest path from signal to source

A useful recruitment dashboard does not make the hiring process look certain. It makes the process easier to inspect. Define the question, audience, source, time frame, drill-down, and action owner for every view. Keep aggregate movement beside evidence readiness, route exceptions visibly, and require a person to verify both the data and every candidate decision.