What You Actually Get From a Google Ads Agency in 2026: The Full Scope, the Real ROAS, and Why It Costs What It Costs

Hiring a Google Ads agency in 2026 does not buy you a campaign. It buys you a measurement system, a feed pipeline, a creative engine, an AI-assisted optimization layer, and the human judgment to point all four at your unit economics. The dashboard is the smallest visible surface of the work. Ninety percent of what you actually pay for sits underneath it.

Most agencies still sell 2019 deliverables. A keyword list. A weekly meeting. A monthly PDF. The actual job has moved on. Performance Max accounts for two thirds of Shopping spend. AI Max for Search went generally available in April. Consent Mode v2 has been mandatory in the EU since March 2024. Enhanced Conversions, server-side tagging, and a clean product feed are the load-bearing layer that decides whether Smart Bidding has the data it needs to compound or collapses inside the learning phase. If your agency is not building those layers first, you are paying retail for a 2019 service while the platform charges 2026 prices.

APEX DIGITAL has run Google Ads for online stores and lead-generation businesses since 2015. Over a decade on the platform, more than EUR 100 million in client revenue moved through Google and TikTok Ads, twelve countries served, and Google Partner status held continuously since 2021. This post is the full inventory of what we deliver, what it costs, and why the contribution-margin ROAS we hold for most accounts sits above 6:1 even when the platform-reported number reads closer to the 4:1 industry average.

What you are actually paying for

A Google Ads agency in 2026 sells six things, only one of which shows up in the platform UI. The first is data infrastructure: a custom data layer, server-side tagging where the spend justifies it, Enhanced Conversions, and Consent Mode v2 so the data the algorithm sees is accurate and durable. The second is account architecture: brand defense separated from generic, Performance Max split by margin tier, Standard Shopping reserved for sniper queries, Demand Gen for upper-funnel YouTube and Shoppable CTV. The third is asset production: copy and creative engineered for the auction, refreshed on a cadence the algorithm rewards. The fourth is continuous optimization: weekly n-gram analysis, daily anomaly detection, monthly asset refreshes, quarterly bidding-strategy reviews. The fifth is compliance and measurement: GA4 attribution alignment, Customer Match upload via the Data Manager API, policy and feed disapproval triage. The sixth and most under-priced is business judgment: when to push spend, when to pull back, when the channel is wrong for the goal.

Tinuiti's Q1 2026 cross-industry benchmark shows advertisers using Target ROAS Smart Bidding achieve 38 percent higher return on ad spend compared to manual CPC. The lift is not in the bidding strategy itself. It is in everything you must put underneath it for the strategy to actually work: clean conversion data, audience signals, asset diversity, and segmented product groups. An agency that ships the bidding strategy without shipping the foundation under it is delivering a Ferrari engine on a chassis that cannot hold it. Our Google Ads PPC management service starts at EUR 500 per month and delivers all six layers, not just the dashboard.

The custom data layer: where every good agency starts

A data layer is the structured event payload your site emits when something meaningful happens. A product view fires view_item with item_id, price, and item_brand. An add-to-cart fires add_to_cart. A purchase fires purchase with the full transaction object. The Google Tag Manager default templates handle the basic case. They do not handle yours.

A custom data layer is one engineered for the specific business model. An e-commerce store needs item-level fields that Smart Bidding can use as audience and bidding signals: margin tier, inventory health, customer segment (new vs returning), order rank (first order, second order, loyalty cohort). A lead-generation business needs lead quality scores, deal-stage progression, and offline-conversion identifiers so closed-won deals can be imported back into Google Ads. A SaaS business needs trial activation, paid conversion, and contraction events. The same generic GTM template will misfire on all three.

The reason this matters: a generic browser-only setup loses 15 to 40 percent of conversions to ad blockers, iOS Intelligent Tracking Prevention, browser cookie restrictions, and consent-revoked sessions. Those losses are not random. They concentrate on your highest-intent buyers, who are disproportionately likely to use privacy tools. When Smart Bidding sees fewer conversions from those users, it bids less aggressively to reach them, and the platform spirals into a worse equilibrium. Server-side tagging and a custom data layer are the cleanest way to recover the data Google needs to bid correctly. As Simo Ahava's reference work on server-side tagging documents, this is now the practitioner consensus, not a power-user experiment.

We deploy the data layer in week one of every engagement. Without it, every other lever you pull is working with degraded signal. With it, the ground floor is solid and the rest of the work compounds. Google's official server-side tagging documentation covers the setup mechanics. The judgment of what to capture, when to gate it, and how to reconcile it across platforms is the part you hire an agency for.

Conversion tracking that survives 2026

Conversion tracking in 2026 is not the click-and-thank-you-page model from 2019. The four layers we implement on every account are Enhanced Conversions, offline conversion imports, Google Tag Gateway, and a GA4 attribution model aligned with the Google Ads attribution model. Skip any of them and Smart Bidding starts learning from a partial picture, which is worse than learning from no picture at all because the bias is invisible.

Enhanced Conversions sends hashed first-party data (email, phone, name, address) from the conversion page back to Google, where it is matched against logged-in user signals to recover conversions the cookie-based default would lose. In June 2026 Google collapsed Enhanced Conversions to a single on-off toggle for new accounts, removing the method-selection screen. Web and Leads converged into one switch. This is now the default, not a power-user upgrade. Accounts without it are running on degraded signal by definition.

Offline conversion imports are how lead-generation businesses close the loop between a Google-Ads-attributed click and a closed-won deal that materialized 30 to 90 days later in their CRM. Without them, the algorithm optimizes for form fills (a leading indicator) rather than revenue (the actual goal). Google now explicitly recommends new builds use the Data Manager API for offline conversions, not the legacy Google Ads API offline endpoint, with the official documentation stating they do not recommend implementing new offline conversion workflows using the Google Ads API. We migrate accounts to the Data Manager API as part of standard onboarding.

Google Tag Gateway (formerly First-Party Mode) routes Google tag traffic through your first-party domain before relaying it to Google. Average data signal uplift after enabling Google Tag Gateway is 11 percent, with some accounts reporting 10 to 15 percent conversion lift purely from more conversion hits actually landing. Cloudflare ships a native integration. The setup is a one-click toggle in Google Tag Manager once the underlying domain is configured correctly. Server-side GTM is the heavier-lift option for accounts where the spend justifies a dedicated server container.

QA cadence is the part that separates an agency from a freelancer. We test the full conversion stack the first seven days post-launch, then weekly, with side-by-side counts in Google Ads, GA4, the data layer payload, and the source-of-truth (Shopify, WooCommerce, the CRM). Any variance over 2 percent triggers a root-cause review. Our conversion tracking and analytics service builds this stack as a one-off implementation for clients running ads in-house.

Consent Mode v2: the EU compliance line nobody crosses cleanly

Consent Mode v2 has been mandatory since March 6, 2024 for any site running Google ad products and targeting end users in the European Economic Area, which includes Romania. Google's official Consent Mode v2 documentation states that "advertisers receiving data from EEA end users must collect consent and share consent signals with Google to use tags and SDKs for measurement, ad personalization, and remarketing features." That is not a recommendation. It is the policy.

The two new signals introduced by v2, ad_user_data and ad_personalization, sit alongside the original ad_storage and analytics_storage. All four must flow from the same source of truth, typically a Google-certified Consent Management Platform. The mistake we see most often on audits is a CMP that fires ad_storage correctly but forgets analytics_storage, which causes GA4 data to silently drift away from Google Ads data. Attribution starts lying. The drift is invisible until the quarterly reconciliation, by which point three months of bidding decisions have been made on bad data.

The two implementation modes are Basic and Advanced. Basic blocks tags entirely until consent is granted and produces no measurement at all from non-consenting users. Advanced loads tags in a cookieless mode that sends pings without identifiers, and Google models the conversions for you. Google strongly recommends Advanced. We deploy Advanced on every account because it preserves measurement on the 30 to 50 percent of EU users who decline cookies. Without Advanced Consent Mode you cannot do remarketing, conversion modeling, audience building, or Enhanced Conversions for EU traffic. The accounts that are still running v1, or running v2 misconfigured, are the accounts that quietly underperform on EU spend month after month and never figure out why.

Keyword optimization in the AI Max era

Match types in 2026 are not what they were in 2022. Google has aggressively redefined exact, phrase, and broad to interpret intent across longer queries and semantic neighbours. Exact match [emergency plumber] now triggers on "water heater burst need help." Phrase match has absorbed most of what broad-match-modifier used to do. Broad match interprets entire query intent. AI Max for Search took this further in April 2026 and made search-term matching a campaign-level setting rather than a keyword-level setting. Google's match types documentation covers the current behaviour.

The practical implication: negative keywords are now your only real intent control. The accounts that scale in 2026 are the accounts with disciplined, weekly negative keyword maintenance. The accounts that bleed are the ones that set up a campaign in 2022 and have not touched negatives since. Google raised the PMax negative keyword limit from 100 to 10,000 per campaign in 2025, and account-level negative lists now support 5,000 entries. The headroom is there. The discipline to use it is the part you hire an agency for.

The scalable approach is n-gram analysis. An account with a million unique search terms typically has only 30,000 to 50,000 unique n-grams (one to three word fragments). Sort those by CPA, ROAS, and conversion rate and you find the waste in minutes rather than the days it takes to scrub a raw search-term report. Adalysis has a clear walkthrough of the methodology. We run n-gram analysis weekly on every account, feed the negatives into both campaign-level and account-level lists, and document the rationale so the next analyst can read the history.

On AI Max specifically: brand exclusions at the campaign level are now the cleaner way to redirect brand traffic to a dedicated brand campaign. Brand suitability controls give you up to 25 term exclusions and 40 messaging restrictions per campaign so the AI does not generate ad copy you would never approve. Set those before you flip the switch, not after.

Product segmentation that beats the algorithm

Performance Max is a black box that rewards structure. The single highest-impact restructure we make on most e-commerce accounts is splitting PMax retail campaigns by margin tier. Most accounts run a single PMax campaign with the full catalog at one Target ROAS. The algorithm then spends aggressively on whatever converts at the highest dollar value, which on most catalogs is the lowest margin SKUs. Blended profitability collapses while platform ROAS looks fine on paper.

The fix is custom labels in the feed. We populate custom_label_0 with margin tier (high, mid, low), custom_label_1 with seasonality, custom_label_2 with inventory health, and custom_label_3 with launch status. Then we build separate PMax campaigns by listing-group filter. High-margin SKUs run at a higher tROAS (say 500 percent) with the Customer Acquisition Goal enabled and "bid higher for new customers" turned on. The catalog-tier campaign covers the rest of the SKUs at a more conservative tROAS (around 350 percent). The algorithm now optimizes inside two envelopes that match the actual unit economics, not against a single average that masks both.

Asset groups are the second segmentation layer. Themed by category, not dumped together. A clothing retailer running shoes, accessories, and outerwear in one asset group is starving each category of the creative diversity it needs to perform. Three asset groups, each with their own headlines, descriptions, images, and lifestyle video, deliver materially better results than one mega-group. Listing groups inside each asset group split by intent so the algorithm has the segmentation it needs to bid differently across the funnel.

The patience point matters: changes greater than 20 percent on budget, any switch on bid strategy, and any addition of a new asset group reset the learning phase. We sequence changes to never trigger more than one reset at a time, and we hold tROAS targets stable for at least two weeks before pushing higher. Our Google Ads for e-commerce service builds this segmentation into the account on day one rather than retrofitting it after the algorithm has cemented bad habits.

Feed engineering: the lever that still moves the most

Every audit we run on a paid e-commerce account in 2026 starts with the feed. Bidding strategies and campaign structure multiply the feed. A perfect Smart Bidding strategy on a 60 percent complete feed delivers worse results than a default Maximize Conversion Value strategy on a 99 percent complete feed. The math does not have a workaround.

The under-used title formula is still Brand + Product Type + Key Features + Size or Color, kept under 150 characters. Google's official product data specification documents the field requirements. Optimized titles produce a 15 to 30 percent increase in impressions and a 10 to 20 percent CTR lift versus vendor-default titles. Stop using the manufacturer's marketing title. Write a title for the buyer who is searching, not for the brand sheet.

GTINs are the second-biggest lever. Products with correct GTINs see a 20 to 40 percent performance bump compared to identical products missing them, because Google can match the listing to its product knowledge graph. Aim for 90 percent or higher catalog coverage. For handmade, custom, or generic items where a GTIN does not exist, set identifier_exists = false on the feed to avoid disapprovals. Do not invent UPCs. That is the fastest way to get an account suspended.

Supplemental feeds are how we run promotional pricing, sale labels, and seasonal attributes without touching the primary feed. The primary feed stays clean and stable. The supplemental feed overrides specific fields on a window. When the promo ends, the supplemental feed is removed and the catalog returns to baseline without a feed regeneration. This is the difference between an agency that handles your feed and a feed that handles your agency.

The most under-discussed concept is the Golden Record store: catalogs at 99.9 percent attribute completion. These stores are seeing three to four times higher visibility inside AI Overviews and AI Mode than catalogs at the 70 to 80 percent baseline. Feed completeness alone expands impression share by 40 to 60 percent with no bid changes. The marginal cost of the last 10 percent of attributes is low. The marginal value is enormous. We run a feed-completeness audit monthly and a structural rebuild quarterly.

Copy and creative: the half nobody itemises

Look at most agency proposals. Copywriting and creative are not on the line items. The implicit assumption is that the account manager will write the headlines between bid adjustments, or the client will provide copy, or the creative will be whatever stock asset gets dragged into the asset group. This is why most PMax accounts plateau. The algorithm rewards asset diversity and punishes accounts that under-feed it. Under-asseted PMax campaigns cap out at impression-share ceilings that a properly fed campaign would never hit.

APEX runs in-house copywriters and creative production on every Google Ads engagement. That includes responsive search ad headlines and descriptions refreshed weekly, Shopping titles rewritten quarterly, Performance Max asset group refreshes monthly, vertical creator-style video for YouTube Shorts and Demand Gen produced in production cycles, landing page hero copy coordinated with the campaign rotation, and email and remarketing creative kept consistent with the ad-side voice. The same brand voice runs across Search, Shopping, YouTube, Demand Gen, and email so the buyer who sees four touches sees one company, not four agencies.

The numbers behind asset diversity: PMax campaigns with 11 or more headlines and 4 or more lifestyle images produce 30 to 50 percent more conversion volume at similar CPA than campaigns running the minimum required asset count. Demand Gen rewards vertical video heavily. Shoppable CTV inventory rewards short-form lifestyle creative over product-on-white. Without an in-house creative function, the agency cannot produce the volume the algorithm rewards, and the campaign caps out at a performance ceiling no bid optimization can break. Our visual identity and brand systems work carries directly into ad creative for clients who run both engagements.

AI-assisted continuous optimization

We run a proprietary in-house AI optimization layer on every Google Ads account. It is the second biggest reason our analyst-to-account ratio outperforms typical agencies, and the reason our contribution-margin ROAS holds steady at the level it does without our hourly fees scaling with account complexity.

What it does daily: surfaces spend, CPA, and ROAS anomalies the moment they trigger; audits search-term reports for new negative keyword candidates and proposes them ranked by waste impact; drafts ad copy variations for analyst review based on the products, audiences, and queries currently performing; flags feed gaps and disapprovals as they happen rather than at the next monthly review; identifies bidding outliers (campaigns where the bid the algorithm is placing has drifted outside a defensible range); and watches for policy disapprovals, account suspensions, and competitive spend shifts. Research from Princeton's GEO study on how AI engines surface citations confirms what every agency that runs AI tooling already knows: structured data, statistics, and citable claims compress decision cycles by an order of magnitude. The same logic applies internally to account management.

Crucially: the AI does not run the account. A Google Partner-certified human analyst makes every change. The AI compresses the analyst review cycle from days to hours. A waste pattern that a human would catch on the next monthly review surfaces in the daily anomaly report. A new negative keyword that would otherwise wait for the weekly QA queue gets proposed within hours of showing up in the search-term report. A creative gap that would normally be noticed when the quarterly asset refresh kicks off gets flagged the moment a competitor's ad volume changes. The throughput we deliver per account would require three times the headcount on a manual workflow.

The discipline is in the boundary. AI proposes. Humans dispose. Every change is logged with the rationale, the analyst who approved it, and the result two weeks later. This is the same workflow we describe in our companion piece on making your website visible to AI, translated into the optimization domain rather than the discovery domain.

Strategy and business consulting, not just ad management

A Google Partner that has run hundreds of accounts is the cheapest business consultant a scaling e-commerce operator can hire. The agency sits on top of more comparable account data than any single in-house team will ever see. That perspective is worth more than any single optimization delivered inside the platform.

The strategy work we deliver on every engagement, included in the management fee, covers pricing architecture (an AOV that does not support the channel will not survive any campaign change), margin analysis (which SKUs to push, which to throttle, which to stop advertising entirely), channel mix (when more Google spend is wrong and Meta, TikTok, email, or affiliates would compound better), landing page conversion rate (we route specific clients to our landing page optimization service when the LP is the bottleneck rather than the campaign), and competitor intelligence via our competitor analysis service. Quarterly business reviews look at cohort LTV, customer acquisition cost payback, and contribution margin trend across 12 months, not at last week's CTR.

This is the line that separates a Google Partner from a "PPC manager." A PPC manager runs the campaigns. A Google Partner partners with the business and points the campaigns at the part of the business that benefits most from advertising spend. The work is harder. The fee structure has to reflect both the management labour and the consulting time. Our pricing on Google Ads for e-commerce (EUR 750 per month plus 10 percent of net conversions) aligns the incentive: we earn more when the business sells more, and we have a structural reason to refuse work that will not produce a return.

Real ROAS vs reported ROAS: the metric that actually matters

Platform ROAS is revenue divided by ad spend, with no costs deducted. It is the number Google Ads shows in the dashboard. It is the number every agency reports in their monthly PDF. It is also the number that funds nothing.

Contribution-margin ROAS is the ROAS that pays salaries. The formula is (Revenue minus cost of goods, shipping, payment processing fees, returns, and VAT) divided by ad spend. Saras Analytics has documented the framework in detail and concluded that "ROAS does not equal profit and can often mask real financial outcomes, creating an illusion of success while silently undermining margin and sustainability." That sentence is worth printing and pinning above every analyst's desk.

The arithmetic on a typical e-commerce account: a platform ROAS of 4.0:1 sounds healthy. EUR 1,000 of ad spend produces EUR 4,000 of attributed revenue. Strip out 50 percent COGS, 5 percent shipping, 2.5 percent payment processing, 5 percent return rate, and the embedded VAT, and the actual contribution to the business is closer to EUR 1,200 to 1,400 per EUR 1,000 of spend. That is a 1.2:1 to 1.4:1 contribution-margin ROAS. The dashboard says 4:1. The P&L says 1.3:1. The agency that walks past the gap is selling you a story.

On most APEX-managed accounts, the contribution-margin ROAS sits above 6:1. We get there by structuring the account around margin tier rather than around revenue, by reporting on the metric that matches the P&L rather than the metric that matches the dashboard, and by routinely killing campaigns where the platform-reported ROAS is acceptable but the contribution-margin ROAS is underwater. Platform ROAS at 4:1 with contribution-margin ROAS at 6:1 is a different account from platform ROAS at 4:1 with contribution-margin ROAS at 1.3:1, even though the dashboard cannot tell them apart. An agency worth its fee builds the second number into the reporting from day one. Google's pillar piece on Google Ads for e-commerce in 2026 covers the platform-side benchmarks that frame the conversation.

What this looks like in practice

The Online Groceries case study is a Canadian grocery retailer where we ran Performance Max, Shopping, and dynamic remarketing layered over a full GA4 plus Consent Mode v2 implementation. Total sales grew from CAD 941K in 2023 to CAD 1.06M in 2024 (+13 percent), orders climbed +43 percent year over year, and Q1 2025 sales accelerated +29 percent versus Q1 2024 with orders up +51 percent. The unit economics in online grocery are tight (typically a 4:1 to 8:1 platform ROAS is the realistic range), and the lift was driven by feed completeness, offline conversion imports for catering and bulk orders, and a custom data layer that captured the repeat-purchase behaviour grocery customers actually exhibit.

The PISCINE Pool Supplies case study is a Romanian retailer in a hard-seasonality vertical (pools are bought between April and August). We rebuilt the feed, segmented PMax by margin tier, and ran a tight Standard Shopping overlay on margin SKUs. Result: +228 percent revenue, +183 percent orders, +143 percent units sold. The seasonal compression mattered. We shipped the foundation in March and the campaign carried the season. Without the foundation, we would have spent April optimizing instead of selling, and the season would have been half over before performance stabilized.

What a real engagement looks like

Weeks 1 to 2: onboarding. Account audit (the same 12-point self-audit further down this post), custom data layer deployment, conversion tracking validation, Consent Mode v2 implementation if not in place, feed pass for e-commerce accounts, account architecture rebuild with brand defense separated from generic and PMax split by margin tier. No bidding strategy changes in this phase. The foundation has to be solid before we touch the layer above.

Weeks 3 to 8: stabilization. Smart Bidding requires roughly 30 to 50 conversions per month per campaign to exit the learning phase reliably. Performance Max needs at least 30 conversions in the past 30 days at the campaign level for Target ROAS to function. We hold the structure stable, refresh creative on the planned cadence, and let the algorithm cement the new equilibrium. Brand campaigns are typically positive within days. Generic Search and PMax retail take three to six weeks to settle.

Month 3 onward: optimization velocity. Weekly cycle: n-gram analysis, anomaly review, asset rotation, bid-strategy adjustments inside the 20 to 30 percent of current actual envelope. Monthly cycle: full performance report with platform ROAS and contribution-margin ROAS side by side, asset group refresh, feed audit. Quarterly cycle: business review covering cohort LTV, CAC payback, channel mix recommendations, and any structural changes to the account.

Pricing is transparent. Our Google Ads PPC management starts at EUR 500 per month for non-e-commerce accounts. Google Ads for e-commerce is EUR 750 per month plus 10 percent of net conversions, which aligns the incentive: we earn more only when you earn more. A one-off account audit is EUR 400 with three-day delivery, and the audit fee is credited against the first month of management if you transition to a retainer. No contract lock-in. Month-to-month from day one.

When you should not hire an agency

Below roughly EUR 2,000 per month in ad spend, a careful in-house operator with time to learn can outperform an under-resourced agency. The reason is straightforward: a EUR 500 per month management fee on EUR 1,500 of ad spend is a 33 percent overhead. The agency cannot afford to deliver the full scope at that price. They will set up campaigns, run a default Smart Bidding strategy, and check in monthly. That is fine, but the same outcome is achievable in-house with five hours of weekly attention from someone who reads the Google Ads help docs and knows the business.

Above EUR 2,000 per month in ad spend, the math flips. The leakage from a misconfigured PMax (you have run the wrong tROAS for six months), broken conversion tracking (you have been bidding on the wrong outcome), missing Consent Mode v2 (your EU traffic is invisible to remarketing), or a feed at 70 percent completeness (you are losing impression share to competitors at 99 percent) typically exceeds the agency fee within the first 30 days. Above EUR 5,000 per month in spend, an account manager, a feed engineer, a measurement specialist, and a creative team are paying for themselves on the leakage prevented, before any optimization upside is even counted.

The honesty here is itself an EEAT signal. Google's Search Quality Rater Guidelines state that "Trust is the most important member of the E-E-A-T family because untrustworthy pages have low E-E-A-T no matter how Experienced, Expert, or Authoritative they may seem." If we tell every prospect that paying us is the right answer, we are selling. If we tell some prospects that they should run ads in-house for now, we are advising. The second is the only relationship worth starting.

The 12-point self-audit you can run today

You can run this checklist yourself in ninety minutes if you have admin access to Google Ads, Merchant Center, GA4, and Tag Manager. Score each item zero or one. Anything below 9 needs urgent attention. We run the same 12 questions at the start of every paid engagement and during the free 30-minute audit.

Foundation (4 points)

1. A custom data layer is deployed with item-level fields, not the GTM template defaults.
2. Enhanced Conversions is enabled and reporting "Recording" with no health warnings in Google Ads.
3. Consent Mode v2 (Advanced) is implemented with all four signals firing from a Google-certified CMP.
4. Google Tag Gateway or server-side GTM is in place and validated against a baseline conversion-count test.

Account structure (3 points)

5. Brand and generic queries are split into separate campaigns with brand exclusions on the generic side.
6. Performance Max retail campaigns are split by margin tier via custom labels in the feed, not lumped together.
7. Standard Shopping is running in High priority alongside PMax for margin SKUs and clearance.

Feed and creative (3 points)

8. Feed is at 90 percent or higher attribute completion, with GTINs, sizes, colors, materials, and category mappings populated.
9. Performance Max asset groups are themed by category, not lumped, and contain at least 11 headlines plus 4 lifestyle images each.
10. RSA assets are refreshed at least monthly, with weekly variant rotation tracked in a change log.

Reporting (2 points)

11. Monthly reporting includes both platform ROAS and contribution-margin ROAS, with the inputs (COGS, shipping, fees, returns, VAT) documented per SKU group.
12. n-gram analysis is run weekly and the resulting negative keywords are pushed to both campaign-level and account-level lists.

Frequently asked questions

What does a Google Ads agency actually do in 2026?

A real Google Ads agency in 2026 sells far more than campaign management. The deliverables fall into ten buckets: a custom data layer that captures clean, durable conversion data; Enhanced Conversions and offline imports; Consent Mode v2 implementation for EEA traffic; keyword and negative-keyword strategy with weekly n-gram analysis; product segmentation by margin tier inside Performance Max; Google Merchant Center feed engineering; copywriting and creative production for Search, Shopping, YouTube, and Demand Gen; AI-assisted continuous optimization that compresses the analyst review cycle; account architecture and bidding strategy; and quarterly business consulting on margin, channel mix, and unit economics. Anything less is a 2019 deliverable wrapped in a 2026 invoice.

How much does a Google Ads agency cost in 2026?

Budget agencies charge 500 to 1,500 euros per month, mid-tier agencies charge 1,500 to 5,000 euros per month, and premium agencies charge 5,000 to 20,000 euros per month or more, with most engagements priced as a flat retainer or 10 to 20 percent of ad spend. APEX DIGITAL prices Google Ads PPC management at EUR 500 per month, Google Ads for e-commerce at EUR 750 per month plus 10 percent of net conversions, and a one-off account audit at EUR 400 with three-day delivery. Below EUR 2,000 per month in ad spend, a careful in-house operator may outperform an under-resourced agency. Above that threshold the cost of misconfigured tracking, broken Consent Mode v2, or an under-engineered feed typically exceeds the agency fee.

What is a custom data layer and why does it matter for Google Ads?

A data layer is the structured event payload your site emits when something meaningful happens: a product view, an add to cart, a checkout, a purchase. A custom data layer is one engineered for your specific business model rather than a default GTM template. It captures item-level fields like SKU, price, margin tier, and customer segment so Smart Bidding has signal to optimize against, and so server-side relays and Enhanced Conversions can recover the 15 to 40 percent of conversions that browser-only setups lose to ad blockers, iOS Intelligent Tracking Prevention, and consent-revoked sessions. A clean custom data layer is the load-bearing foundation of every other lever in the account.

Is Consent Mode v2 mandatory in Romania and the EU?

Yes. Consent Mode v2 has been mandatory since March 2024 for any site running Google ad products and targeting end users in the European Economic Area, including Romania. Google states that advertisers receiving data from EEA end users must collect consent and share consent signals with Google to use tags and SDKs for measurement, ad personalization, and remarketing features. Without it you lose remarketing, conversion modeling, audience building, and Enhanced Conversions for EU traffic, and your account is subject to policy compliance warnings and potential serving restrictions. Consent Mode v2 is not optional. It is the EU compliance line every account must cross cleanly.

What is contribution-margin ROAS and how is it different from platform ROAS?

Platform ROAS is revenue divided by ad spend as Google Ads reports it, with no costs deducted. Contribution-margin ROAS is revenue minus variable costs (cost of goods, shipping, payment processing fees, returns, and VAT) divided by ad spend. They measure different things. Platform ROAS is a top-line number that funds nothing. Contribution-margin ROAS is the only ROAS that pays salaries. APEX DIGITAL holds a contribution-margin ROAS above 6:1 on most client accounts, even when the platform ROAS reads closer to the 4:1 industry average. The two numbers are both true. Only one of them is the one you should manage to.

Do you handle copywriting and creative production in-house?

Yes. Copywriting and creative are not a side task handled by the account manager between bid adjustments. APEX DIGITAL runs in-house copywriters and creative production for Search responsive ads, Shopping titles, Performance Max asset groups, YouTube and Demand Gen video, and landing page hero blocks. We refresh RSA variants weekly, refresh PMax asset groups monthly, and produce vertical creator-style video for YouTube Shorts and Shoppable CTV inventory. Creative diversity is the single biggest under-performing input on most under-asseted PMax accounts, and asset diversity is what separates accounts that scale from accounts that plateau.

Do you use AI to manage Google Ads accounts?

We use a proprietary in-house AI optimization layer on every account. It does not run the account. A human Google Partner-certified analyst makes every change. The AI compresses the analyst review cycle from days to hours: it surfaces spend, CPA, and ROAS anomalies, audits search-term reports daily for new negative keyword candidates, drafts ad copy variations for analyst review, flags feed gaps and disapprovals, identifies bidding outliers, and watches for policy changes. The result is a weekly optimization velocity that competing agencies cannot match without three times the headcount. The discipline of human judgment plus the throughput of automation is what allows us to deliver above-average contribution-margin ROAS at sustainable fees.

How long until a new Google Ads agency engagement is profitable?

Plan for two to four weeks of onboarding (audit, custom data layer, conversion tracking validation, Consent Mode v2 implementation, feed pass, account architecture rebuild) and 60 to 90 days before performance stabilizes. Smart Bidding requires roughly 30 to 50 conversions per month per campaign to exit the learning phase reliably. Performance Max needs at least 30 conversions in the past 30 days at the campaign level for Target ROAS to function well. Brand campaigns turn positive almost immediately. Generic Search and PMax retail typically need a feed pass, conversion-tracking validation, and at least one full bidding-cycle reset before cost-per-acquisition lands inside target. The accounts where engagements pay back fastest are the ones where the agency rebuilds the foundation first and only then optimizes the campaigns running on top of it. Our pillar piece on Google Ads for e-commerce in 2026 covers the platform-side dynamics in full.