Should You Run a Marketing Audit Before Switching Agencies? And How?

A marketing audit done right is a decision tool. It tells you whether the problem is the agency, the strategy, the measurement, or the business infrastructure behind marketing.

It is built for the buyer who is 60 days from switching agencies and not sure that is the answer. It also serves CFOs and finance leaders who own the budget line, and it holds up in private-equity diligence, where a marketing program that looks healthy at signing becomes an EBITDA hole six months later.

A full-stack audit runs seven layers: attribution, creative and landing pages, organic and AI visibility, martech, channel mix, team capability, and business infrastructure. Across 200+ audited accounts, roughly half underreport paid conversions, often by 50% or more. Switch without fixing that first and the next agency looks just like the last.

A marketing audit is a structured diagnostic that evaluates every layer of the system against a benchmark and identifies what’s broken, who owns the break, and what it would cost to fix. It is a forensic review of where spend goes, how it’s measured, and what the data says compared to what the platforms report. Treating an audit like a strategy deck, a list of quick wins, or a performance review is the most common reason agency-led “audits” produce no change.

A strategy review answers “what should we do next.” An audit answers “what is the ground truth right now” against the marketing goals already on the table. Agencies selling audits as a sales tool routinely conflate the two; the buyer ends up with a roadmap before knowing whether the foundation is sound. A real audit produces three artifacts: a ranked list of findings with financial impact, a clear read on whether the current agency caused each finding, and a switch-or-fix recommendation per layer. Anything less is a pitch document with a different cover.

The cleanest test for whether an audit is diagnostic or pitch-driven is what it does with the broken thing. A pitch audit hides behind “opportunities for improvement” and makes a new retainer the only logical conclusion. A diagnostic audit names the broken thing, names who broke it, and tells the reader what they can fix without hiring anyone. It provides value without the need for an engagement. That’s what a diagnostic-first approach looks like.

Quick takeaways:

• An audit measures ground truth; a strategy review picks next moves.
• Three artifacts: ranked findings with dollars, agency-fault read, switch-or-fix call per layer.
• Diagnostic vs pitch: one names the broken thing, the other sells the retainer.

Audit before switching whenever the cause of underperformance is genuinely unclear, the current agency owns the measurement infrastructure, or spend is about to step up materially. Three conditions in particular trigger an audit: the buyer cannot say with confidence whether the problem is the agency, the current agency owns the My Client Center (MCC) account or pixel implementation, or the business is about to scale spend two times or more inside the next six months. Skip the audit and switch directly when transparency has already been refused twice, the account has gone six months with no meaningful change-log activity, the current contract expires inside 30 days and a qualified replacement is in hand, or measurement hasn’t been rebuilt in 18 months or more.

Most buyers switch for the wrong reason. A bad quarter that turned out to be a measurement bug, a landing-page load-time regression that killed every channel, or a single product launch that mistargeted the audience routinely get blamed on the agency. Switching costs 60 to 90 days of lost momentum, knowledge handoff, and re-ramp; that cost makes an audit cheap at any price unless one of the trigger signals above is already present. An 18-month-stale tracking setup is itself an audit finding. If the agency owning it hasn’t raised or fixed it, the layer is theirs whether the audit happens or not.

Quick takeaways:

• Audit when the cause is unclear or the agency owns measurement.
• Switch directly only on transparency refusals, account autopilot, or expiring contracts.
• Stale tracking is on the incumbent, audit or no audit.

A full-stack marketing audit has seven layers. Most agencies audit only the marketing channels they sell, which is why a cross-layer audit catches leaks a channel-specific agency cannot see. In rank order of financial impact across accounts audited at scale:

1. Attribution and measurement. What the data actually says.
2. Creative and landing pages. Paid economics where they live.
3. Organic and AI visibility. Search traffic, traditional and AI.
4. Martech and data infrastructure. Tools, integrations, and signal flow.
5. Channel mix. Interaction effects and incrementality.
6. Team and agency capability fit. Who is in the seat day to day.
7. Business infrastructure. Sales, CRM, pricing, and offer testing.

Layer 1 is where the largest financial leaks usually surface, even though buyers expect the leak to be in creative or targeting. Layers interact: a broken attribution setup hides a landing-page problem that hides a channel-mix problem. The ranking shifts by business type. Direct-to-consumer (DTC) ecommerce leans on Layers 2, 5, and 4. Business-to-Business (B2B) Software-as-a-Service (SaaS) leans on Layers 1, 3, and 7.

Quick takeaways:

• Layer 1 carries the biggest dollar leaks; the buyer rarely expects that.
• Layers compound; channel-only audits miss cross-layer cause.
• Mix shifts by model: DTC on 2/5/4, B2B SaaS on 1/3/7.

Attribution is where the largest financial leaks hide and the layer that has to be cleaned first. If this layer is broken, every other audit finding is downstream noise. The audit checks conversion-action setup in Google Ads and Meta, conversion imports, server-side tracking and Conversion API (CAPI) coverage, Urchin Tracking Module (UTM) hygiene, Google Analytics 4 (GA4) channel attribution, and the gap between platform-reported and blended performance numbers. Across 200+ audited accounts, roughly half have shown underreporting of paid conversions, with the gap reaching 50% or more in many of them. Platform Return on Ad Spend (ROAS) frequently disagrees with P&L ROAS by a factor of ten or more in the same month for the same account; common pattern: platform ROAS at 1,140%, P&L ROAS at 77%, same month, same account.

The Apple App Tracking Transparency (ATT) framework is still the single largest source of platform underreporting. Many agencies never rebuilt tracking after ATT and compensated by expanding attribution windows instead of fixing the source data. A good agency volunteers the platform-vs-blended distinction; a pitch agency conflates them and leaves the wider gap unexplained.

Benchmarks matter inside this layer. Non-brand Google Ads ROAS holds at a 1.5x floor for ecommerce. For B2B, cost per qualified lead should be benchmarked against target customer acquisition cost (CAC) divided by expected close rate. In one B2B SaaS rebuild, cost per qualified lead (CPQL) fell from $512 to $103 once the conversion stack and attribution windows were corrected and offline conversion imports were wired through (covered in the full-funnel paid media program). The pattern across audited accounts is consistent: fixing measurement first changes which channels look profitable and which look broken, and frequently inverts the previous month’s prioritization.

Recommended audit checks:

• Conversion-action and import setup in every paid platform.
• Server-side tracking and CAPI coverage for the major channels.
• Platform ROAS vs blended ROAS variance, by month and account.

Many paid-media problems at scale are creative and landing-page problems, not audience problems. Meta cost per mille (CPM) is up 30 to 40% year over year across $30M+ a month managed across six channels, so the margin for weak creative has vanished. A landing page loading above 2.5 seconds on mobile kills roughly 10% of leads before the form is even seen. Inside an Opascope audit, creative is examined only as it serves the paid program; creative is not sold or audited as a standalone service line because the question that matters here is whether the creative is doing the job the spend needs it to do, not whether it is good in the abstract.

Creative checks look for at least three distinct hook angles per active campaign, a static-to-video ratio between 40/60 and 60/40, creator and whitelisted ads in the mix, and formats fitted to placement. Landing-page checks look for mobile load under 2.5 seconds on 4G, form length matched to offer, the first call to action above the fold, and Core Web Vitals green for Largest Contentful Paint (LCP), Interaction to Next Paint (INP), and Cumulative Layout Shift (CLS). Economic checks compare CAC against average order value (AOV) or lifetime value (LTV), CPQL against target, and channel-by-channel contribution margin. Advertorial pages frequently drop cost per acquisition (CPA) 30 to 40% versus branded product pages for DTC; the audit asks whether the account has tested them at all.

Recommended audit checks:

• Creative volume and variation per campaign.
• Mobile load and Core Web Vitals on top-spend landing pages.
• CAC, CPQL, and contribution margin against vertical benchmarks.

Organic search and AI search are now a single audit layer. 25.11% of Google queries trigger AI Overviews [1]. 85.7% of brands don’t appear in AI-generated recommendations in their own category, and GA4 misclassifies 70.6% of AI-referred traffic as Direct, with 87.4% of that referral traffic coming from ChatGPT [2]. Most marketing audits at the top of Google still skip this layer entirely. That omission is the single largest audit gap across the SERP.

Traditional SEO checks cover technical health, page speed, schema coverage, internal linking, backlink profile, content freshness, and organic keyword coverage. Answer Engine Optimization (AEO) checks ask whether the brand appears in ChatGPT, Perplexity, Claude, Gemini, Google AI Overview, and Copilot for the top ten commercial prompts in the category. The audit also verifies that the brand’s pages are crawlable by LLM crawlers (Cloudflare blocks many by default), and that meta descriptions carry the answer weight LLMs use for page selection. That’s the substance of an AI search visibility audit.

The traditional and AI checks overlap more than most teams expect, and recognizing the overlap is what allows one set of fixes to serve both. Content freshness is a ranking signal in classic SEO and a citation signal in AI engines; updating cornerstone pages to current dates and current claims lifts both surfaces at once. Structured data improves Google rich-result eligibility and gives LLMs the entity scaffolding they cite from. Site architecture and internal linking concentrate authority for traditional rankings and define the entity graph LLMs traverse during retrieval. A page that loads slowly is a problem for users and for crawlers, traditional and AI alike. The audit treats these overlapping items as one set of investments rather than two parallel programs.

AI traffic accounting is a separate gap. Most brands still report AI-referred traffic as Direct in GA4 because referrer headers are stripped at the LLM client. Until the audit separates AI traffic from true Direct, the growth curve is invisible. Peer-reviewed research from Princeton, Georgia Tech, and Allen AI shows that AEO-style optimization (the same approach the academic literature calls GEO, or Generative Engine Optimization) can boost visibility up to 40% in AI responses, with citations, statistics, and structured data carrying the most weight [3].

Recommended audit checks:

• Traditional health: speed, schema, linking, freshness, keyword coverage.
• AI visibility: presence across ChatGPT, Perplexity, Claude, Gemini, Google AI Overview, Copilot.
• Overlap: freshness, structured data, architecture, page speed serve both surfaces.

The martech stack is where audits find the most overlap, the most unused licenses, and the most tracking that never finished implementation. The overlap is not redundancy in the catalog sense; it is duplication of function across tools that nobody has rationalized. Two analytics tools tracking the same events on the same site, two tag managers firing the same pixels, a customer data platform (CDP) doing identity resolution that the customer relationship management (CRM) is also doing on a different schema. On average, half the tools in a mid-market stack do not feed into any decision anyone reads on a weekly basis. The audit question is not “which tools do you have.” It is “which tools does anyone read output from.”

Tool inventories include CRM, analytics, tag manager via the Google Tag Manager (GTM) container, CDP, email, attribution, data warehouse, dashboarding, session replay, A/B testing, and marketing automation. Integration health asks whether platform events are flowing into the CRM, whether offline conversions are flowing back to the ad platforms, and whether lead scores are consumed downstream by a sales process. Unused licenses surface fast: 20 to 30% of mid-market martech spend typically does not survive a rigorous audit, and the savings frequently fund the rebuild work that fixes Layers 1 and 5.

Recommended audit checks:

• Tool inventory check: every active license, owner, and weekly read.
• Integration health check: platform-to-CRM, offline conversions back, lead scores downstream.
• Unused license check: which seats and tools survive a six-week usage audit.

Marketing channels are an interconnected system, not a set of isolated reports. Turn off LinkedIn ads and branded Google search drops. Cut TikTok and creative fatigue accelerates on Meta. Channel-mix decisions made 18 months ago are usually wrong by now because audience behavior, platform CPMs, and creative supply have all shifted. The audit question is what happens to the system when one input changes, not what the ROAS of an isolated channel is on the platform-reported dashboard.

Incrementality testing is the test the layer turns on. If the agency has never run a geo holdout or a channel-pause test, the mix is theoretical and they make mix decisions on platform-reported numbers alone. Top-of-funnel versus bottom-of-funnel balance also belongs in this layer: a 98% prospecting and 2% retargeting split is wrong for high-AOV, high-consideration products, and 20 to 40% retargeting is closer to right at that level. Emerging channels, including connected TV, AppLovin, and TikTok Shop, sit inside the same audit lens. The audit checks whether the account has incrementality-tested the ones that apply rather than added them by inertia.

Recommended audit checks:

• Incrementality testing history: geo holdouts and channel-pause tests run in the last 12 months.
• Prospecting vs retargeting split against AOV and consideration cycle.
• Emerging-channel decisions: incrementality-tested or added by inertia.

The team layer is where the switch-agencies conversation usually settles. The audit checks who is actually in the account day to day, how many accounts they manage, whether seniority of the people who pitched matches seniority of the people who work on the account day by day, and what change-log depth looks like in a comparable account. Below 20 changes a month in an account is de facto on autopilot. An account-to-analyst ratio above 8:1 is the working quality line; above that, the person in the seat does not have the hours to do the work the audit is asking about.

Cross-channel work requires senior coverage that a single-channel agency cannot provide on its own. A pay-per-click-only agency cannot run the Layer 5 channel-mix audit because the data they have is the data the audit is questioning. Cross-layer work requires senior marketers on the account who span paid media, organic search, content marketing, and analytics, and the audit verifies that this seniority is in the seat rather than only on the pitch deck.

Recommended audit checks:

• Account-to-analyst ratio under 8:1.
• Change-log depth above 20 meaningful changes per month.
• Seniority match between pitch and day-to-day seat.

Half the problems blamed on marketing agencies are business-infrastructure problems. Dormant lead nurture, CRM hygiene, sales handoff, pricing logic, and offer testing all sit outside the marketing function but inside the number marketing is being measured on. If marketing drives leads into a sales process that converts at 2% when the benchmark is 15%, no agency switch will fix the gap.

Dormant leads are the most common single finding in this layer. In one audit of an account-based-marketing (ABM) program, 63 enterprise deals in year one came from leads that had been sitting untouched in the CRM for 12 to 24 months. Marketing had done its job by sourcing and qualifying those leads; the leads were never picked up by sales. Once the dormant-lead motion was activated, marketing’s contribution rose to 70% or more of new business monthly against a 30% SaaS benchmark. Built-from-scratch ABM programs routinely outperform their inbound counterparts on the same logic. One generated 10,000 qualified leads and $3.65M closed-won in 90 days, with landing pages converting at 21%. When the audit flags pricing, packaging, or offer issues, responsibility doesn’t shift to marketing. It means an absence of testing here is producing underperformance the business is currently blaming on marketing.

Recommended audit checks:

• Dormant-lead inventory: deals untouched 12 to 24 months and the nurture motion behind them.
• Sales handoff: marketing-qualified-to-sales-qualified pickup rate.
• Offer testing cadence: pricing, packaging, and offer experiments in the last 12 months.

After running the seven layers, sort every finding into one of three buckets: agency caused this, the business caused this, or measurement hid this. Count the agency-caused findings. Fewer than three meaningful findings means the switch will not solve the problem. More than five and the switch is the right call. Anything between needs a conversation with the current agency about a specific, time-boxed fix.

The common misattribution is treating a measurement problem as an agency problem. A serious incoming agency rebuilds measurement as part of onboarding, and the picture clears inside 60 to 90 days. The trap is switching to another agency running the same broken measurement or attribution playbook, in which case the picture never clears. Even when the leak is “just” measurement, an incumbent agency that hasn’t rebuilt tracking in 18 months or more is responsible for the gap. The true-positive case for switching is rarely a single bad month: it is long change-log droughts, refusal to share account data, stale creative without testing, no channel-mix opinion, and no experimentation calendar.

Time-box the fix when the buckets are mixed. If the current agency is given a measurable 30 to 60 day list and misses it, the switch decision is made for them. If they hit it, the audit findings carry into the renewed engagement instead of into a transition.

To put the framework on a specific account, book an audit. The seven-layer review takes two to six weeks and ends in a written findings document with a switch-or-fix recommendation per layer.

Quick takeaways:

• Sort findings into agency, business, or measurement caused.
• Under three agency-caused findings: do not switch.
• Five or more: switch. In between: time-box a measurable fix list.

Q: What is a marketing audit and what should it include? A marketing audit is a structured diagnostic that evaluates every layer of the marketing system against a benchmark and produces a ranked list of findings with financial impact. The full-stack version covers seven layers: attribution and measurement, creative and landing pages, organic and AI visibility, martech and data infrastructure, channel mix, team capability, and business infrastructure outside marketing. Single-layer audits catch only the leaks inside that channel. Done right, a full-stack audit tells the business whether the problem is the agency, the measurement, or the business itself.

Q: How do I know if my marketing agency is underperforming? Underperformance shows up across several signals, and no single one is enough on its own. Common examples include a wide gap between platform ROAS and P&L ROAS in the same month, missing or stale change-log activity for the channels the agency owns (the Google Ads change log is one specific example for paid search agencies; an SEO agency’s content cadence and indexation report is the equivalent), a refusal to volunteer attribution gaps, no incrementality testing in the last 12 months, no creative or landing-page testing calendar, and no cross-channel point of view. None of these alone proves underperformance. Three or more in combination usually do, and the audit is what separates a measurement problem from an agency problem.

Q: Should I audit my marketing before switching agencies? Yes, in almost every case. Switching agencies costs 60 to 90 days of momentum, knowledge handoff, and re-ramp. An audit costs days or weeks and reveals whether the problem is even the agency. The common pattern: the audit surfaces underreporting of paid conversions in roughly half of accounts examined, which made the current agency look weaker than it actually was. Skip the audit only when transparency has already been refused twice, the account has been on autopilot for six months, or a qualified replacement is already in hand and the contract expires inside 30 days.

Q: How much does a marketing audit cost and how long does it take? Good audits run from free to roughly $15,000 depending on scope, take two to six weeks, and involve one to two senior marketers. Free audits can be genuinely diagnostic when run by teams that also operate accounts day to day. A $5,000 to $15,000 paid audit typically covers a full seven-layer review with a written findings document and a switch-or-fix recommendation. Anything longer than six weeks usually indicates junior staffing on the work. Anything shorter than a week usually indicates a pitch audit that skipped layers.

Q: Is AI search visibility part of a marketing audit? It belongs in every modern audit. 25.11% of Google searches now trigger AI Overviews [1]. AI-referred traffic converts at roughly 11x the rate of organic search for sign-ups [2], since prompt-driven referrals arrive with intent already formed. 85.7% of brands are invisible when AI tools make recommendations in their category, and GA4 misclassifies 70.6% of AI-referred traffic as Direct because most LLM clients strip the referrer header [2]. The audit checks whether the brand appears in classic Google search and across ChatGPT, Perplexity, Claude, Gemini, Google AI Overview, and Copilot for the top ten commercial prompts, and whether the site is crawlable by LLM crawlers.

Q: What’s the biggest mistake buyers make during a pre-switch audit? Confusing a measurement problem with an agency problem. The single most common audit finding across paid attribution is significant underreporting, with the gap frequently reaching 50% or more in the affected accounts. Buyers then read the underreported dashboard as agency underperformance and switch. If the new agency rebuilds tracking on day one, the picture clears. If they inherit the broken setup, the cycle repeats. An incumbent agency that hasn’t rebuilt measurement in 18 months or more is accountable for the gap.

Q: What’s the difference between a marketing audit and a SWOT analysis? A SWOT analysis lists strengths, weaknesses, opportunities, and threats at a high level. A marketing audit verifies what’s actually true in the data: whether conversion tracking is correct, whether landing pages convert at benchmark, whether the channel mix is incremental. A SWOT belongs inside an audit’s findings document; it does not replace one. A marketing audit produces numbers; a SWOT produces a list.

Q: What does a marketing audit action plan actually look like? The action plan is a ranked list of fixes with owner, deadline, and expected dollar impact next to each item. A typical plan opens with attribution rebuild (four to six weeks, owned by the next agency or analytics lead), followed by landing-page speed and creative testing fixes (two to four weeks), AI visibility and SEO content updates (four to eight weeks), and martech consolidation (four to twelve weeks). Items without a named owner and a date are not part of the plan; they are a wishlist.

Q: How does a marketing audit fit into the broader marketing strategy and marketing plan? The audit feeds the marketing strategy by replacing assumptions with measured ground truth across marketing channels. The strategy then sets the next 12 months of priorities, and the marketing plan translates those priorities into quarterly campaigns, budgets, and team assignments. Without the audit, the strategy is built on platform-reported numbers that routinely overstate paid performance by meaningful margins, which is one of the reasons three-year strategies built without audits often unwind inside 90 days.

1. Conductor 2026 AI Overview Study, 21.9M queries analyzed. https://www.conductor.com/learning-center/ai-overview-study/
2. Loamly State of AI Traffic 2026, 2,014 companies and 446K visits analyzed. https://loamly.com/state-of-ai-traffic
3. Pradeep Dasigi et al., “GEO: Generative Engine Optimization,” KDD 2024 (Princeton, Georgia Tech, Allen AI). https://arxiv.org/abs/2311.09735
4. Google Ads Help Center, change history and offline conversion imports. https://support.google.com/google-ads/answer/2998031