Table of contents
Before you ever see it, your marketing data is capped, sampled and half-hidden. Google Analytics samples your reports once traffic climbs. Search Console hides the search terms behind almost half your clicks and forgets anything older than 16 months. Google Ads, LinkedIn and your CRM each keep their own slice behind their own login, with their own version of the numbers.
That is the problem a marketing data warehouse solves. And the barriers that once put one out of reach have fallen away: it costs a fraction of what it used to, you no longer need SQL or a database background, and you no longer need a data engineer to set it up.
Data privacy
If you have ever exported your top queries from Search Console, added up the clicks, and found they don’t match the total on the chart, you haven’t made a mistake. Google does this on purpose.
Search Console hides what it calls anonymized queries: searches not issued by more than a few dozen users over a two-to-three-month period. It drops them to protect people’s privacy, which is reasonable. The problem is what it does next. As Google explains in its own deep dive into Search Console’s data:
“While the actual anonymized queries are always omitted from the tables, they are included in chart totals, unless you filter by query.”
Daniel Waisberg, Search Advocate at Google
So the rows you can see never sum to the total, and the moment you filter by query, the hidden clicks vanish entirely. Google’s own worked example: your visible queries add up to 450 clicks, but the chart says 550. Filter to queries containing a given word and you get 175; exclude that word and you get 275. Add those up and you have 450, still not 550.
This isn’t a rounding error. An Ahrefs study of billions of clicks found that in April 2025, anonymized queries made up 46.77% of them. Close to half your search data can be missing from any query-level report. An AI cannot reason its way around data it has no idea is missing.
Caps, sampling and a short memory
Anonymized queries are only the start. GA4 and Search Console both ration your data in ways that distort what any tool, human or AI, sees:
| The limit | GA4 | Search Console | A BigQuery export |
|---|---|---|---|
| Row caps | High-cardinality dimensions (500+ unique values a day) push rows past the report limit into an “(other)” bucket. | Shows 1,000 rows in the UI, up to 50,000 a day via the API. | No row cap. Every row is kept. |
| Sampling | Samples reports and explorations once you pass its event quota. | Not sampled, though rare queries are hidden (see below). | No sampling. Raw, event-level data. |
| History | Keeps user-level exploration data for up to 14 months (2 months by default). | Keeps roughly 16 months. | Kept forever, from the day you switch it on. |
| Hidden queries | Not applicable (GA4 has no search-query data). | Withholds anonymized query rows from tables and the API. | Clicks and impressions are kept (the query text stays hidden), so the totals reconcile in full. |
Feed that to an AI and you get a fluent answer built on a partial, stale sample. None of these limits exist because your data is unavailable. Google has all of it. These are limits on what the interfaces and APIs will hand back to you.
It’s not just GA4 and Search Console
The same three failure modes (row caps, short retention and modelled or sampled numbers) run through the rest of the stack. Salesforce reports display a maximum of 2,000 rows. HubSpot caps custom reports at 1,000. Meta Ads Manager keeps 37 months of data, models your attribution and narrows which attribution windows you can even see. Shopify’s dashboards often don’t match its own exports.
The interface always shows less than the platform holds. A warehouse is where you get the rest back, from every source at once.
The fix: a marketing data warehouse
The fix is to get all of that data into one place you own, with none of those limits. That place is a marketing data warehouse: a central store where you keep the raw, first-party data from all your marketing tools and query it directly, with SQL or with plain-English AI. The usual sources are the big ad platforms (Google Ads, Meta Ads, LinkedIn Ads), your web and search analytics (GA4, Search Console), and your CRM and email tools (HubSpot, Salesforce, Klaviyo). Instead of a silo per tool, each with its own login, its own limits and its own version of the numbers, everything sits in one store you control.
Better still, the on-ramp is free and already built. Search Console has a Bulk Data Export that pushes your full performance data to BigQuery every day, with no row limit, so you get the complete long tail of queries and pages. GA4 has its own free BigQuery export that streams raw, unsampled, event-level data. Neither is capped, neither is sampled, and both accumulate for as long as you like rather than rolling off after a year or so. Turning them on mostly means clicking through a few menus.
Two honest caveats. First, the exports are not retroactive: they only contain data from the day you turn them on, with no historical backfill. That is the single best argument for switching them on today, even if you don’t query them for months. Second, Search Console’s export still hides the anonymized query strings themselves, but it keeps their clicks and impressions, so your totals finally reconcile.
This is the direction the whole data industry has moved. As Tristan Handy, founder and CEO of dbt Labs, puts it:
“We need a product like Google Analytics that, instead of plugging into Google’s proprietary back-end, plugs into your data warehouse.”
Tristan Handy, founder and CEO of dbt Labs
Your analytics tools, and your AI, should point at data you own, not the other way around.
More than Google: PostHog, Amplitude, Mixpanel and Snowflake
BigQuery is the obvious home if you already live in the Google stack, since the free exports land straight into it. But it isn’t the only option. A marketing data warehouse comes in two flavours: a general-purpose cloud warehouse you pipe everything into (BigQuery, Snowflake), or a product-analytics platform that doubles as a warehouse (PostHog, Amplitude, Mixpanel).
Here is where each one tends to shine:
| Tool | What it is | Where it shines |
|---|---|---|
| BigQuery | Google’s cloud warehouse | The default if you are in the Google stack. The free GA4 and Search Console exports land straight in, and it is cheap at marketing volumes. |
| Snowflake | Cloud-neutral warehouse | Consolidating many sources across clouds and teams, especially when your data lives well beyond Google. |
| PostHog | Product analytics with a built-in warehouse | Engineering-led teams that want SQL over raw event data (plus session replay and feature flags), with an open-source, self-host option. |
| Amplitude | Behavioural analytics platform | Serious funnel, retention and cohort analysis at scale. Can run warehouse-native on data you already own. |
| Mixpanel | Behavioural analytics platform | Fast, self-serve event reports for teams that want answers without writing SQL. |
How you get your data in matters most when you have no data engineer to build pipelines. On marketing sources, native coverage has improved across the board:
| Platform | Native marketing sources | Coverage |
|---|---|---|
| PostHog | Google Ads, Meta Ads and LinkedIn Ads, plus a dedicated marketing-analytics view | Strong |
| Amplitude | Google, Meta and LinkedIn ad spend, clicks and impressions | Strong |
| Snowflake | Native ingestion via Openflow: Google, Meta, LinkedIn and Amazon Ads, plus Salesforce and HubSpot | Good |
| BigQuery | Google Ads and GA4 natively; Facebook Ads and Salesforce as third-party transfers | Moderate |
| Mixpanel | Ad data via warehouse sync or a custom export, with no one-click ad connector | Thin |
For anything a platform doesn’t cover natively, an ETL layer fills the gap, and that is where the raw breadth lives: Fivetran offers 750+ connectors, Airbyte over 600, and the marketing-focused Supermetrics 150+. The upshot: most of these now pull the big ad platforms natively, and an ETL tool covers the long tail.
These are not rivals. A common setup pairs a product-analytics tool for fast, visual questions about behaviour with a warehouse underneath as the source of truth. Route by job: use the purpose-built tool where its interface saves you time, and the warehouse where you need to own, join and govern the data across all your marketing channels.
Why now is different
The first half of this article was about one barrier: GA4 and Search Console cap, sample and forget your data, and a warehouse doesn’t. But for years the warehouse itself was a barrier too. Setting one up meant a data engineer, fluent SQL, and a cloud bill to match. Three things have changed that.
- It got cheaper and easier. The GA4 and Search Console exports are free and switch on from a settings menu, and they usually stay inside BigQuery’s free tier (10 GB of storage and 1 TB of queries a month).
- You no longer need SQL. AI in marketing analytics has gone conversational: ask a question in plain English and the tool writes the query for you. BigQuery has conversational analytics, Snowflake has Cortex Analyst, Microsoft Fabric has Copilot, and the product-analytics tools have their own: PostHog’s Max AI, Amplitude’s AI Agents, and Mixpanel’s Agent (formerly Spark). None of them ask you to write a line of SQL.
- The competition is healthy. This is no longer a two-horse race between Google and Microsoft. PostHog, Amplitude, Mixpanel and others are competing hard on exactly this: raw data access, and now natural-language AI on top of it. That competition is what keeps pushing the capability cheaper and better.
Put it together: free exports, no caps, no SQL, and real competition. The wall that used to sit between “we have the data” and “we can use it” is largely gone.
Garbage in, garbage out
An AI answer is only as good as the data and the definitions behind it, which is the real reason the warehouse matters.
In the tool-per-silo world, your metric logic ends up scattered across dozens of calculated fields in Looker Studio that nobody remembers writing. One colleague defines a “session” one way, another defines it another way, and when your two dashboards disagree you lose an afternoon working out why. Hand that mess to an AI and it will confidently pick one and never tell you.
A warehouse flips that. Metric definitions live once, in code you can review and document, so the number is the same on Monday as it is on Friday, and the same for the AI as it is for you. You can join first-party data that was never meant to be joined: search queries against product usage, ad spend against pipeline, email engagement against revenue. You can hold the input and output metrics that matter in one place. That is the difference between an AI that guesses and one you can trust to help you decide which marketing campaigns to run next.
The bottleneck was never the AI
The models keep getting better, faster and cheaper. That part is handled, and it is not where you have any edge.
Your edge is the data, and you already own it. Years of first-party search and behavioural data are likely sitting inside your vendor tools right now, doing nothing for you. A marketing data warehouse just moves it somewhere you, your team, and your AI can finally use it, which turns “ask your data a question” from a party trick into a habit.
The models are ready. The only question is whether your data is.
About Growth Method
This is the bet Growth Method is built on: AI marketing that starts with your data, not a blank chat box. It is the agentic marketing platform for B2B teams, with one-click integrations for Google Analytics, Google Ads, Search Console, PostHog, Amplitude, Mixpanel and more, so its agents can plan, launch and analyse campaigns from your real performance data.
A warehouse and an agentic platform are two sides of the same idea: your data belongs somewhere complete, owned and ready to act on. Give the agents access to your data and they surface fresh campaign ideas, prioritise them by likely impact on your goal and tell you what worked and why.
Apply for early access, or book a call to see it running on your own numbers.
Frequently asked questions
Can I use AI to analyse my Google Analytics and Search Console data?
Yes, and increasingly well. Tools like BigQuery’s natural-language analytics let you ask questions in plain English and get forecasts, clusters and trends back. The quality depends entirely on the data you point them at. Fed the capped, sampled data from the GA4 or Search Console API, an AI gives confident answers built on a partial picture. Fed a complete warehouse export, it can reason over years of unsampled first-party data.
What is a marketing data warehouse?
A central store, usually Google BigQuery or Snowflake, where you keep the raw first-party data from all your marketing and analytics tools and query it directly with SQL or plain-English AI. Instead of a separate silo per tool, you get one source of truth that you own and that other tools, including AI, can read.
Is the BigQuery export from GA4 and Search Console free?
Yes. Both Google Analytics 4 and Search Console can export your data to BigQuery for free, with no row caps and no sampling. You only pay BigQuery’s own storage and query costs, which at typical marketing volumes often sit inside the free tier. The catch is that the exports are not retroactive, so they only capture data from the day you switch them on.
Why don’t my Google Search Console numbers add up?
Search Console hides anonymized queries (searches made by only a few dozen users) to protect privacy. It removes them from tables and the API but still counts them in chart totals, so your visible rows never sum to the total, and any query filter drops them. An Ahrefs study found these hidden queries accounted for nearly 47% of clicks.
Do I need a data warehouse for a small website?
Often no. If you never hit the row caps, sampling never kicks in, and 16 months of history is enough, GA4 and Search Console on their own are fine. But the free BigQuery exports are not retroactive, so it is worth switching them on now to bank complete history for whenever you want to point AI at it.
Where do PostHog, Amplitude and Mixpanel fit?
They are product-analytics platforms that increasingly double as warehouses. PostHog offers a SQL editor over your raw event data, while Amplitude and Mixpanel provide purpose-built behavioural analysis and can run warehouse-native or sync to one you already have. Many teams use one for fast visual analysis and a warehouse like BigQuery as the underlying source of truth.