Short answer
Organizational intelligence helps leaders answer a defined question about how work is happening and what to do next. It combines relevant records, employee and stakeholder accounts, professional expertise and operating context. A useful process preserves the source, coverage and uncertainty of each input, tests competing explanations and records an accountable human decision without turning employees into profiles.
Start with the decision that is blocked
“Understand the organization” is not an actionable brief. Name the decision, the work moment and the evidence threshold before collecting more material.
| Brief field | Example entry |
|---|---|
| Decision | Whether to test a revised returns handover in three stores |
| Decision owner | Regional operations director |
| Population | Returns colleagues and duty managers in the three stores |
| Period | The previous six weeks |
| Known observation | Unresolved returns remained open past the next shift |
| Open question | Is the cause unclear ownership, missing information, system access or another condition? |
| Evidence needed | Handover records, process version, employee accounts and manager context |
| Decision limit | No conclusion about an individual's capability |
| Review point | After two weeks of a controlled local test |
The US Federal Data Strategy practices advise identifying the key question and the data needed to answer it, documenting quality and provenance, and aligning quality with intended use. These are federal data practices, not an employment standard, but the sequence is useful: a question gives data quality a purpose.
Use four evidence lenses
One source rarely carries enough context for an organizational decision. Build the decision file from evidence that answers different parts of the question.
- Organizational records: process dates, staffing coverage, work events or service measures. These can show what was recorded, under the definitions used.
- Employee and stakeholder accounts: recent examples, explanations, disagreement and local language. These show what participants report and how they interpret the work.
- Professional expertise: knowledge from HR, operations, legal, safety, data and other relevant owners. Expertise helps test feasibility and risk.
- External research: applicable research or guidance that informs the method or the proposed action.
The CIPD evidence-based practice factsheet describes research, organizational data, professional expertise and stakeholder perspectives as evidence sources to consider together. Agreement does not automatically mean independence. Two reports may repeat the same system record, while a manager summary may already include the employee account later counted as a second source.
For each input, record its origin and what it can support.
| Source check | Question |
|---|---|
| Provenance | Who or what created the record, when and for which purpose? |
| Coverage | Which people, places, events and dates are present or missing? |
| Definition | What exactly does the field, theme or event mean? |
| Transformation | Was the source translated, coded, summarized, joined or filtered? |
| Independence | Does this add evidence or repeat another source? |
| Decision fit | Which part of the decision can it reasonably inform? |
Separate an account from an observation
An employee conversation can be current, detailed and valuable. It remains an account from a participant. It may describe a work event that can be checked against a record or observed in context, but follow-up questions do not automatically verify every statement.
Keep three fields distinct:
- Source material: what a person said or what a record contains.
- Observation: the bounded event or condition directly supported by the source.
- Interpretation: a possible explanation that may require another check.
The GOV.UK research analysis guide recommends recording what was seen or heard separately from what it means before forming findings and actions. Use that separation for employee context too.
For example, “The approval owner was not listed in the handover document used on Tuesday” is a checkable observation. “Managers do not communicate” is a broad interpretation. The second statement requires a defined standard, wider evidence and counterexamples before it can guide action.
Copy this decision-context worksheet
Complete one row for each material claim in the decision brief.
| Field | What to record |
|---|---|
| Claim | A bounded statement about the work, population and period |
| Supporting evidence | Source IDs and the relevant records or passages |
| Contrary evidence | A case that weakens, narrows or changes the claim |
| Missing perspective | A role, shift, site or stakeholder not yet represented |
| Alternative explanation | Another mechanism consistent with the evidence |
| Confidence | What is established, plausible and unresolved |
| Proposed action | A proportionate decision, check or reversible test |
| Owner | Person accountable for the action and communication |
| Review | Date, evidence to collect and continue/change/stop rule |
For a shorter meeting-ready version of this structure, use the HRBP data briefing worksheet.
Avoid building a permanent employee label from a situational account. A comment about one handover is evidence about that event. It is not a stable trait, a forecast or a basis for an automated employment decision.
Fictional example: a returns handover
Northbank Retail is fictional. A regional HRBP is asked why unresolved returns are carrying into the next shift at three stores.
The initial dashboard shows 28 carry-overs in six weeks. The definition changed in week three, so the earlier and later counts are not directly comparable. Eleven voluntary employee conversations include seven accounts of an unclear approval owner, but evening-shift participation is low. Two duty managers say the owner is covered in the process document. A document check finds that one store still displays the previous version. Another store has the current document and still records carry-overs.
The HRBP writes a bounded finding: approval information may contribute at one store, while the current evidence does not explain all three. The evidence does not support ranking stores or identifying a “best” employee.
The operations director approves a two-week test at the store with the outdated document. The team replaces it, names the duty owner in the handover and records whether the next shift can resolve the same defined cases. Evening-shift colleagues are included in the review. The other stores continue source checks rather than receiving the same intervention by assumption.
Check freshness and quality separately
Recent information may be incomplete, inconsistently defined or based on a narrow group. Older material may be accurate for its original period but stale for today's decision. The UK Government Data Quality Framework defines quality as fitness for purpose and treats completeness, uniqueness, consistency, timeliness, validity and accuracy as separate dimensions. It also notes that timeliness can trade off against completeness or assurance.
Display the source date, availability date and decision period. If a definition or population changed, show that beside the comparison rather than smoothing it into one trend.
Where employee conversations fit
A focused employee conversation can add recent examples and missing work context to a defined question. Managers receive a prepared brief, not raw employee responses. HR uses grouped material only when at least five respondents are represented, and that floor does not remove the need to consider whether context could identify someone.
The Lontra product overview explains the employee conversation and manager-brief workflow. Use conversation material as one declared source in the decision file. Human owners still check provenance, compare relevant records, test alternatives and decide what follows.
Organizational intelligence becomes useful when a leader can see why a conclusion was reached, what remains uncertain and which evidence will determine the next step. Its value lies in better-framed decisions and institutional learning, not in producing more data or profiles.