Agency Playbook: Multi-Client Taxonomy Governance

An in-house marketer manages one taxonomy. An agency account manager switches between five clients in a single morning, each with its own approved source values, its own campaign naming pattern, its own analytics platform, and its own opinions about naming conventions.

Every governance problem in this guide, an agency lives at multiples.

This page is the architecture that makes that sustainable: a base taxonomy, workspace isolation, a half-day onboarding, and cross-client reporting. Build it once and a new client comes under governance in hours, with performance benchmarks that run across your whole portfolio.

Why Multi-Client Governance Is Hard

The failure modes are unique to agencies:

Cross-contamination. A source value from Client A’s taxonomy gets accidentally used for Client B. Client B’s analytics shows phantom traffic from a source they have no relationship with, and their first question is why you didn’t catch it. (Chapter 8 tells the full story of one such incident.)
Convention drift. Without centralized standards, each account team grows its own conventions, and the agency’s “taxonomy” becomes five incompatible systems that can’t be compared, audited, or maintained.
Onboarding friction. Every new client means building a taxonomy from scratch. If that takes two weeks to design and document, that’s two weeks of links created without governance.
Turnover vulnerability. When an account manager leaves, their clients’ taxonomy knowledge leaves with them, unless it lives in a system instead of a person’s head.

The Agency Taxonomy Architecture

The model that works: a standardized agency base taxonomy with client-specific customizations layered on top. Consistency where it matters (parameter structure, medium values, naming conventions), flexibility where clients genuinely differ (source values, campaign naming, custom dimensions).

Agency Base Taxonomy (shared across all clients)

utm_medium:: controlled vocabulary (cpc, paid_social, social, email, display, affiliate, referral, video, offline)
Naming rules:: lowercase, underscores within, hyphens between, no spaces
Campaign pattern:: [initiative]-[product]-[date]
Forbidden patterns:: no PII, no internal codes

Client A Customization

utm_source: google, facebook, linkedin, newsletter, partner_acme

Products: platform, api, enterprise

Regions: us, emea, apac

Client B Customization

utm_source: google, facebook, instagram, tiktok, email_blast

Products: shoes, apparel, accessories

Regions: us, uk, de

Client C Customization

utm_source: google, bing, linkedin, webinar_platform

Products: saas_core, saas_enterprise

Regions: (none: single market)

What the base taxonomy standardizes:

Element	Agency standard	Why
`utm_medium` values	Controlled vocabulary aligned with GA4 channel groups	Medium values are universal: `cpc` means the same thing for every client
Naming format	Lowercase, underscores within segments, hyphens between	One format means agency-wide audits work and format errors stop happening
Campaign pattern	`[initiative]-[product]-[date]`	A standardized structure is what makes cross-client reporting on campaign types possible
Forbidden values	No PII, no internal abbreviations, no platform-specific codes	Protects clients and prevents unintelligible values

What each client customizes:

Element	Client-specific	Why
`utm_source` values	Varies by client’s channel mix	Client B uses TikTok; Client A doesn’t. Client C has a custom webinar platform.
Product/brand dimensions	Varies by client’s offerings	Each client’s product taxonomy is unique
Regional dimensions	Varies by client’s market scope	Some clients are single-market; others are global
Custom parameters	Varies by client’s analytics setup	Some clients use `cid`, `utm_id`, or custom dimensions

Client A and Client B from the diagram above are both running below. Read them as a pair. The medium vocabulary and the [initiative]-[product]-[date] pattern are identical because the base taxonomy says so. The sources and products are different because the clients are. That split is the whole architecture, and it is easier to see in two builders than in any table.

Client A: B2B SaaS.

Live demoOpen full view

Loading the live builder…

Client B: ecommerce.

Live demoOpen full view

Loading the live builder…

Client Onboarding: The 4-Step Process

A standardized onboarding process turns a two-week taxonomy design project into a half-day configuration exercise.

Step 1: Discovery questionnaire (30 minutes).

Before the first strategy meeting, send the client a short questionnaire:

What analytics platform(s) do you use? (GA4, Adobe, Matomo, etc.)
What channels are you actively running? (Paid search, paid social, email, display, affiliate, organic social, events, partnerships?)
What are your product lines or brand segments?
What regions/markets do you operate in?
Do you have existing UTM conventions? (If yes, get a sample of current tagged URLs.)
Who on your team creates UTM-tagged links?
Do you have a CRM or marketing automation platform that captures UTM data?

Step 2: Taxonomy configuration (2–3 hours).

Using the discovery answers:

Fork the agency base taxonomy into a client-specific workspace.
Populate the utm_source approved values based on the client’s active channels.
Add product and region values if the client’s campaigns need those dimensions.
Review any existing UTM conventions. If the client has existing values in their analytics, map legacy values to the new taxonomy (see Chapter 5’s “When You Inherit a Mess” section).
Configure cascading dropdowns: when utm_medium = email, only email-relevant sources appear.

Step 3: Documentation (1 hour).

Produce a client-specific taxonomy document containing:

Approved values for each parameter (with definitions)
Campaign naming pattern with 3–5 examples using the client’s actual campaigns
Rules for requesting new values (who to ask, turnaround time)
Who owns the taxonomy (agency-side and client-side contacts)

This document lives in the client’s shared workspace, not in the account manager’s head.

Step 4: Handoff and training (30 minutes).

Walk the client’s team through the taxonomy, the link builder, and the “how to request a new value” process. If the client’s team creates links (not just the agency), give them access to the builder with their workspace pre-selected. If the client’s team does not create links, make that explicit: all link creation goes through the agency.

Total onboarding time: ~4–5 hours. Without a standardized base taxonomy, budget 2+ weeks for the same result.

Workspace Isolation

Every client gets their own workspace. This is a structural requirement, and it is the thing that prevents cross-contamination.

What workspace isolation provides:

Separate approved value lists. Client A’s sources can’t appear in Client B’s dropdowns.
Separate link history. Each client’s generated links stay isolated and searchable within their workspace.
Separate user permissions. Client-side users see only their workspace. Agency users switch between workspaces.
Separate audit trails. Who created what, when, for which client.

The master taxonomy relationship:

Each workspace inherits from the agency base taxonomy and can only narrow it, never widen it. If the base taxonomy defines utm_medium values as cpc, paid_social, social, email, display, affiliate, referral, video, offline, a client workspace can remove affiliate (they don’t run it) but can’t add paid_search (a value the base doesn’t define). Consistency across clients survives; per-client customization stays possible.

Cross-Client Reporting

The payoff for all that discipline: cross-client performance benchmarking.

Because the base taxonomy standardizes utm_medium values and campaign naming patterns across all clients, you can answer questions like:

“What’s the average CPC channel conversion rate across our portfolio?”
“Which clients get the best email engagement relative to list size?”
“How does paid social ROAS compare across our e-commerce clients vs. our B2B clients?”

The prerequisite: this only works if the base taxonomy is actually enforced. The moment one client uses social-paid instead of paid_social, cross-client channel grouping breaks. That’s why enforcement is non-negotiable: every benchmark you show a client stands on it.

Cross-client reporting model:

SELECT
    c.client_name,
    s.channel_group,
    SUM(s.sessions)      AS total_sessions,
    SUM(s.conversions)   AS total_conversions,
    SUM(s.revenue)       AS total_revenue,
    SUM(s.revenue) / NULLIF(SUM(s.spend), 0) AS roas
FROM fact_sessions s
JOIN dim_client c ON s.client_id = c.client_id
WHERE s.date BETWEEN '2025-01-01' AND '2025-03-31'
GROUP BY c.client_name, s.channel_group
ORDER BY c.client_name, total_revenue DESC;

Privacy note: Share cross-client data in aggregate only. Never show one client another client’s specific campaigns, spend levels, or conversion rates. Benchmarks stay anonymized (“our e-commerce portfolio averages X”), never attributed.

Handling Client-Specific Analytics Platforms

Not every client uses GA4. The agency base taxonomy stays platform-agnostic; the implementation varies:

Client platform	Key differences	Adjustments needed
GA4	Auto-tagging overrides manual UTMs for Google Ads. Channel classification runs on the source/medium pair, not the medium alone.	Ensure `utm_medium` uses GA4-recognized values. Govern Google Ads campaign naming (Ch 7b).
Adobe Analytics	Uses `cid` parameter and classification rules. UTMs get captured only once you configure them (campaign variable or processing rules).	May need to maintain both UTM parameters and `cid` on the same links.
Matomo	Reads standard UTM parameters natively. On-premise option for privacy-sensitive clients.	No special adjustments: standard UTMs work.
Mixpanel / Amplitude	Event-based; UTMs land as user properties, with first-touch values kept alongside the latest touch.	Ensure UTM capture is configured in the client’s Mixpanel/Amplitude setup.
Multiple platforms	Client uses GA4 for web analytics and HubSpot for marketing automation.	Both read UTMs independently, so governed links work for both. Ensure field mappings in HubSpot match the UTM parameter names.

The base taxonomy’s utm_medium values (aligned with GA4 channel groups) work well across most platforms. Where a client uses a non-GA4 platform exclusively, you may relax the GA4-specific medium values, but only inside that client’s workspace.

Scaling Agency Governance: Common Patterns

Pattern 1: Centralized link creation (agency creates all links).

The agency team creates every UTM-tagged link and delivers them to the client as part of the campaign brief or ad setup. The client never touches the link builder.

Best for: Clients without marketing ops staff. Smaller engagements. Clients who don’t want to learn another tool.

Risk: Bottleneck if the agency team is slow to turn around link requests.

Pattern 2: Guided self-service (client creates links in governed builder).

The client’s team has access to the builder with their workspace pre-selected. Dropdowns enforce the taxonomy. The client creates links without agency involvement.

Best for: Clients with marketing ops staff who run their own email, social, or content campaigns. High-volume accounts where the agency can’t create every link.

Risk: Client users may create links for campaigns the agency doesn’t know about, leaving tracking gaps in reporting.

Pattern 3: Hybrid (agency creates paid media links, client creates owned media links).

The agency handles all paid campaign UTMs, where spend is involved and accuracy is critical. The client handles owned media (email, blog, social) using the governed builder. Both work from the same taxonomy.

Best for: Most mid-to-large agency engagements. Clear division of responsibility. Both teams contribute links without stepping on each other.

The Agency Quarterly Review

Build taxonomy governance into your quarterly client review. Not as a separate agenda item; as part of the performance story.

What to include:

Naming health score (from Ch 5’s audit framework). One number for the quarter’s taxonomy compliance. The trend line shows improvement or decay.
Fragmentation rate. Unique values per parameter divided by true entities. Give the client the number: “We had 12 unique sources in Q1, matching our 12 actual sources. Zero fragmentation.”
Attribution coverage. What percentage of the client’s conversions carry complete UTM attribution? Rising coverage means the tracking is working.
One governance win. A specific insight or budget decision that clean data made possible: “Your podcast sponsorship drove 3x more demo requests than LinkedIn content, a signal that was invisible before we started tagging podcast links.”

Presented this way, taxonomy governance reads as a capability the agency provides, not overhead the client tolerates. It justifies your fees, and it builds trust.

Competitive Advantage: Why This Matters for Agencies

Good taxonomy governance is an underrated competitive moat for agencies:

Faster onboarding. A standardized base taxonomy means new clients are operational in hours, not weeks.
Fewer errors. Workspace isolation and controlled vocabularies prevent the cross-contamination and naming drift that erode client confidence.
Better reporting. Clean data means dashboards that tell clear stories without asterisks, caveats, and manual cleanup.
Cross-client insights. Portfolio benchmarks give clients intelligence they can’t get from their in-house team, or from any agency running on ungoverned data.
Reduced churn. Clients who trust the numbers in your reports, because the numbers are consistently clean, don’t spend renewal season questioning your value.

The agency that can say “We govern campaign taxonomy across all our clients, which means your data is clean from day one and comparable across channels” has a materially stronger pitch than the one still typing UTMs into URL bars by hand.

Key Points

One standardized agency base taxonomy, customized per client: that’s what buys fast onboarding, cross-client benchmarking, and consistent quality

Workspace isolation is a structural requirement; it is what stands between Client A’s taxonomy and Client B’s data

Standardize onboarding into four steps (discovery questionnaire, taxonomy configuration, documentation, handoff) and a new client takes 4–5 hours, not two weeks

Put taxonomy metrics (naming health score, fragmentation rate, attribution coverage) in every quarterly review, and position governance as a capability, not overhead

Cross-client reporting is the most valuable thing clean agency data buys you, and it works only when the base taxonomy is enforced for every client

Action Item: No base taxonomy template at your agency? Create one this week. Start with the utm_medium controlled vocabulary (it’s the same for every client), add your standard campaign naming pattern, and fork it for your three highest-value clients. The next new client onboards in half the time.