BigQuery + Querri
Ask about traffic, ad spend and revenue in plain language, and the answer follows your rule for which conversions count. Querri copies the BigQuery tables you pick into one Library with the rest of your business and refreshes them on a schedule.
Querri only reads from BigQuery, with read-only Google access and no key files. SOC 2 Type II, HIPAA and ISO 27001:2022 certified.
Why BigQuery reporting gets hard
BigQuery stores and scans data very well. The friction is everything around it: who can ask, what the numbers mean, and what each look costs.
Marketing and finance know what they want to ask, but getting it out of BigQuery still means writing or checking SQL. Requests pile up with the analysts, and a quick question waits days for a slot.
The GA4 export writes one events table per day, and event parameters sit in a repeated, nested record. Even sessions by source means unnesting and stitching days together.
On-demand queries are billed by the bytes they read, and on a non-clustered table a LIMIT doesn't reduce what's read. Open-ended exploration by many people adds up quietly.
What counts as a conversion, which revenue is real, which campaigns are brand: it's in dbt models, LookML, or one analyst's memory. Two people querying the same table get two answers.
Business teams download results into Sheets or Excel to finish the job. From then on the numbers stop refreshing, and every copy drifts a little further from the source.
What you can finally answer
Each one is a request marketing, growth and finance teams bring to the data team, why it's slow today, and how Querri handles it.
Why it's hard
GA4 knows where a visit came from; the orders table knows who paid. Joining them means matching users across tables and agreeing on which touch gets the credit.
With Querri
Querri joins GA4 sessions to your orders in the Library, and you state once in Facts & rules which touch earns the credit. Step-by-step setup is in the channel to revenue playbook.
Why it's hard
Google Ads reports cost per conversion, but a conversion there is often a form fill. Cost per customer who actually paid needs the ads export joined to real orders.
With Querri
Sync the Google Ads export tables and your orders, then track cost per new customer as a KPI by campaign. Querri keeps its history from the day you start, so last quarter's comparison is one question.
Why it's hard
The raw export is nested and split by day, so even a simple traffic question is an UNNEST query that most marketers can't write.
Why it's hard
GA4 counts purchase events as they fire; finance counts settled orders net of refunds. Both live in the warehouse, and both are right for something.
With Querri
Querri names which revenue it used and shows the other one, so nobody mistakes a tracking number for a booked one. Bring in NetSuite or QuickBooks to tie it to the ledger. See how answers work.
Why it's hard
Spend anomalies hide in a long list of campaigns. Account-level totals look normal while one campaign burns budget for a week.
With Querri
Each night Querri compares every campaign with its own recent history, using arithmetic, and leaves the outliers in your Inbox by morning, through Nightly insights and monitors.
Why it's hard
Agencies and multi-brand teams keep every client in one dataset. One report per client means one filtered query per client, kept up by hand, then pasted into a spreadsheet.
With Querri
Set the report up a single time as a project. Each Monday it reruns on the newest synced data, and every client gets a copy cut to their own rows, by email or as a formatted Excel workbook. The client reports playbook shows the setup.
Why it's hard
Product events land in BigQuery and conversations live in Intercom. Nobody owns the join, so the question stays a hunch.
With Querri
Connect Intercom next to BigQuery and Querri lines up conversations with each account's activity. The ticket volume playbook shows what to look for.
Why it's hard
An assistant pointed at raw tables guesses what each column means, and every person's prompt produces its own version of revenue.
With Querri
Plug Claude or ChatGPT into Querri. The assistant reads from your Library, uses your definitions, and only sees the rows each person is allowed to see in Querri.
BigQuery plus the rest
BigQuery holds your clicks, sessions and product events. The CRM holds the deals, the ERP holds the money, and the targets sit in a spreadsheet. With all four in one Library, a question about ad spend can be answered in closed deals and booked revenue.
BigQuery pairs well with
Answers you can stand behind
"Revenue last month" can mean GA4 purchase events or settled orders net of refunds. Both sit in BigQuery, and they never quite agree. Querri answers with one, names it and shows the other, so marketing and finance stop comparing different things. See how answers work.
Wake up to what changed
Each night Querri measures your KPIs against their own past and goes looking for the campaign or segment nobody has checked lately. What it finds is in your Inbox by morning, and it doesn't change anything unless you say so.
Custom BigQuery dashboards
Describe the dashboard you need, or start from a screenshot of the report your team checks every week. Changes are a sentence away, the charts use the same definitions as your chat answers, and each client or region can be limited to its own rows.
Reports you can stop rebuilding
Traffic, spend and conversions from your GA4 and ads exports, sent every Monday.
Read the playbookOne version per client from a shared dataset, formatted and on schedule.
Read the playbookSessions and campaigns followed through to the orders and deals they produced.
Read the playbookHow long new accounts take to reach first value, from product events.
Read the playbookCost per new customer by channel and campaign, with history kept.
See KPIsA formatted workbook delivered on a schedule, instead of a CSV someone has to download.
See Excel toolsWhat comes over
Querri copies BigQuery tables into its own storage on a schedule, and questions run on that copy. Access is limited to what your signed-in Google account can read.
What you can sync
Full refresh daily, weekly or monthly, or incremental as often as hourly on a column you choose, such as an updated timestamp. Custom queries always reload in full.
Tables stream through BigQuery's Storage Read API in Arrow format and land as Parquet. Views, external tables and custom queries run as query jobs instead. An optional row limit caps how much each source brings over.
Querri asks Google for read-only BigQuery access. It never writes to, alters or deletes your tables or datasets.
You sign in with Google, so there's no service-account key to create or share. Querri keeps a refresh token so scheduled syncs run without you signing in again.
How to connect BigQuery
On Querri's Connect page, type "connect BigQuery" to the agent, or click BigQuery in the catalog.
Grant read-only BigQuery access in Google's own sign-in; your password stays with Google.
Choose one Google Cloud project. Keep every table, pick specific ones, or add custom SQL queries.
Nothing loads until you approve. Choose full refresh or incremental, and how often.
Your BigQuery tables show up in the Library. Ask about them in chat, or build a dashboard or report on them.
BigQuery FAQ
Ask the Connector Agent on Querri's Connect page to set up BigQuery, or choose it from the catalog. Sign in with Google and grant read-only access, choose a project, then pick all tables, specific tables or custom queries. Nothing loads until you approve the load and a schedule.
By default, every table in every dataset of the project you choose. You can instead pick tables by dataset and name, include views and external tables, or save named custom SQL queries as extra sources. Querri can only read what the connected Google account has access to.
Full refreshes are scheduled daily, weekly or monthly. Incremental syncs can run hourly, daily or weekly on a column you choose, loading only rows above the last value seen. Incremental syncs don't catch deleted rows; a full refresh does. A manual sync is there whenever you need fresher numbers.
No. Querri only reads from BigQuery. The Google access it asks for is read-only, so it can't create, update or delete your tables or datasets. Your records stay exactly as they are; Querri works on its own copy.
Querri sees only the projects, datasets and tables the Google account you sign in with can read, so connect with an account whose access fits what you want in Querri. Once the data is in Querri, row access comes from Querri's own sharing and row-level rules; a person's Google Cloud roles don't follow them in.
No. Querri copies the tables you choose into its own storage on a schedule, and questions, dashboards and reports run on that copy. That also means answers are as fresh as the last sync, so pick a schedule that matches how current the numbers need to be.
It can. Syncs read from BigQuery, and views, custom queries and incremental checks run query jobs in the project you selected, which bills that project. Querri doesn't set a byte cap or show a cost estimate before a job runs, so check what a custom query scans in the BigQuery console before you save it.
Connect both, and Querri lines up your GA4 or Google Ads data with HubSpot contacts and deals, for example by UTM campaign or email. You can then follow spend through to pipeline. The query behind each answer shows how records were matched, so anyone can check the join.
Connect both, and Querri joins warehouse tables such as web sessions or product usage to Salesforce accounts and opportunities in the Library. State once what counts as an active account or a qualified visit, and every answer, dashboard and report after that uses the same definition.
Connect the project that holds your GA4 export. Because the export stores event parameters in nested records, many teams save one custom query that pulls out the fields they use. Tell Querri once what counts as a conversion, then ask, or schedule the report to arrive every Monday.
Sync the Google Ads export tables and your orders or CRM deals. Querri matches spend to the customers each campaign produced, with your rule for what a new customer is, and can track the result as a KPI so you see how it moves month to month.
Yes. Save the analysis as a project and give it a schedule. Each run uses the latest synced tables and delivers a formatted Excel workbook by email, split into one version per person or client when you need that. Nobody has to download a CSV again.
Yes. Describe what you want to see, or begin with a screenshot of a report you already rely on. Edits are plain-language requests, charts can combine BigQuery with CRM and finance tables, and access can be narrowed so each person or client sees only their rows.
Yes. Once Querri is connected to Claude or ChatGPT, the assistant pulls from your Library and follows your definitions, limited to the rows that person can see in Querri. It works from the synced copy, so a question in Claude doesn't send a new query to BigQuery.
Not today. The BigQuery connector uses Google sign-in only, with two read-only scopes, and keeps a refresh token so scheduled syncs keep running. If you want a dedicated identity, sign in with a Google account set up for Querri and grant it access to the datasets you want.
No. Each BigQuery connection points at one project, which you pick from a list after signing in. To bring in data from another project, add a second connection. Both land in the same Library, so questions can still span them.
Custom queries can't sync incrementally, there's no Test connection button, and Querri doesn't add partition filters or byte limits of its own; the only cap is an optional row limit per connection. It never writes back to BigQuery, and questions run on the last synced copy rather than live.
Querri is SOC 2 Type II, HIPAA and ISO 27001:2022 certified. Access is granted through Google's own sign-in with read-only scopes, so no password or key file passes through Querri. Data is encrypted in transit and at rest, and it is never used to train a model.
Keep exploring
Sign in with Google and pick your tables. Querri only reads your data; your tables and datasets stay exactly as they are.