Supermetrics review 2026: what the Google Ads connector does — and where it stops
The native Google Ads connector in Looker Studio is free, and for a while it is enough. Then the familiar wall: a report across fifteen accounts takes four minutes to refresh, half the fields you need are missing, and putting Google Ads next to Meta and GA4 in one table is not on the menu. That is when most people start pricing paid connectors, and Supermetrics is the first name in the results. This Supermetrics review works strictly from the vendor’s own documentation and pricing page: what it charges for, which limit bites first, where its numbers are guaranteed to disagree with the Google Ads interface, and when paying is still premature.
A word on method. Every number in this Supermetrics review comes from the vendor’s own docs or pricing page. Where the vendor talks about itself in marketing material, that is labelled a vendor claim rather than a verifiable spec. Where the documentation is silent, I say so — for a paid tool, a missing number is itself a finding.
What it is, and what it is not
Supermetrics is not an ad management platform and not a tracker — it optimises nothing and bids on nothing. Its job is to pull data out of an account over the API and drop it where you read it: Google Sheets, Looker Studio, Power BI, Excel, a warehouse or an AI assistant. It is transport, and should be judged as transport: completeness, predictability, cost of ownership.
That line matters practically — it stops you buying the wrong category. A management platform like Optmyzr changes campaigns by rules. A tracker like Keitaro sits in the click path and counts what the ad platform never sees. A connector does neither: it moves numbers the platform already calculated. If your pain is “I cannot see what happens after the click”, this will not fix it.
An honest fork at the entrance: with one Google Ads account feeding one Looker Studio dashboard, the native Google Ads connector is enough. A paid connector pays for itself on two things — many accounts, and several sources in one report.
What you actually pay for: four limits, not one price
The central thing about Supermetrics pricing: there is no single billing unit. You buy a package of four independent allowances and hit whichever runs out first. The pricing page calls them destinations, users, data sources and accounts — the last defined as “The number of accounts you can connect per data source”.
| Package | Destinations | Users | Data sources | Accounts per source | Per month, billed yearly | Per month, billed monthly |
|---|---|---|---|---|---|---|
| Starter | 1 core destination | 1 | 3 | 3 | $44 | $55 |
| Growth | 1 core destination | 2 | 6 | 7 | $177 | $222 |
| Enterprise | negotiated | Custom user limit | negotiated | Custom account limit | “Get a Quote” — no published price | |
The gap between yearly and monthly billing is stated plainly on the pricing page: “Save 20% with yearly plan”. Read that the other way round — the annual price is the real price, and paying monthly costs you about a quarter more.
Each allowance tops up separately, with its own ceiling: an extra destination from $49 (two in total), a user from $37 (three) or $99 for the wider bundle (five), a data source from $37 (nine), an account from $11 (fifteen). Treat those as indicative — the page shows two add-on columns without labelling which is monthly and which yearly, so budget the upper figure.
The consequence is simple. Do not price the package — price the package plus the add-ons your real configuration needs. An agency with six clients and a dashboard each does not run out of data sources; it runs out of destinations and accounts, exactly where the ceilings are lowest (two and fifteen).
For anyone paying on invoice rather than by card: the primary method is credit card, and the docs warn that “an additional administration fee will if applied if the total invoice amount is under €/$2,400” (the source’s own wording and typo). At modest volumes, invoicing costs more than the subscription by that fee.
The agency trap: Google Ads accounts are counted one by one
The most expensive single fact in this review is not on the pricing page but in the connector docs: “In the Google Ads connector, each Ad account is considered a separate account within the connection”.
An MCC does not count as one account. A manager account with twenty client accounts consumes twenty units of your quota — and the Starter package allows three.
Now the second rule: “All data sources share the same account limit” — the allowance is pooled across sources. Connect Google Ads, Meta and GA4 and they draw from one counter. An agency with ten clients, each with a Google Ads account, a Meta account and a GA4 property, needs thirty units against seven on Growth and fifteen as the add-on ceiling. Beyond that it is Enterprise, at a price nobody publishes.
A third detail runs the counter up quietly. For Google Ads, Google Analytics, GA4 and Microsoft Advertising paired with Google Sheets, “All accounts” is on by default, and the docs spell it out: “This means that every data source account will be included in every query”. Every query pulls every connected account, even when the report needs one.
What happens when you go over? The docs show one behaviour — an error reading “You’ve reached your plan’s limit of 3 Facebook Ads accounts”. No overage billing, no auto-upgrade, no grace period is described. That asymmetry favours the customer against platforms that bill overage silently: you get a refusal, not a surprise invoice. But the work stops.
Prioritised accounts can be swapped up to ten times per month, the counter resetting on the first. For an agency rotating clients that belongs in your process, not someone’s memory.
Quotas: Google Ads has none, the rest of your stack does
Good news that also tells you where you will actually be stopped. In the daily request quota table Google Ads is listed as “Unlimited operations / day” — for daily reporting across a large portfolio, that matters.
But a connector rarely works alone, and its neighbours have finite ceilings:
| Source | Published limit |
|---|---|
| Google Ads | Unlimited operations / day |
| Google Analytics 4 | 200,000 tokens per property per day (GA360: 2,000,000); 40,000 per hour; 10 concurrent requests per property |
| Google Display & Video 360 | 2,000 requests per day |
| Google Campaign Manager 360 | 30,000 requests per user per day; 30 simultaneous reports |
| Google AdSense | 10,000 requests per day |
| Google Search Ads 360 | 3,000 queries per minute; 150,000 per day per project |
| Google Ads Keyword Planner | 1 request per second |
| Facebook Ads | dynamic limit based on account usage and ad spend |
| Google PageSpeed Insights | 200 queries per day per subscription |
And the caveat that makes “request” a deceptive unit: “A single Supermetrics query can result in many requests to the data source API”, the count depending on the metrics and dimensions picked. One heavy Sheets report can burn a quota you had budgeted for hundreds of calls. Hence the documentation’s down-to-earth remedies: shorten the date range, drop fields, query fewer accounts at once.
What the docs do not say: what counts as one request generally, whether a query fails or queues at the cap, when Supermetrics’ own quotas reset, and whether scheduled refreshes consume the same quota as manual ones. For Google Ads it does not matter — there is no cap. For GA4 in the same report it matters, and I found no answer.
The 37-month history wall that landed in June 2026
The newest and most awkward change for anyone doing serious analysis, and it comes from Google rather than Supermetrics. The docs: “In early June 2026, Google will be making changes for the data retention policies for Google Ads”, and “shorter breakdown periods such as hourly, daily, and weekly will now be restricted to the last 37 months”.
The restriction hits any query containing at least one of these breakdowns: Date, Hour, Today, Day of week, Week (Mon–Sun), Week (Sun–Sat), Week (Mon–Sun, dates) and both Year & week variants — nearly every trend report you would actually build.
There is a way around it at the cost of granularity: “If you want to fetch older data to a maximum of 11 years, you have the following options” — switch to a Month, Quarter or Year breakdown, or drop the date dimension. Queries without a date dimension carry their own rule: beyond 37 months the start date must be the first day of a month and the end date the last. Other combinations will not run.
Two further ceilings are easy to miss while designing a report:
- Unique users and Impression frequency — “These metrics specifically only allow data for the last 3 years”.
- Click view report — “The Click view report can only be fetched for up to the last 90 days”. If you build GCLID-level analysis on it, the history is gone after a quarter.
The conclusion is architectural: if history deeper than three years matters to you, store it yourself rather than expect it from the API. That is a direct argument for loading Google Ads into BigQuery and accumulating your own tables: the connector feeds the warehouse daily and you simply outlive the 37-month window.
Where this Supermetrics review has to be blunt: seven reasons your numbers won’t match
The commonest question asked of any connector is “why don’t these figures match the interface?”. To its credit Supermetrics answers more honestly than this class usually does, in its “Good to know about Google Ads” page. You want this list before a client asks first.
1. Attribution is not under your control. “Currently, Supermetrics does not support user-controlled attribution settings in the Google Ads data source connector.” Changed the attribution model in the interface? The connector will not mirror it.
2. Conversions arrive as an estimate. “Google makes an estimation of conversions and conversion values”, then the key sentence: “The more dimensions included in the query, the higher the change of discrepancies” (source’s own typo). The finer the breakdown, the further totals drift — the docs show rows for Finland and Sweden summing to 25.5 while the total for the same query reads 25.61.
3. Warehouse transfers drift more than dashboards. “This affects mainly Data Warehouse transfers as, by default, they get a higher level of granularity.” Same mechanism as point 2, but teams building a warehouse expect the opposite — that is where they assume the “most accurate” numbers live.
4. Parent-account labels do not arrive. If a child account uses labels owned by the manager account, they show in the interface but, per the docs, “they will not be reported via the Google Ads API”. Your label-based report is incomplete, and the connector is not the cause.
5. Zero-impression rows are filtered out by default. “The Google Ads API no longer supports this feature” — the old mechanism is gone and the connector drops those rows. The include_zero_impressions setting brings them back, with a warning: “Please note that this may greatly increase the volume of data returned.” And if every metric for an object is zero and the query includes a segment-type dimension, the API may return nothing.
6. GCLID is not one-to-one with clicks. “The association between GCLID to clicks is not exactly 1:1”, and with manual tagging one GCLID can carry many clicks. Reconciliation built on unique GCLIDs breaks here.
7. The refresh window must match your attribution window. The documented recommendation: set the re-fetch window “to the maximum of your attribution windows on the platform”. If your conversion window is 30 days and the connector re-pulls only the last 7, some conversions never land in your report — they are credited after you stopped looking.
Plainly: none of the seven is a defect in Supermetrics. Six are properties of the Google Ads API that the vendor conscientiously documented. For the owner of the report the difference is nil — the numbers diverge and you explain them. If that is a recurring problem, I covered the diagnostic order in the piece on reconciling Google Ads spend.
What the connector cannot do at all
Flat refusals from the documentation. Read before paying — none is a settings problem:
- Experiments. “Experiment campaigns are not currently supported.” Tests run through Experiment Center will not reach your report.
- DSA together with keyword text fields. “Dynamic Search Ads Campaigns currently do not work in conjunction with KeywordsText fields.”
- Custom columns with calculations. “the Google Ads API does not support them at this time” — rebuild those formulas in the destination.
- App install conversions need the connected user to have access; accounts with no App campaigns usually do not.
On deprecated fields lurking in inherited templates: as of 18 June 2025, Extension type, Attribute values, Extension status, Feed ID and Feed item ID return no value; earlier casualties include Average position, Average frequency, Max CPA and the assisted-conversion fields. Taking over someone’s dashboard, check these first — they do not error, they quietly return blanks.
Destinations and the “core destination” catch
Named across the vendor’s docs and site: Google Sheets, Looker Studio, Power BI, Excel (via an add-in), BigQuery, Snowflake, Amazon Redshift, Azure Synapse, Supermetrics’ own storage, API access in JSON or CSV, and the AI assistants — Claude, ChatGPT, Copilot and Gemini.
The billing subtlety: a package includes “1 core destination”, and a second is an add-on capped at two in total. So “a Looker Studio dashboard for the client plus a BigQuery export for ourselves” is already two destinations and will not fit on Starter without paying more — the most common reason a quote lands above expectation.
In the API source list, Google Sheets, BigQuery and Snowflake also appear as readable sources, and there is a dedicated BLEND source for combining data. Some are tagged “(early access)” — ChatGPT Ads and Google Ads Account Explorer among them; I would not build client-facing reporting on anything carrying that tag.
What changed during 2026
| Date | Change |
|---|---|
| 12 Jan 2026 | Fields AI Max headline, AI Max landing page, AI Max search term and All conversions value per cost added, with a published list of dimensions and metrics compatible with the AI Max report type |
| June 2026 | The 37-month history restriction on short breakdowns takes effect (a change on Google’s side) |
| 2 Jul 2026 | Connector upgraded to Google Ads API v22; a new “Use targeting expansion view (AI Max)” setting switches to the report that surfaces “Total: AI Max landing page matches” the way the interface does |
| 21 Sep 2026 | A separate “Google Ads reporting changes” entry linking to its own page; its contents are outside this review |
The practical takeaway: if you already run AI Max in search, check which of the two report versions is active for you — otherwise your landing-page-match figure will differ from the interface purely because of a connector setting.
MCP and AI access: where the spec ends and the promise begins
Through 2026 the vendor has sold its AI layer hard: an MCP server plus connections to Claude, ChatGPT, Copilot and Gemini. The factual part is that those destinations exist and are named in the docs. Past that we are in vendor-claim territory.
The boldest claim: in its own comparison material Supermetrics says its MCP offers write access — “you can adjust budgets, pause & resume campaigns, create completely new campaigns”. That is running an ad account from a chat window, and it deserves the caution due any automation with write permission: a dedicated user with minimum rights, nothing pointed at production until proven, a change log. I could not verify it against technical documentation — the pages I read detail no write scopes.
Second: how AI access maps to packages is never clearly stated. One page advises teams needing “advanced AI and activation features” to look at higher tiers, but there is no explicit “MCP is available from plan X” — a question for a salesperson before you pay. On splitting work between a human and a model in reporting, I went deeper in the piece on AI dashboards in Google Ads.
The vendor contradicts itself on its own numbers
While gathering sources, something surfaced that is useful in its own right: Supermetrics’ own pages quote different figures about Supermetrics.
| Item | Version A | Version B |
|---|---|---|
| Starter, billed yearly | $39/mo | $44/mo |
| Starter, billed monthly | $49/mo | $55/mo |
| Growth, billed yearly | $159/mo | $177/mo |
| Connector count | “hundreds” / 150+ | 170+ |
| Tier names | Starter, Growth, Business | Starter, Growth, Enterprise |
The pricing page itself shows $44 and $177 on annual billing — version B. The likely explanation is page age: the vendor’s comparison posts are dated July and September 2026, with pricing marked “checked” in October and August. The pragmatic rule: pin price and package contents to the pricing page and your written quote, never to the vendor’s blog — and that applies to any SaaS, not just this one.
The same caution goes for competitive comparisons. Supermetrics publishes a nine-platform table (Adverity, Cometly, Fivetran, Funnel, Improvado, Measured, Twilio Segment, Whatagraph and itself) with prices and connector counts. A useful rough map of the market; not a basis for choosing, since one of the compared parties wrote it.
Who needs it, who is too early
| Profile | Verdict | Why |
|---|---|---|
| One Google Ads account, one Looker Studio dashboard | too early | The native connector covers it |
| One advertiser, 2–3 sources, report in Sheets | yes, Starter | 3 sources and 3 accounts fit the allowance exactly |
| Agency, 5–7 clients, a dashboard each | price carefully | You run out of destinations (ceiling 2) and accounts, not sources |
| Agency, 10+ clients, an MCC each | top tier only | Every child account is a separate unit; 15 accounts is the add-on ceiling |
| You need history deeper than 3 years | yes, with a warehouse | The 37-month API window is beaten only by accumulating your own tables |
| Your pain is “what happens after the click” | no | That is a tracker’s job, not a connector’s |
| You want automated rules acting on campaigns | no | That is the management-platform category |
What the documentation does not say
- Data retention and refresh frequency per tier — absent from the pricing page and the subscription docs.
- Behaviour on exceeding the allowance — only an error; nothing on overage, auto-upgrade or grace period.
- Contract term and cancellation.
- What counts as a single request in general, and when Supermetrics’ own quotas reset.
- Whether scheduled refreshes consume quota like manual ones.
- Whether API access is in every tier, and its authentication and rate limits.
- How MCP and AI access map to packages.
- How an MCC is counted — the per-account rule is in the connector docs, but the pricing page, where the buying decision happens, carries no such warning.
One thing deserves credit, because it is rare in this class: the “good to know” material for Google Ads is detailed and written against the vendor’s own sales interest, listing where its numbers will not match the interface and giving a numeric example of the drift. That is well above a market where this surfaces in month three of production use.
Where this blog draws the line
A data connector is neutral: it reads what the ad platform already computed. It has nothing to do with evading review, antidetect setups or disguising traffic, and none of that appears here. Workable infrastructure for paid traffic means Google Ads agency accounts and clear reporting on top, not attempts to fool platform checks. Other tooling sits in the services section, and further breakdowns are in the blog.
Verdict
Supermetrics is solid data transport with unusually honest documentation and one structural quirk that decides whether it fits at all: accounts are counted individually and the allowance is pooled across every source. For a single advertiser with a couple of sources it is a cheap, fast way to get real reporting. For an agency carrying ten MCCs the arithmetic breaks even on the upper packages.
Second, the 37-month history window is the new reality of 2026, and it is a Google Ads API constraint rather than a Supermetrics one. If you need history you have to accumulate it. The connector is a fine way to do that — but then you are buying warehouse ingestion, not reporting, and the warehouse belongs in the same budget line.
Third, general: a SaaS price with no single billing unit always costs more than the pricing page suggests. Price the package plus add-ons against your real configuration, before the trial rather than after.
FAQ: the Supermetrics Google Ads connector
How much does Supermetrics cost for Google Ads?
The pricing page publishes two packages: Starter at $44 per month billed yearly or $55 monthly, and Growth at $177 and $222. Enterprise is quote-only. Add-ons for extra destinations, users, sources and accounts sit on top, so the final figure is almost always higher than the headline.
Does an MCC count as one account or several?
Several. The documentation states that in the Google Ads connector each ad account is considered a separate account within the connection. A manager account with twenty child accounts consumes twenty units — and Starter allows three.
Is the account allowance separate for each data source?
No, it is pooled. The docs put it as “All data sources share the same account limit”, so connected Google Ads, Meta and GA4 accounts all draw from one counter.
Is there a daily request limit on Google Ads?
No. The quota table lists Google Ads as “Unlimited operations / day”. Other sources do have finite limits, and a single Supermetrics query can trigger many API calls, so in a multi-source report the bottleneck will not be Google Ads.
Why don’t the figures match the Google Ads interface?
Seven main reasons: attribution cannot be controlled from the connector, conversions arrive as a Google estimate that drifts with every added dimension, warehouse transfers drift more than dashboards, parent-account labels are not returned by the API, zero-impression rows are filtered by default, GCLID is not one-to-one with clicks, and your re-fetch window may be shorter than your attribution window.
How far back can Google Ads history be pulled?
With a day, hour or week breakdown, the last 37 months — Google restricted short breakdowns to that window from early June 2026. Going deeper, up to eleven years, needs a month, quarter or year breakdown or no date dimension, and then the period must start on the first day of a month and end on the last.
What can the connector not do at all?
Experiment campaigns are not supported, Dynamic Search Ads do not work together with keyword text fields, custom columns containing calculations are not transferred, and app install conversions require the connected user to have access to them.
Will Supermetrics replace a tracker or a management platform?
No. It is data transport: it moves what the platform already counted. It does not sit in the click path like a tracker, nor change campaigns by rules like a management platform. If your question is what happens after the click, you need a different tool.
Is there a free trial?
Yes, fourteen days, and no credit card is needed. What the trial includes is not separately documented — a good reason to run it on your real destination and account configuration rather than a toy setup.
Can campaigns be managed from Supermetrics through AI?
The vendor claims its MCP server offers write access and lets you adjust budgets, pause and resume campaigns and create new ones. The technical documentation I read details no write scopes. Treat such access like any automation with write permission — a dedicated user, minimum rights, proven outside production first.
Is it worth buying if I only need Google Ads reporting?
If that means one account and one dashboard, probably not — the native connector will do. A paid connector earns its keep in two scenarios: many accounts in one report, and several data sources side by side.
