Company values become useful when people can recognize the behavior a principle calls for and apply it to a real choice. A value such as transparency, for example, might require explaining who a policy change affects, which assumptions shaped it, and what tradeoffs were considered. Announcing the change without that context states an intention, but does not show how the intention guides a decision.
The practical method is to define each value in plain language, connect it to observable behaviors and anti-examples, test it against realistic conflicts, and assign an owner to review evidence over time. Keep stated intent, operational evidence, employee sentiment, and business outcomes separate. None automatically proves the others.
This guide shows how to create a value-to-behavior contract, involve employees without turning values into a popularity contest, apply principles to hiring and management, and maintain a decision record that can be reviewed without overstating what the evidence proves.
What company values are, and what they are not
Company values are principles an organization intends to use to guide behavior and decisions. They can help employees understand expected ways of working, but a short list of attractive words is not enough to guide a particular choice.
How do we intend to act?
Principles such as explaining consequential decisions or taking ownership of a customer problem.
Why, where, and what rule?
A mission describes purpose or core work, a vision describes a desired future, and a policy sets a specific requirement or restriction.
A useful definition test is simple: could an employee use the wording to choose between two plausible actions? If not, the phrase may be an aspiration or label that needs a behavior and a decision context. Values also do not override law, safety requirements, privacy obligations, employment requirements, or formal policy.
A public values page demonstrates stated intent. It does not establish consistent execution. For current examples, use official company sources and label them as stated values. HubSpot’s careers page currently describes its Culture Code as humble, empathetic, adaptable, remarkable, and transparent. Airbnb’s careers page lists four current values, and Gusto’s current careers page lists its own. Company wording changes, so older lists should be dated rather than presented as current.
Turn each value into a value-to-behavior contract
For each proposed value, record five elements:
- Plain-language definition: what the principle means in this organization.
- Observable behaviors: what someone could see, hear, or verify.
- Anti-examples: behavior that conflicts with the principle.
- Decision context: a realistic situation where the value affects a tradeoff.
- Owner and review date: who maintains the wording and when it should be reconsidered.
Then test the value against a conflict involving speed, cost, revenue, privacy, safety, or legal obligations. A useful value helps explain how a tradeoff is handled, not merely which outcome sounds desirable.
A value is operational only when people can identify the behavior it calls for, the behavior it rules out, and how it applies to a real tradeoff.
Consider an illustrative contract for Transparency:
- Definition: Explain the reasons and tradeoffs behind consequential decisions.
- Behavior: When changing a major policy, explain the assumptions, affected groups, and relevant constraints.
- Anti-example: Announce the decision without explaining its rationale or tradeoffs.
- Decision context: A change to customer-support coverage affects customer access, operating cost, and employee schedules. The decision record identifies which interests were considered and why.
- Owner and review: A policy owner maintains the approved version, effective date, and next review trigger.
Ask employees outside the drafting group to apply the wording to a realistic scenario. If they reach materially different interpretations, clarify the behavior or name the tradeoff explicitly. Google’s official philosophy page is an example of a company presenting principles with explanations. Use it as evidence of Google’s stated philosophy, not proof of a particular business outcome.
Develop values with employees, not just for them
Leadership remains accountable for approving the final values, but employees can provide concrete examples of decisions and behaviors that reflect the organization at its best. Look for recurring themes in examples rather than treating a poll as a vote on polished slogans.
Before collecting identifiable feedback, define its purpose, who can access it, how long it will be retained, and what level of reporting employees will see. Anonymous or aggregated feedback may be appropriate when candid responses could expose an employee to pressure. Raw feedback should not be mixed with approved value statements.
This is an organizational process, not a verified vendor automation. A source of record might be a controlled policy document or values register already used by the organization. Record candidate wording and feedback separately so a draft theme cannot be mistaken for an approved value.
Put values into the operating cycle
Translate approved behaviors into job-relevant interview prompts, onboarding examples, manager expectations, and performance discussions. Avoid using “culture fit” as a general hiring criterion because it can reward similarity instead of evidence of job-related behavior.
For a role that requires explaining difficult policy changes, connect the value to a structured interview question such as: “Tell us about a time you had to explain a decision that affected different groups differently. What did you communicate, and why?” Use the same rubric for candidates and require the interviewer to record specific evidence rather than a general impression.
The hiring manager owns the assessment. The people team maintains the rubric and checks that it is applied consistently. In onboarding, show a real and appropriately shareable decision example, then ask what behavior the value called for. In performance discussions, evaluate documented actions and context, not whether someone appears to embody a slogan.
Managers should also know what to do when values conflict or a policy creates a legitimate exception. A value should not be rewarded in isolation from safety, privacy, legal, or operational requirements. Organizations translating principles into repeatable practices may find operations and systems consulting relevant. This is a service link, not a prebuilt values workflow.
Prove practice with decisions, not slogans
Use a “proof before praise” rule. Support a claim that a value is practiced with a policy, documented decision, program, governance process, or disclosure. A public values page supports the narrower claim that an organization states that value.
Keep four evidence grains distinct:
- Value record: one approved value and its version.
- Decision record: one concrete decision, its context, date, and owner.
- Citation record: one source artifact supporting or explaining that decision. A decision may have several citation records.
- Outcome record: an aggregated measure for a defined population and period, with its method and source.
Employee sentiment and business outcomes are separate measures. Do not infer that values caused a change in retention, revenue, productivity, customer trust, or employee happiness without suitable evidence and a defined method.
The following is an illustrative internal data design, not a vendor schema. It records one value-to-decision relationship at the evidence grain and keeps citations and AI review runs separate:
{
"evidence_id": "ev_391",
"organization_id": "org_12",
"value_id": "val_07",
"value_version": "v2",
"decision_id": "dec_208",
"event_date": "2026-10-12",
"artifact_version": "policy-v2",
"reviewer": "operations_lead",
"review_status": "pending",
"citation_ids": ["cit_882", "cit_883"],
"ai_run_ids": ["run_44"]
}At this grain, the database uniqueness rule could be organization_id + value_id + decision_id + artifact_version. Enforce it with a database unique index or transactional upsert. A lookup followed by creation is not race-safe when concurrent writers operate. Citation rows should have their own identifiers, and each AI run should have a separate run identifier so repeated evaluations do not overwrite the underlying observation.
- The specific decision, date, context, and owner are identified.
- The relevant value version and artifact version are identified.
- Each supporting source artifact is retained as a separate citation record.
- A reviewer and review status are recorded.
- Any aggregated outcome is reported separately with a defined population, period, method, and source.
Coca-Cola’s current Code of Business Conduct page connects its code with employee actions and company values. That supports a claim about a governance document. It does not, by itself, show that every decision consistently follows every stated value.
Review values when the organization changes
Set a regular review date and triggers for material changes such as a merger, rapid growth, leadership transition, major shift in work practices, or repeated conflict between stated principles and decisions. Distinguish a wording refresh from a change in operating expectations.
For every revision, record what changed, why, who approved it, its effective date, its revision date, and which version it supersedes. Do not overwrite older wording if existing policies or evidence refer to it. A version history should make it possible to answer which value was in force when a decision occurred.
Use current official sources for company examples. HubSpot says its Culture Code has been revised dozens of times, while its current careers page describes the current wording. Adobe’s current Code of Business Conduct uses “Create the future,” “Own the outcome,” and “Raise the bar,” which differs from older Adobe value lists. Airbnb, Gusto, and Coca-Cola also publish current materials that should take precedence over undated historical lists.
Use AI only for bounded support
AI may help cluster de-identified open-ended feedback or draft candidate behavior descriptions for human review. Its role should be limited to suggesting themes, organizing source material, or proposing language. It should not identify respondents, judge whether an individual employee embodies a value, choose which values the organization adopts, or automatically change policies or performance outcomes.
Use deterministic checks for required-field validation. For example, confirm that an owner, date, source URL, approved status, and allowed status value are present. These checks do not require a language model. Preserve source provenance and require a named human reviewer before a suggestion becomes an approved value or an employee-facing decision.
If a bounded AI-assisted process is being designed, AI agent consulting may be relevant. The link describes a consulting service, not an existing values-classification product or verified workflow.
Questions leaders commonly ask
How many company values should we have?
There is no universally correct number. Choose a set employees can remember and apply without merging distinct principles into vague slogans. Values can be words, phrases, or sentences. Pair each format with behaviors, anti-examples, and decision contexts so people know what it means in practice.
Should values be single words or complete sentences?
Either can work. A single word is memorable but may be ambiguous. A sentence can be clearer but harder to remember. Test the wording against realistic decisions rather than selecting a format by convention.
What if employees disagree with a value?
Capture the disputed wording or decision, identify the competing values and affected groups, and assign an owner to review it. The response may be to clarify the wording, change a practice, or formally revise the value. Disagreement alone is not evidence of poor culture fit. It may reveal an unclear principle or a gap between stated expectations and actual decisions.
Can values conflict with revenue, speed, or customer demands?
Yes. That is why scenario testing matters. Ask what happens when transparency conflicts with confidentiality, customer responsiveness conflicts with safety, or speed conflicts with legal review. The value contract should state the escalation path and identify which policy or decision owner controls the exception.
Do company values improve business results?
Values can provide a shared reference for decisions and communication, but publishing them alone does not establish improved revenue, retention, productivity, customer trust, or employee happiness. To make an outcome claim, define the metric, population, measurement period, method, and comparison. Distinguish correlation from causation.
Start with one value that regularly appears in decisions. Define its behavior and anti-example, test both against a real tradeoff, assign an owner, and document the first decision that can be reviewed against it. That turns a statement into a practice leaders can inspect and improve.