Product
Freshness checks
Tessel learns when each table normally lands and warns you before a late load reaches a dashboard.
How it works
A pipeline can succeed and still be wrong: the run finished, but the source it read was yesterday’s. Freshness checks catch that. Tessel watches when each table is updated and how many rows arrive, learns the normal pattern over two weeks, and flags anything that drifts from it.
You don’t set thresholds by hand. A table that loads hourly on weekdays and once on Saturdays gets a different expectation on each day, and month-end spikes don’t trigger false alarms.
For the tables your dashboards read, defined as dbt exposures or tagged in Tessel, you can set a promise in plain words, such as “ready by 7:00 Dubai time”. Tessel warns the owner when that promise is at risk, not after it is broken.
Step by step
-
Learn
Fourteen days of update times and row counts give each table its own expected pattern.
-
Watch
Every load is compared with that pattern, allowing for weekends, holidays and month end.
-
Warn
When a table your dashboards read is at risk of missing its promise, the owner hears first.
By the numbers
- of history to learn each table
- 14 days
- Starts alerting on day one with simple rules
- average warning before a dashboard goes stale
- 25 min
- Across Team plan customers in 2026
- thresholds to set by hand
- 0
- Override any table if you want to
Freshness checks learned our load schedule on their own. Finance stopped messaging me at 9:05 every morning.
Questions about freshness checks
Does Tessel query our tables to check freshness?
It reads table metadata such as last-modified times and row counts from your warehouse’s information schema. It never reads the rows themselves.
How does it handle holidays?
Pick a holiday calendar per project. Tessel also learns from history, so a quiet Eid week doesn’t surprise it twice.
Can we set our own freshness rules?
Yes. Any table can have a fixed rule, such as “updated within 2 hours”, alongside or instead of the learned pattern.
See it on your own pipelines
A 30-minute demo on a copy of your project, or start free with up to 25 models.