Retrieving a Random Row in ActiveRecord

Note (2025): This post was originally written in 2011. The approaches here are still valid, but in modern Rails (6/7/8), you should consider dataset size and database engine:

Offset method is fine for small datasets.

ORDER BY RANDOM() (Postgres) or ORDER BY RAND() (MySQL) works but is expensive for large tables.

For very large datasets, use database-specific sampling (TABLESAMPLE in PostgreSQL) or precomputed random IDs.

Offset-based method

This approach uses the total row count and a random offset:

offset = rand(Model.count)
Model.offset(offset).first

This is fast and requires no custom SQL. It picks the n-th row by offset, not by ID, so gaps in IDs don’t matter:

# IDs: 1, 2, 4, 8
# If random offset is 2, returns 3rd row (ID=4)

Using `ORDER BY RANDOM()` (Postgres)

For small to medium datasets:

Model.order("RANDOM()").first

⚠️ Caution: On large tables, this can be very slow since it randomizes the entire result set. Always pair with .limit(1).

Modern Alternatives (2025)

Postgres TABLESAMPLE: For very large datasets, this avoids full table scans:
```
SELECT * FROM models TABLESAMPLE SYSTEM (1) LIMIT 1;
```
This picks ~1% of the table randomly, then you limit further.
Precomputed random ID: If you know ID ranges are dense and continuous:
```
id = rand(Model.maximum(:id)) + 1
record = Model.find_by(id: id)
```
But this can return nil if IDs are sparse.

Quick Comparison

Method	Pros	Cons	Best for
Offset with `rand`	Simple, avoids SQL hacks	Requires 2 queries (count + row)	Small datasets
`ORDER BY RANDOM()` (PG) / `RAND()` (MySQL)	One-liner, clean ActiveRecord	Slow on large tables	Small–medium datasets
`TABLESAMPLE` (Postgres)	Very efficient	Approximate randomness	Large datasets
Random ID	Fast if IDs are dense	Fails with sparse/deleted IDs	Dense PK sequences

Wrap-up

For most Rails apps, offset + .first or ORDER BY RANDOM() is good enough.
For large datasets, prefer TABLESAMPLE in Postgres or rethink your randomness strategy.
Avoid RAND() in MySQL on large datasets without LIMIT, as it’s extremely costly.