Export your A/B test data to BigQuery

Every night, Mida puts your raw experiment data into your own Google Cloud Storage bucket, ready for BigQuery. Join it with your orders, CRM and product data, and analyze tests your own way.

Data Warehouse Data out Server-side Works with Mida A/B Testing & Personalization

About BigQuery

BigQuery is Google Cloud's data warehouse. It runs SQL over very large datasets, and can query files in Google Cloud Storage directly as external tables.

How the BigQuery integration works

Mida keeps a copy of each project's experiment data as it happens. Shortly after midnight UTC it adds the previous day to your bucket as four tables, one file per day: exposures (each visitor put into a variant, and when), events (goal conversions and your own events, such as purchases, with their properties), results_daily (the totals per variant, as your Mida report shows them that day) and tests (your tests with their variants and goals). In BigQuery, one set of external tables over the bucket picks up each new day on its own; the help article has the SQL to create them. The first delivery also includes everyone already in your live tests, so a test that was running before you turned the export on can be analyzed in full.

Requirements

  • A paid Mida plan.
  • A Google Cloud Storage bucket, in the same location as your BigQuery dataset (for example US or EU).
  • Access to change the bucket's permissions in Google Cloud.
  • A BigQuery dataset you can create tables in.
  • Owner or Admin role in Mida.

How to use BigQuery with Mida

  1. In Google Cloud, give the Mida service account shown in Mida the Storage Object User role on your bucket.
  2. In Mida, open Integrations and choose BigQuery under Data Warehouse.
  3. Enter the bucket name and, if you like, a folder, then click Check access and turn on.
  4. In BigQuery, run the SQL from the help article once to create external tables over the bucket. New days show up on their own.

Use cases

Measure tests against revenue you trust

Join exposures with the orders in your own database, so a test is judged on refunds, repeat purchases or margin, not only on the conversion Mida tracked.

Follow tested visitors over time

See what visitors in each variant did weeks later: whether they upgraded, churned or came back. Exposures keep the first time each visitor saw a test, so long-term effects are easy to measure.

Report on every test in one place

Put results_daily into your dashboards next to the rest of your company's numbers, and track every running test day by day without opening Mida.

Run your own analysis

Segment results by any customer attribute you hold, or apply your own statistics. Exposures and events share the same visitor ID, so the join is a single line of SQL.

Frequently asked questions

What data does the export include?

Four tables: exposures (each visitor put into a variant), events (goal conversions and the events you track, with their properties and page), results_daily (visitors, conversions and revenue per variant for each live test) and tests (names, variants and goals). Page views and engagement pings are left out to keep the files small.

How often is the data delivered?

Once a day. Shortly after midnight UTC, Mida adds the previous day's files and a small manifest that lists how many rows each table has, so you know the day is complete.

Do I get data from before I turned it on?

The first delivery includes everyone already in your live tests. Events start from the day you turn the export on.

Can I export to Amazon S3 or Azure?

Not yet. The files go to Google Cloud Storage, which BigQuery reads directly.

Which Mida plans include it?

Every paid plan. Many testing tools keep a warehouse export for their enterprise plans only.

Does it slow down my site?

No. The export runs on Mida's servers. Nothing changes on your site or in the Mida tag.

Ship A/B tests 10x faster — connected to your stack

Mida is the lightweight A/B testing and personalization platform that integrates with the tools you already pay for.

Decorative graphic Decorative graphic Decorative graphic Decorative graphic