Setup Considerations: what analytics needs from your data

What this achieves

Almost every analysis in this section compares two things you recorded elsewhere — what your roles require against what your people have. Deciding how those two sides get maintained is what separates a report you can act on from one that quietly reports nothing.

The failure mode to design against

An analysis with nothing to compare does not say “no data”. It shows an empty result, and an empty result reads like good news.

Option	Description
No gaps found	Either every requirement is met, or no requirements were recorded.
No readiness scores	Either nobody has been scored, or no job profile states what a role requires.
No succession candidates	Either nobody has been nominated, or no position is marked critical.
Empty heatmap	Either nobody holds those skills, or nobody has recorded them.

Example: HC Corp’s Sales department shows no skill gaps. If Sales has 21 people and no job profile in Sales lists a required skill, that clean result is measuring nothing at all.

Before you circulate any empty result, establish which of the two explanations applies. This is the single most valuable habit in this section.

Decision 1: which roles state their requirements

Gap analysis, readiness, and succession all read requirements from job profiles. A job profile without skill and competency requirements is invisible to all three.

You do not need every profile complete on day one, but you do need to decide the order. Start with the roles you would struggle to fill, because those are the ones where an analysis has something to tell you.

Example: completing requirements for the engineering profiles at HC Corp before the administrative ones means the first gap analysis you run answers a question somebody actually asked.

Decision 2: how skills get recorded, and by whom

The other half of every comparison is what people hold. Decide up front whether self-assessment, manager assessment, or formal assessment is your source of record, because mixing them without deciding produces a heatmap where a level means something different in each department.

Also decide who chases it. A skills inventory that nobody maintains degrades quietly, and the analysis will not tell you it has — it will keep comparing against whatever was last recorded.

Decision 3: your revalidation cadence

Skills age. A level recorded three years ago still counts as a level in gap analysis, on the heatmap, and in readiness scoring, whatever its freshness status says elsewhere.

Decide how often skills must be revalidated and hold to it, or accept that your capability analysis describes a version of the company that no longer exists.

Decision 4: which analyses are live and which are generated

This is the distinction people get wrong most often, because both kinds look identical on screen.

Analysis	When it is calculated
Skill gap analysis, skill heatmap, demand and supply, headcount by department	Derived when you open the screen. Always current.
Workforce snapshots	Only when generated. Fixed at that date forever.
Flight-risk scores	Only when calculated. Carry their calculation date.
Predictions and skill-trend analyses	Only when generated. The screen shows the latest one for the scope.
Readiness scores	Only when calculated. Do not refresh when an employee’s skills change.

The consequence: a generated analysis is a statement about the day it ran. Decide who regenerates each one and how often, and write it down, because nothing in the product will chase you.

Example: if HC Corp calculates flight risk once and looks at it six months later, the list is six months old and looks exactly as authoritative as a fresh one.

Decision 5: your snapshot rhythm

Snapshots are the only way to compare two dates honestly, and they cannot be taken retroactively. A snapshot you did not take in March does not exist in June.

Pick a cadence before you need the comparison. Quarterly at company scope is the common choice. Also decide whether you need department-level snapshots — a company snapshot is not the sum of departmental ones, and the executive dashboard reads the company one.

Decision 6: reporting on small groups

Several analyses segment by department, grade, tenure, age group, and gender. Your departments are not all large enough for that to be anonymous.

Example: HC Corp’s People department has 6 people and Finance has 9. A flight-risk breakdown reporting “two at high risk in Finance” is describing two identifiable individuals, and a gender breakdown within a department that size may describe one.

Decide a minimum group size below which you will not circulate segmented figures, and apply it to anything that leaves the room — board packs and scheduled reports especially. The product will show you the small numbers; deciding not to distribute them is yours.

Decision 7: who reads what

Analytics is bounded by what the reader is entitled to see, and company-wide analysis is an administrative view. Managers reach their own team’s position through their team views rather than through company reports.

Two people can quote different numbers and both be right, because they were answering different questions about different populations. Where a figure will be discussed by more than one person, say what population it covers when you quote it.

Decision 8: your AI oversight cadence

If you use AI features, decide now how often bias monitoring is reviewed and who reviews it. Quarterly, marked as reviewed with real notes, is a defensible position. Generating a bias report for the first time in response to a complaint is not.

The same applies to credit monitoring: decide who watches consumption, because the first sign of a problem is usually a projection rather than an outage.

What is expensive to change later

Decision	Cost of changing later
Snapshot cadence	High. Missing snapshots cannot be recreated, so a gap in the history is permanent.
Job profile requirements	Medium. New requirements apply going forward, and past readiness scores were calculated against the old ones.
The source of record for skill levels	Medium. Changing it means the levels recorded before and after are not comparable.
A saved report’s type and filters	Medium. Neither can be changed after creation — you build a new report.
Your small-group reporting threshold	Low to set, high to undo. Figures already circulated cannot be recalled.

What happens next

Once both sides of the comparison are maintained, the analyses in this section start answering questions rather than reporting emptiness. Take a company snapshot as soon as the data is worth capturing — it costs nothing, and it becomes the baseline every later comparison is measured against.