How true must GTM data be before the system can act?

GTM data is true enough when a team can use it for one named commercial decision without guessing what the field means, where the claim came from, whether it is still current, who owns correction, or what happens when sources disagree.

That standard is smaller than a perfect database and stronger than a completed field.

The failure mode is field confidence. A CRM value looks factual because it is structured. A dashboard repeats it. An automation routes from it. An AI agent summarizes it. Yet nobody can inspect the event behind the value or explain when it stops being reliable. The system gains confidence as the evidence gets farther away.

Definition

Definition: Truth in GTM is decision scoped commercial evidence with explicit meaning, an inspectable source, a freshness condition, an accountable owner, and a path for resolving conflict.

Truth is the first layer of the Lorde stack because every downstream layer inherits its quality. A Playbook turns evidence into a normal decision. Architecture records and moves that decision. An Operator maintains the loop and resolves exceptions.

If the evidence is weak, the rest of the system can still run. It simply runs the wrong commercial state with greater consistency.

Truth is not the same as a clean database

A founder does not need every record completed before the business can act. The useful question is narrower: which claims must be reliable for this decision?

Consider a proposal decision. The team may need to know whether the buyer has confirmed the problem, who participates in the economic decision, which constraint shapes scope, and which next step was accepted. An unrelated firmographic field can be missing without blocking the proposal. A polished record with stale buyer intent can be far more dangerous.

The UK Government Data Quality Framework treats quality as fitness for purpose. It describes dimensions such as completeness, consistency, timeliness, validity, and accuracy. The framework is not a sales method, but the lens is useful. Presence alone does not make a claim fit for use.

A commercial record can be complete but stale. It can be current but ambiguous. It can be accurate to the last conversation but unsupported by a source anyone else can inspect. The Truth layer should expose those differences instead of compressing them into a green required field.

This also protects commercial capacity. Teams lose capacity when people repeatedly reopen the same facts, search private messages, ask the founder to remember context, or repair actions triggered by bad evidence. Better Truth reduces reconstruction work before it increases automation.

Build a five part evidence contract

Apply this contract only to claims that can change a consequential action. Start with one claim, not the whole CRM.

1. Meaning

Write what the claim means in observable terms.

“Decision maker identified” is weak if one seller means a job title and another means a person who confirmed authority. A usable definition states the evidence required and what the field does not establish.

Meaning belongs in Truth. The action triggered by that meaning belongs in the Playbook. Keeping those separate prevents a field label from silently becoming a decision rule.

2. Source

Name the event or record that supports the claim.

A source may be a buyer statement, signed document, call note, verified system event, or approved calculation. “The seller knows” may be legitimate context, but it should not become invisible evidence after the opportunity moves to another owner.

Inspectability matters more than centralization. The evidence can live in several systems if the commercial path preserves where it came from and how to retrieve it.

3. Freshness condition

State what event can make the claim unreliable.

Do not invent a universal age limit. Buyer intent may change after a budget review, stakeholder change, delayed implementation, new objection, or long silence. Company facts may remain useful for much longer.

A freshness condition ties review to the decision. It answers, “What would require us to confirm this again?”

4. Owner

Assign the role responsible for keeping the definition usable and correcting normal defects.

Ownership is not permission to rewrite buyer reality. It is responsibility for resolving missing evidence, maintaining the field contract, and ensuring the next user can understand the state.

When every user can edit a consequential claim but nobody owns its quality, Architecture stores change without accountability.

5. Conflict path

Define what happens when valid sources disagree.

The latest note may conflict with a signed scope. A buyer message may conflict with a CRM stage. Two stakeholders may describe authority differently. The system should not choose silently.

The conflict path can prefer one source, request confirmation, hold the action, or escalate judgment. This is where Truth meets Playbook and Operator ownership.

Worked example: economic buyer involved

Suppose the proposal workflow advances whenever the CRM field “economic buyer involved” is marked yes.

The field is complete across the active pipeline, but its contract is missing.

One seller marks yes after seeing a senior title. Another waits for direct participation. A third copies the value from an old opportunity. The proposal automation works exactly as configured, yet the team cannot explain why the buyer is ready for that step.

Repair the claim before repairing the automation:

  • Meaning: The person who can authorize the commercial commitment has been identified, and their role in the decision has been confirmed through an inspectable interaction.
  • Source: Link the buyer event or note that supports the claim.
  • Freshness: Reconfirm after a stakeholder change, a material scope change, or evidence that the decision process has shifted.
  • Owner: The opportunity owner maintains the normal claim. A named revenue role audits recurring defects.
  • Conflict: If title, buyer statement, and account history disagree, hold the proposal trigger and route the exception for judgment.

Now the Architecture can route from a defined state. The Playbook can specify the normal next action. The Operator can inspect exceptions and improve the contract when real cases expose a missing condition.

Decision rule

Use this rule before a commercial claim drives a workflow, dashboard interpretation, or AI action:

If the team cannot state the claim's meaning, source, freshness condition, owner, and conflict path, treat it as context, not as an execution trigger.

Classify the claim in one of three operating states:

  • Trigger: All five properties are explicit. The claim may drive the approved normal action.
  • Context: The source is inspectable, but one property is incomplete. A person may use it with judgment, but the system must not present it as a settled trigger.
  • Unknown: The source is absent, invalid, or in unresolved conflict. Stop the dependent action and collect or resolve evidence.

The failed property still points to the repair. Meaning and source begin in Truth. Divergent action belongs in the Playbook. A defined state that cannot travel belongs in Architecture. Recurring decay and unresolved exceptions need Operator ownership. Automation is earned when the claim survives a handoff without private explanation.

NIST AI RMF Map guidance starts with establishing context, including intended purpose, users, expectations, assumptions, limitations, and deployment conditions. It does not define sales truth. It supports the sequence principle: context should be understood before a system is trusted to act inside it.

Checklist

Run this audit on one consequential claim this week:

  • Name the commercial decision the claim can change.
  • Write the claim in observable language.
  • Identify the evidence required and its inspectable source.
  • State the event that makes the claim require confirmation.
  • Assign one role to maintain the normal contract.
  • Define what happens when sources disagree.
  • Sample current records and mark each failed property.
  • Repair the first property before adding another field or automation.
  • Test whether a new owner can use the claim without private explanation.
  • Record exceptions so the contract can improve from real decisions.

The output is not a data quality program. It is one commercial claim that the system can use honestly.

What this is not

This is not a demand for perfect data. Missing evidence can be an honest state, provided the system does not present absence as certainty.

It is not a universal freshness schedule. The relevant review condition depends on the claim and the decision.

It is not an argument against AI or automation. Both can expand commercial capacity when they inherit defined evidence and bounded authority. The sequence is evidence, decision, architecture, then scale.

FAQ

Does every CRM field need an evidence contract?

No. Start with fields that can qualify, disqualify, route, price, commit, forecast, or trigger external action. Low consequence context can use lighter governance.

How current must commercial evidence be?

Current enough for the decision it governs. Define the event that invalidates or weakens the claim instead of copying a universal time limit.

Can AI decide which source is true?

AI can compare sources, identify conflict, and apply an approved precedence rule. When the conflict requires commercial judgment or changes buyer impact, the system should escalate rather than invent certainty.

If consequential fields keep producing disputes or bad actions, a Lorde GTM diagnosis can trace whether the restriction sits in Truth, Playbook, Architecture, or Operator ownership.

Lorde

Message on WhatsApp