What if running more experiments could actually reduce your cost per A/B test? At Mintminds, we believe the best experimentation programs scale cost-efficiently, lowering the cost per experiment. But a hidden cost killer is lurking in most GrowthBook setups: BigQuery query inefficiency.
In this case study, we’ll show you how we partnered with The Social Hub and GA4Dataform by Superform Labs to slash BigQuery costs by 81.8% while actually improving data refresh speeds and monitoring capabilities.
GrowthBook
Cost Structure
At Mintminds, our mission is simple: build high-quality experiments with reliable data and analysis. GrowthBook’s pricing model allows for a setup where the more you test, the less you pay per experiment. But to optimise costs, you need to understand where money actually flows. Let’s break it down:
Fixed Costs (pricing, as of Nov 2025)
Variable Costs (GrowthBook Cloud):
Self-Hosting Alternative: You can eliminate CDN costs by self-hosting GrowthBook for $11-50/month (depending on your infrastructure choice).
BigQuery
The elephant in the room
Regardless of hosting choice, BigQuery becomes your primary variable cost when using GA4 as your data source. For companies running active experimentation programs with daily updates, unoptimized BigQuery costs can easily reach $200-400/month.
Important context: Even with BigQuery costs included, GrowthBook remains dramatically cheaper than traditional alternatives like Convert ($3,500/month) or VWO ($4,300/month) at comparable traffic levels. GrowthBook is already the smart financial choice. Our optimisation makes it unbeatable.
The problem
GA4 structure wastes BigQuery resources
The default GrowthBook BigQuery integration queries GA4’s standard events_* and events_intraday_* tables. These tables store event parameters in nested structures, forcing BigQuery to process far more data than necessary.
For example when you’re running experiments with:
..BigQuery has to scan through nested arrays and repeated fields to extract the specific event parameters you need. You’re paying to process gigabytes of data when you only need megabytes of relevant information.
GrowthBook does allow custom fact tables and metrics to select only relevant events and parameters. This helps, but optimisations plateau quickly because you’re still querying nested GA4 tables.
Enterprise customers get access to:
But for Pro license users, we needed a different approach.
The solution
GA4Dataform's flattened datasets
At #CH2024 (the conference formerly known as Conversion Hotel), I spoke with Jules Stuifbergen from Superform Labs about this exact challenge. Jules introduced me to GA4Dataform, which offered an elegant solution:
What GA4Dataform Does: The Core Version (free!) creates a customised, flattened dataset optimised for the type of queries that GrowthBook uses.
| Feature | Benefit |
|---|---|
| Fully flattened structure | No nested fields = dramatically faster queries |
| Smart partitioning and clustering | Restricting queries by date and event names will decrease the number of rows scanned |
| Smaller data footprint | Less data processed = lower BigQuery costs |
| Daily automated updates | Fresh data from GA4 events table is appended to the table, using incremental logic |
Key insight: Even though you’re creating a new dataset in BigQuery (which feeds from the generic GA4 table), the flattened structure makes it cheaper to generate AND cheaper to query than repeatedly querying GA4’s nested tables.
Bonus benefit: This same optimised dataset can be used for all your other BigQuery reports and dashboards, compounding the savings.
The test setup
Rigorous A/A experiment
We partnered with The Social Hub—a hybrid hospitality brand offering hotel rooms, co-living spaces, coworking facilities, and creative playgrounds across Europe—to validate this approach with real data.
Laura Semeraro, Digital Analyst at The Social Hub, immediately saw the potential:
“This wouldn’t just reduce GrowthBook costs—it would optimise all our BigQuery reports and dashboards.”
Implementation Steps
1. GA4Dataform Setup – Laura installed GA4Dataform Core (free version). The custom event parameters from GrowthBook were added to the configuration (experiment ID and variation ID). With the daily schedule enabled, GA4Dataform automatically updates the flat events table incrementally.
2. GrowthBook Configuration – Created a new assignment query (for counting experiment visitors). Built fact tables for key conversion events: Add-to-cart and purchase events. Purchase events
3. A/A Test Design – We ran two identical experiments simultaneously:
Configuration:
The Only Difference:
Experiment A: Default GrowthBook queries (nested GA4 tables)
Experiment B: Optimised queries (flattened GA4Dataform dataset)
4. Measurement – GrowthBook usage is automatically labelled in BigQuery, allowing us to track:
Test duration: 1 week
This gave us an objective, apples-to-apples comparison.
The Results
Massive Cost Reduction
When the results came in, we had to verify the numbers multiple times to ensure accuracy.
A Whopping 81.8% Cost reduction and a massive query speed improvement, too. By using the GA4Dataform flattened dataset instead of the default GA4 nested tables, we reduced BigQuery data processing by more than four-fifths.
|
Benefit |
Impact |
|---|---|
| Update experiment results more frequently | Better SRM and MDE monitoring without budget concerns |
| Run updates faster | Flattened queries execute in a fraction of the time |
| Scale experiment volume | The “more you test, less you pay” promise becomes reality |
| Optimize other analytics | Use the same flattened dataset for all BigQuery dashboards |
The compounding effect: Lower per-experiment costs + faster refresh rates = exponentially better experimentation program ROI.
Query Cost Reduction
Faster query speed
Savings
The bigger picture
To show the full impact, let’s take a real-world example from one of our clients with 2.6 million unique users/month and running 5-7 experiments a month. In this example, we are running the GrowthBook JS SDK on Cloudflare pages, which means no limitations on the number of tested visitors for free. Yes, you read it right… for free!
The variable GrowthBook costs are:
Fixed GrowthBook Pro costs for a team of 5 members: 5 * $40 = $200
| Platform | Monthly Cost | Annual Cost | vs. GrowthBook Optimized |
|---|---|---|---|
| Convert.com Pro | $3,488 | $41,856 | 1,050% more expensive |
| VWO Pro | $4,308 | $51,696 | 1,320% more expensive |
| GrowthBook (Unoptimised) | $546 | $6,552 | 80% more expensive |
| GrowthBook (Optimised) | $303 | $3,640 | Baseline |
The bottom line: The 82% BigQuery reduction transforms GrowthBook from “very affordable” to an offer you simply can’t refuse.
Conclusion
Accessible enterprise-grade experimentation
This case study demonstrates that you don’t need an Enterprise license to achieve exceptional BigQuery efficiency with GrowthBook.
By combining:
You can build a cost-effective, high-performance experimentation stack that rivals Enterprise setups—at a fraction of the price. The cost reduction we achieved with The Social Hub isn’t an outlier. It’s the new baseline for optimised GrowthBook implementations.
Big thanks
To our partners
This breakthrough wouldn’t have been possible without:
Laura Semeraro and the team at The Social Hub – Thank you for your partnership, analytical rigour, and willingness to experiment on experimentation itself. Your commitment to data-driven decision-making continues to inspire us.
Jules Stuifbergen and Superform Labs – Your GA4Dataform tool is a game-changer for the GrowthBook community. Thank you for building accessible solutions that democratise advanced analytics optimisation.
Let's collaborate
Cases
More projects
From Browse to Buy: USP Text on Slider Drives 13% Add To Cart Rate Uplift
Experimentation on the Edge unlocks cost per experiment reduction
From Generic Share to WhatsApp: The Trigger That Boosted RPU by 23%
How The Social Hub slashes experimentation costs by 82%
Direct contact
Nijverheidstraat 11-1E
7511 JM Enschede
the Nederlands
+31 53 234 0444
