Who Owns GEO? RACI for SEO, Content, PR, Product Marketing, and Analytics

TL;DR

• Give the program one accountable owner, not one all-purpose department. A CMO or VP owns the investment boundary; a GEO program lead owns the operating method and backlog; specialist functions remain accountable for truth, production, distribution, analytics, controls, and deployment.


• Build the RACI around decisions and artifacts. “Content owns GEO” is too vague. “Product Marketing approves canonical claims; Content accepts the draft; Engineering accepts production deployment” can be operated and audited.


• Separate Accountable from Responsible. Accountable means final acceptance or decision authority. Responsible means doing or coordinating the work. Several people can contribute, but each row should normally have one final accountable role.


• Do not make SEO the owner of facts it cannot approve. SEO/GEO can own buyer-question design, observation methods, diagnosis, and retrieval requirements. Product Marketing or Product should own product truth; Analytics and RevOps should own commercial-event definitions.


• Connect evidence production to maintenance. Research, customer proof, PR interpretation, content packaging, technical retrieval, and refresh require different owners. Publication is a handoff, not the end of the work.


• Use durable records and acceptance evidence. A claim registry, metric dictionary, prompt registry, action register, distribution record, decision log, and dependency register reduce committee memory and make escalation specific.


• Start with the downloadable RACI and replace every illustrative assignment. Adapt it to your company’s reporting lines, controls, team size, market, risk, and systems. A template cannot confer authority your governance model has not granted.

Why Does GEO Not Belong to One Department?

Generative engine optimization crosses how a company is understood, retrieved, cited, recommended, visited, and evaluated. No single marketing function controls that whole path.

SEO can study buyer questions and source environments, but it cannot unilaterally approve a pricing claim. Content can package an expert answer, but it cannot decide whether a customer has consented to public proof. PR can earn interpretation, but it should not redefine the research method. Analytics can define a qualified event, but it cannot make Engineering ship a fix. A CMO can fund the program, but should not become the reviewer for every paragraph.

GEO is an operating system, not a publishing queue

A useful program connects scope → measurement → diagnosis → accepted action → re-observation → value review. The in-house AI-search operating system explains the broader model. This guide goes deeper on decision rights: exactly who accepts the objects moving through that loop.

One person can hold several roles

A smaller company may have one leader acting as GEO lead, SEO owner, content strategist, and analyst. Keep the accountabilities separate even when one name appears several times. That makes missing skills, conflicts, and future handoffs visible.

The org chart is not the workflow

Reporting lines describe management. A RACI describes how a defined object reaches acceptance. The Content team may report to Brand while a technical SEO release depends on Product Engineering. Map the real path instead of forcing the work into a neat department diagram.

What Does RACI Mean for a GEO Program?

RACI is a responsibility-assignment model. Use it at the level where a decision can be accepted, rejected, revised, or escalated.

Use one Accountable role per row

Two executives can sponsor a program, but an individual decision still needs a final acceptance path. If two approvals are independently required—for example, product accuracy and legal permission—make them separate RACI rows. Do not hide a two-gate process inside “A/A.”

Responsibility is not authority

The writer may be responsible for a draft and unable to approve a regulated claim. The analyst may calculate a KPI and be unable to approve its use in an investor narrative. Assign authority according to existing governance, contracts, and policy.

Consultation needs a boundary

A Consulted role should know what input is requested, by when, and what happens if it does not respond. “All stakeholders are consulted” turns a workflow into a waiting room. High-consequence decisions deserve more review; routine decisions should not inherit the same burden.

Which Decision Rights Must the GEO RACI Separate?

“Ownership” hides several different rights. Separate them before naming teams.

Metric ownership is not program ownership

Analytics may own the formula, denominator, exclusions, and data pipeline. The GEO lead owns how that metric participates in a diagnosis. The sponsor owns whether the evidence justifies another quarter of investment.

Action ownership is not claim ownership

Content can own a rewrite while Product Marketing owns the underlying product claim. Record both. Otherwise, the action owner is pressured to invent truth to meet a deadline.

Approval is not contribution

Someone who provides a quote, screenshot, dataset, edit, or engineering estimate has contributed. That does not automatically make them Accountable. Treating contribution as approval makes retrospectives impossible.

Which GEO Team Structure Should You Choose?

Choose the smallest operating model that can complete the measurement-to-action loop. The labels below are patterns, not universal maturity levels.

Default to a hub-and-spoke model

A small hub can own the charter, method, backlog, dependency map, and review cadence. Spokes retain authority for claims, content, engineering, analytics, PR, customer proof, and controls. The hub coordinates; it does not annex the company.

Choose the hub by bottleneck

If unreliable evaluation is the bottleneck, SEO/GEO or Analytics may lead. If contradictory product facts are the bottleneck, Product Marketing may lead. If approved action never ships, the program may need a delivery-oriented owner. Revisit the structure when the bottleneck changes.

Keep an internal owner when using a partner

An agency or platform cannot approve internal truth, production access, risk exceptions, CRM definitions, or investment. The build, buy, or partner guide can help locate the boundary. The internal owner remains accountable for the company’s decisions.

What Should the CMO or Executive Sponsor Own?

The sponsor owns why the program exists, what it may consume, which decision it must produce, and who can stop it.

Own the decision, not the dashboard

The sponsor should be able to say what decision the evidence will support. “Improve our AI score” is not an executive decision. “Fund a second market after the pilot demonstrates reliable measurement, deployable actions, and qualified-demand evidence” is closer.

Give the program lead real authority

The lead needs authority to request named owners, reject method-breaking shortcuts, pause unsupported reporting, and escalate blocked decisions. Responsibility without authority creates a coordinator who can document failure but not change it.

Protect stop authority

A program should be able to stop when measurement is unreliable, truth is unavailable, controls prohibit necessary data use, actions cannot ship, or the cost exceeds expected decision value. Stopping is an operating decision, not an admission that citations were lost.

What Should the GEO Program Lead Own?

The GEO program lead is Accountable for the connected operating method. This role may sit in SEO, Growth, Product Marketing, Content, or a program office.

Maintain the connective tissue

The program lead keeps identifiers aligned across prompts, observations, claims, pages, actions, deployments, reruns, and commercial events. The systems can remain separate; the records need a traceable route.

Prevent one workstream from defining success

SEO visibility, PR coverage, content output, technical tickets, and qualified pipeline are different signals. The lead prevents whichever function has the best dashboard from turning its metric into the whole program.

Preserve disagreement

If Analytics calls a result not comparable, Product Marketing disputes a claim classification, or Legal narrows a reuse condition, retain the decision and reason. Governance should make uncertainty inspectable rather than producing false consensus.

What Should SEO and the GEO Method Owner Own?

SEO/GEO should usually own the buyer-question framework, observable answer method, retrieval diagnosis, and action hypothesis—not every downstream approval.

Own methods that can be inspected

Record answer products, modes, locales, clocks, panel eligibility, missing states, sampling, relevance, repetitions, providers, versions, and limitations. The Community’s guide to what AI-search dashboards really measure shows why a visible score may reflect provider and collection choices as well as brand outcomes.

Diagnose before assigning content

An absent recommendation may involve discovery, retrieval, reranking, answer composition, source availability, evidence, claim fit, landing continuity, or product fit. “Write a blog” is an action category, not a diagnosis.

Hand off truth questions

When the method reveals inconsistent pricing, positioning, availability, integration, or evidence, route the truth question to Product Marketing or Product. SEO can show inconsistency; it should not decide which fact becomes official.

What Should Product Marketing and Product Own?

Product Marketing or Product should own canonical product truth and buyer-positioning boundaries. The precise role depends on the company.

Maintain a claim registry

Claims about features, prices, integrations, service areas, certifications, availability, customer outcomes, and leadership change. Give each material claim an owner, evidence route, limitation, valid date, review date, public destination, and retirement path.

Approve reusable truth, not every keyword

Product Marketing does not need to approve a list of semantically related phrases. It needs to approve the substantive statements those phrases express. Keep review focused on consequence.

Correct the source chain

The Community’s analysis of claim drift across the AI answer pipeline is useful because drift can begin inside the company. When sales decks, help docs, product pages, and press releases disagree, no prompt tactic repairs the truth layer.

What Should Content Own?

Content should own editorial transformation and production acceptance: turning approved truth and evidence into an accurate, useful decision asset.

Own the editorial judgment

Content is not a transcription service for SEO briefs or executive notes. The editor decides whether the asset explains the problem, shows evidence, names uncertainty, respects the audience, and avoids inflated conclusions.

Separate draft acceptance from production acceptance

A perfect document in a content system can fail through rendering, navigation, links, structured data, scripts, or deployment. Content accepts the editorial artifact; Web/Engineering accepts production behavior.

Make refresh an assigned object

Every material evidence asset needs a review trigger: product change, method change, new contrary evidence, claim expiry, broken source, market change, or scheduled review. “Published” is a state, not a permanent truth certificate.

What Should PR and Communications Own?

PR and Communications should own truthful distribution and relationship context, not the underlying validity of a study they did not conduct.

Earn interpretation, not duplicated wording

Ten copies of one announcement do not create ten independent confirmations. Give specialists enough method and context to test, qualify, or disagree with the finding. PR owns the route; the independent party owns its interpretation.

Keep coverage and AI outcomes separate

Coverage can expand source availability and category interpretation. It does not guarantee retrieval, citation, recommendation, referral, or revenue. Record those events separately and avoid crediting one pitch for a later outcome without suitable evidence.

Route corrections back to the source

If external coverage reveals ambiguity or error, update the primary source, claim registry, source packet, and derivative assets. A correction hidden in one email does not repair the evidence system.

What Should Analytics and RevOps Own?

Analytics owns data and metric contracts. RevOps owns CRM-stage definitions and reconciliation. They protect the program from converting visibility into revenue by label.

Write the event contract first

Define event, unit, eligibility, exclusions, clock, data source, identity rule, missing-data treatment, version, and owner before reporting trend. GA4 AI-traffic setup is one observable referral layer, not a complete causal system.

Own comparability decisions

If the prompt panel, provider, analytics channel definition, consent mode, CRM stage, or attribution window changes, Analytics should say whether the comparison remains valid, directional, or not comparable.

Give executives a confidence statement

Report what is observed, what is associated, what is inferred, what is missing, and what decision the evidence can support. The GEO KPI scorecard for CMOs provides a layered reporting design.

What Should Web and Engineering Own?

Web and Engineering own production behavior, access, release acceptance, and rollback for the systems they control.

Require an executable brief

An action should specify affected route, current behavior, expected change, dependencies, acceptance evidence, test environment, production owner, rollback, and re-observation. “Improve AI crawlability” is not a ticket.

Do not make release access informal

The GEO program should follow existing security, change-management, and release controls. Urgency does not grant marketing production authority.

Preserve deployment facts

Record what changed, when, by whom, where, under which version, and with what acceptance result. Without a deployment record, a later answer change cannot be responsibly aligned to the intervention window.

What Should Legal, Privacy, Security, and Brand Control?

Control functions should own decisions within their mandate, with service boundaries that keep routine work moving.

Use risk tiers

A regulated product claim, customer performance statement, internal data export, and generic educational definition do not need identical review. Define which tier requires preapproval, sampling, post-review, or no specialist review.

Publish the control path

Teams need the intake format, reviewer, service expectation, deputy, escalation route, and required evidence. An unnamed “Legal review” cell creates delay without accountability.

Preserve the reason

Record why wording was approved, narrowed, rejected, or expired. The reason helps future reuse owners avoid reopening the same decision or stripping away a material limitation.

What Should Customer Success and Sales Own?

Customer-facing teams own relationship knowledge, evidence permissions, buyer-language signals, and commercial-stage discipline—not the GEO method.

Customer Success can originate evidence

Customer proof needs consent, scope, dates, method, and owner. Customer Success or Customer Marketing is often best placed to manage the relationship; Product Marketing and Legal may still approve reuse.

Sales should not silently change qualification

If the definition of an accepted lead changes after a campaign, version the change and avoid rewriting history. RevOps owns the stage contract; Sales supplies operational feedback.

Use field evidence without pretending it is a survey

Five repeated objections can justify investigation. They do not establish market prevalence. Label qualitative input and use it to form a bounded question.

What Does the Master GEO RACI Look Like?

The table is illustrative. Replace role names and split rows where your company requires independent approvals.

Split combined cells during implementation

“A/R” is acceptable when the same role both performs and accepts the work. If different people do those jobs, separate them. The downloadable model uses named columns so you can replace every combined label.

Add deputies and effective dates

A role-only RACI can fail during leave, reorganization, or vacancy. Maintain named owners, deputies, effective dates, and a vacancy rule in the operating register.

Test the RACI with one real action

Walk one material issue from observation to value review. If the team cannot identify the next owner, acceptance gate, source record, or escalation date, the RACI is descriptive rather than operational.

Who Owns the Original-Research Supply Chain?

The GEO Community’s AI-search supply-chain model for original research uses six stages: question, method, primary publication, independent interpretation, category reuse, and maintenance. This article transforms those stages into ownership gates.

Start with a question that can disappoint the company

Research should be able to complicate the preferred marketing conclusion. The Accountable research owner protects that property; the sponsor protects the team’s permission to publish a null or mixed result.

Keep the method beside the finding

PR and Content can compress a study for different audiences, but a repeatable claim should retain the dates, sample, selection, coding, comparison conditions, and limitations necessary to interpret it.

Assign maintenance before launch

A source needs an owner and review trigger before derivative pages, sales decks, PR, and AI answers reuse it. Otherwise, a time-bound observation becomes an ownerless company fact.

Who Owns the Evidence Content Flywheel?

The Community’s evidence content flywheel describes evidence → packaging → distribution → retrieval → refresh. Treat it as an operating loop, not a universal ranking law.

Do not centralize every stage in Content

Content can coordinate the asset, but it should not self-approve customer consent, product truth, research validity, distribution relationships, technical release, and commercial interpretation.

Design the return path

Distribution can reveal questions. Sales can reveal objections. Measurement can reveal claim mismatch. Support can reveal outdated facts. Route those signals back to the evidence or truth owner instead of only adding topics to a calendar.

Measure the loop in layers

Track evidence health, accepted assets, distribution outcomes, retrieval observations, qualified demand, maintenance status, and cost separately. A higher citation count does not prove the flywheel caused revenue.

How Do You Separate Metric Owners From Action Owners?

Every material metric should have a definition owner. Every accepted action should have a delivery owner. Every investment decision should have an executive owner.

Use separate IDs

Give prompts, observations, claims, actions, deployments, and events stable identifiers. Linking them is more reliable than squeezing them into one giant score.

Preserve null and adverse results

If an accepted change produces no comparable movement, a mixed result, or a regression, retain it. The action owner completed the test; the program learned something. Do not delete the row to protect a team KPI.

Make proprietary metrics inspectable enough to govern

GeoZ may use proprietary algorithms and metrics, but the program still needs the metric’s unit, eligibility, components, exclusions, version, owner, and appropriate use. The GeoZ Metrics Dictionary separates observable events from diagnostic constructs.

How Should GEO Handoffs Work?

A handoff is complete when the receiving owner accepts an object against declared criteria. Sending a document or ticket is not acceptance.

Write acceptance criteria before work starts

An action brief should define what the receiver will inspect. Acceptance can cover accuracy, accessibility, technical behavior, evidence linkage, control approval, analytics instrumentation, and rollback.

Distinguish blocked from rejected

Blocked means a dependency prevents a decision. Rejected means an accountable owner decided the object does not pass. Deferred means it is valid but not currently prioritized. These states require different next steps.

Stop reopening accepted work informally

If new evidence requires a change, create a versioned revision. Do not let an endless comment thread erase who accepted what and when.

What Cadence Keeps the RACI Alive?

Use meetings only where a decision benefits from synchronous discussion. Every forum needs input, decision, output, and owner.

Every cadence and duration above is illustrative. Adapt it to work consequence, existing forums, team size, and release reality.

Remove status-only attendees

Informed roles can read the record. Consulted roles attend when their input is material. Responsible and Accountable roles attend when an actual decision is scheduled.

Protect maker time

Use asynchronous artifacts for routine updates. Synchronous time should resolve disputed evidence, blocked dependencies, acceptance, risk, or investment.

Audit decisions, not attendance

A well-attended meeting that produces no owner, acceptance state, next date, or record has not advanced the operating loop.

Which Systems of Record Does the Team Need?

The program can use existing content, analytics, product, project, and CRM systems. It needs declared systems of record and stable links between objects.

Avoid five competing spreadsheets

The prompt registry and claim registry do not need to share a database. They do need clear authority and identifiers. Duplicate unofficial copies create hidden method and truth changes.

Give every record a maintenance owner

Ownership includes adding, correcting, versioning, archiving, granting access, and responding to disputes. Creation alone is not stewardship.

Keep confidential and public evidence distinct

Mark what may be published, shared with a partner, summarized, aggregated, or retained internally. Retrieval value never overrides contractual, privacy, security, or ethical limits.

How Should the Team Escalate Blockers?

Escalation should produce a decision before the blocked work loses value. Escalate by consequence and latest useful decision date, not by job title alone.

The labels and response windows are illustrative. Use the company’s incident and governance policies for actual severity.

Escalate the decision package

Provide the blocked object, owner, due date, consequence, evidence, options, recommended route, latest useful decision time, and default if no decision occurs.

Do not escalate ordinary prioritization as an incident

Not every deferred article or ticket is a leadership crisis. The program lead should resolve routine tradeoffs within the charter.

Close the loop after escalation

Update the action, dependency, decision, method, claim, and communication records affected by the outcome. A decision made in chat is not complete until the operating system reflects it.

How Does Ownership Change as the Program Matures?

Before changing the organization, test whether the current RACI works. The audit below scores operating evidence, not individual performance.

Use an illustrative 80-point acceptance test

Each of the 40 checks has an illustrative weight of 2 points. Award 2 when the inspected evidence passes, 1 when it is partial, and 0 when it is absent or contradictory. The sample size, decision window, repair window, and retest window are planning inputs—not universal service levels. Replace them based on consequence and company policy.

Use these illustrative interpretation bands only after replacing the inputs: 0–39 means ownership is too incomplete for scaled work; 40–55 means a bounded pilot may expose repair priorities; 56–69 means the loop can operate with named gaps; 70–80 means the inspected sample is well governed, not that AI outcomes are guaranteed.

Complete a launch checklist

• Confirm the sponsor’s decision and stop authority in writing.


• Name the GEO program lead, deputy, and escalation route.


• Map every generic role to a real title and person.


• Remove RACI rows that are outside the accepted scope.


• Split any row that hides independent approval gates.


• Confirm every Accountable role has legitimate authority.


• Confirm every Responsible role has capacity and system access.


• Give every Consulted role a question and response boundary.


• Route Informed roles to records instead of standing meetings.


• Declare systems of record for prompts, claims, metrics, and actions.


• Add versions and effective dates to governed objects.


• Define blocked, rejected, deferred, accepted, and retired states.


• Write handoff acceptance evidence before production begins.


• Define the latest useful decision date for dependencies.


• Preserve null, mixed, adverse, and incomparable results.


• Separate visibility observations from commercial events.


• Publish the legal, privacy, security, and brand review routes.


• Test one real issue through the full ownership chain.


• Record the test’s delay, conflict, exception, and decision.


• Schedule the first ownership repair and executive review.

Roles can change as the bottleneck moves. Do not confuse more staff with greater maturity.

Centralize standards, federate truth and delivery

A central GEO hub can own method contracts and governance fields. Product teams remain closer to truth; regional teams remain closer to market context; delivery teams remain accountable for their systems.

Add specialization when consequence justifies it

A dedicated research ops, analytics engineering, customer evidence, or technical GEO role can help at scale. Do not create a role because the label is fashionable; create it because a recurring object lacks an owner.

Re-run the RACI after reorganization

New reporting lines, vendors, products, acquisitions, consent rules, or release systems can invalidate assignments. Review ownership on change, not only annually.

How Do You Implement the GEO RACI in 90 Days?

Use 90 days as an illustrative governance window, not a universal implementation promise. The enterprise GEO first-90-days playbook covers the full program; this sequence focuses on ownership.

Days 0–30: map reality

Interview the people who actually approve claims, publish assets, deploy changes, define events, manage customer permission, and make budget decisions. Do not design from titles alone.

Days 31–60: run an ownership test

Choose one bounded buyer route and one material issue. Move it through observation, diagnosis, truth, action, acceptance, deployment, rerun, and value review. Measure handoff delay and rejection reason as operating evidence—not staff-performance ranking.

Days 61–90: repair and authorize

Remove duplicated approvals, split overloaded rows, name deputies, publish control paths, connect records, and set escalation defaults. The sponsor then decides whether the structure can support another operating period.

How Do You Use the Downloadable GEO RACI?

Download the GEO ownership RACI CSV. It includes 36 illustrative rows for program, measurement, truth, research, evidence, content, technical delivery, distribution, analytics, controls, operations, and partner work.

Replace roles before using it

Map generic labels to actual roles and names. Confirm that each Accountable role has the authority the row assumes. Add deputies, effective dates, risk tiers, and local approval requirements.

Split rows where gates differ

If product accuracy and legal permission require separate decisions, use two rows. If Web and Data Engineering accept different parts of a deployment, separate their objects. Precision is more useful than a compact chart.

Delete work that is outside scope

The template is not a maturity checklist. A small pilot may not need PR distribution, customer evidence, or partner access. Removing irrelevant rows makes live accountability clearer.

How Can GeoZ Support the GEO Team Structure?

GeoZ can provide a Value as a Service layer across its in-house tools, proprietary algorithms and metrics, LLM Taste, diagnosis, content and technical execution, and operating review. The internal team still owns company truth, approvals, production authority, data permissions, CRM definitions, and investment.

Use GeoZ when the connective layer is missing

A team with mature measurement, diagnosis, execution, and governance may not need a partner. A team that only lacks monitoring may need software. GeoZ fits when the organization needs a more complete measurement-to-action operating layer.

Write the partner RACI before access

Define data, methods, systems, deliverables, acceptance, response, approvals, branding, confidentiality, escalation, transition, and deletion before work starts. How GeoZ Works shows the service loop; the contract must still state the actual boundary.

Keep the company accountable for company decisions

GeoZ can recommend and execute within scope. It cannot confer legal approval, invent product truth, control answer engines, redefine qualified pipeline, or guarantee citation, ranking, traffic, leads, pipeline, or revenue.

Who Should Own GEO in Your Company?

One executive should own the investment decision. One program lead should own the connected method and backlog. Specialist functions should remain accountable for the decisions only they can legitimately make.

Do not assign GEO to a department and declare governance complete. Assign the charter, prompt registry, measurement contract, metric dictionary, claim library, content brief, research method, distribution plan, deployment record, commercial-event contract, dependency register, and value decision. Test the handoffs with real work. Then revise the model where authority, evidence, or execution fails.

If the team has many dashboards but no accepted action—or many actions but no trustworthy measurement—contact GeoZ. Bring one product, one ICP, one buyer route, the current RACI, five material claims, and the backlog that is not moving. The first useful output is a clearer responsibility boundary, not another visibility promise.

FAQs

Should SEO own GEO?

SEO or a GEO lead can own buyer-question design, measurement methods, retrieval diagnosis, prioritization, and the connected program backlog. It should not automatically own product truth, customer permission, legal approval, engineering deployment, CRM definitions, or investment. Use a hub-and-spoke structure with one program owner and distributed specialist accountability.

What is the difference between Accountable and Responsible in a GEO RACI?

Responsible roles do or coordinate the work. The Accountable role has final acceptance or decision authority for that object. Several people may be Responsible or Consulted, but each row should normally have one Accountable role. If two independent approvals are required, make separate rows rather than hiding them in one cell.

Who owns AI-search metrics and reporting?

Analytics or a named metric owner should maintain unit, formula, denominator, source, exclusions, missing-data rule, version, and data quality. The GEO lead owns how the metric supports diagnosis and action. The CMO or sponsor owns the investment decision. Mention, citation, recommendation, referral, accepted lead, pipeline, and revenue must remain separate.

Should PR or Content own third-party evidence for GEO?

Ownership is split. Research, Product Marketing, or Customer Success may own the underlying evidence and permission. Content owns editorial packaging. PR owns truthful distribution and relationship context. Independent publishers own their interpretation. The original evidence owner remains responsible for corrections, versioning, and retirement.

How often should a GEO RACI be updated?

Update it when roles, products, markets, vendors, controls, systems, methods, or operating scope change. A monthly ownership check and quarterly executive review can be useful illustrative cadences, but they are not universal. Maintain named owners, deputies, effective dates, and vacancy rules so the RACI survives organizational change.

When should an in-house team work with GeoZ?

Work with GeoZ when the team needs a connected measurement, diagnostic, execution, or Value as a Service layer and has internal owners for truth, approvals, access, analytics definitions, and investment. Use a narrower tool or internal approach when that is sufficient. GeoZ does not guarantee citations, rankings, traffic, leads, pipeline, or revenue.