A/B Testing Database Storage Per-User Calculator FAQ
Answers to common questions about estimating per-user A/B test assignment and event data storage.
Use these answers to understand the calculator inputs, results, assumptions, and common capacity-planning considerations.
General questions
Basic information about the storage estimate.
What does this calculator estimate?
It estimates database storage for one user's experiment assignments and retained exposure and conversion events.
Is the result a total database size?
No. It is a per-user estimate. Multiply it by an appropriate user count and add other storage components for a broader estimate.
What data is excluded?
The estimate excludes shared experiment definitions, dashboards, backups, replicas, archives, and data warehouse copies.
Inputs and event volume
How the main inputs affect the calculation.
What are active experiments per user?
This is the average number of concurrent experiments for which a typical user has an assignment record.
What is an exposure event?
An exposure event is a stored record that indicates a user saw, was evaluated for, or was otherwise exposed to an experiment or variant.
Why are conversion events entered per experiment?
The calculator combines conversion activity with exposure activity for each active experiment before estimating retained event records.
Can exposure events per day be a decimal?
Yes. A decimal represents an average rate across users or days, such as 0.5 events per day.
Formula and retention
How assignment storage and event storage are combined.
How is assignment storage calculated?
Active experiments per user are multiplied by the average assignment record size.
How is event storage calculated?
Active experiments are multiplied by daily exposure plus conversion events, retention days, and average event record size.
Does retention affect assignment records?
Not in this model. Assignment storage is calculated as one record per active experiment, while retention applies to event-level records.
What happens if I double retention days?
With all other inputs unchanged, retained event storage doubles. Assignment storage remains unchanged.
Accuracy and planning
How to interpret a capacity-planning estimate.
How accurate is the estimate?
It is as accurate as the averages you enter. Actual usage can differ because of compression, schema design, indexes, replication, and data lifecycle processes.
What should the overhead percentage cover?
It can cover indexes, row headers, metadata, storage-engine overhead, and other costs outside the raw record-size inputs.
Should I include replicas in overhead?
It is clearer to calculate primary storage first and add replica capacity separately, since replication topology can vary.
Should I use compressed or uncompressed record sizes?
Use sizes that best reflect the storage layer being planned. Be consistent about whether your inputs already include compression effects.
How do I estimate total A/B testing storage?
Multiply the per-user result by the relevant number of retained users, then add separate capacity for replicas, backups, shared configuration, and other data stores.
Explore Related Questions
Ready to see what you can calculate?
Open the calculator and get personalized results in seconds.