Sync frequency usually gets set once, when an integration is built, and then left alone. It looks like a simple dial: turn it up for fresher data, turn it down to reduce load. But frequency fails quietly at both ends.
A sync that runs too often can use up a base’s API capacity and create more chances for data collisions on fields that more than one system writes.
A sync that runs too rarely leaves other systems working from old data. In both cases, every run can succeed and every log can stay clean.
Update frequency is one of the five factors in InAir's Integration Load framework, part of InAir's broader guide to enterprise Airtable integrations. This post covers what goes wrong at each end, where the workable middle sits, and a check you can run on your own landscape.
Frequency means different things depending on how a sync is triggered.
Most of this post is about scheduled syncs, because that's where the interval is a decision someone makes. For event-triggered syncs, the equivalent questions are whether the real run rate, taken from run history, still matches what the team assumes, and whether the sync is set up to safely handle a spike in volume, since a bulk update in the source system can trigger thousands of sync requests at once.
The risk of running often depends almost entirely on what each run does. A sync that pushes only the records that changed can run frequently at little cost. A sync that rewrites every record on every run is where frequency becomes expensive.
Take a CRM that syncs 500 customer records into Airtable by rewriting all of them on each run. Airtable's API accepts up to 10 records per request, so each run needs 50 requests. Two things follow.
It takes up the base's shared capacity. Airtable allows 5 requests per second per base, and every system calling that base draws from the same limit. Fifty requests is at least ten seconds of the base's full capacity. Run that sync every ten seconds, and it leaves almost no capacity for any other system. Requests over the limit are rejected with an HTTP 429, and the sender, along with every other system sending requests to that base, has to wait 30 seconds before its requests succeed again.
That 429 typically shows up in the sending system's logs, not in anything the Airtable operator looks at, which is why throttling often goes unnoticed. InAir describes this congestion stage in What Breaks First at Enterprise Scale: the base doesn't fail loudly, and the backlog is often found only when someone asks why a record hasn't updated. Getting that signal to the right person is covered in alert ownership.
It can use up a plan's monthly API calls. On Team plans, Airtable caps API usage at 100,000 calls a month. Business and Enterprise Scale plans have no monthly cap. The same 50-request sync running every two minutes makes 720 runs a day, or 36,000 requests. On a Team plan, that one sync would use up the month's allowance in under three days.
For the people working in Airtable, none of this shows up as an error. Runs complete, rejected requests are either retried or dropped in another system's logs, and the records look current.
Running less often reduces the load, and it reduces the likelihood of a data collision on shared fields. But it creates a different failure: stale data.
Suppose the CRM syncs to Airtable once a day at 9 am, and Airtable syncs to the billing system once a day at 8 am. A customer's billing email is changed in the CRM at 10 am. It misses that morning's sync, reaches Airtable at 9 am the next day, misses that day's 8 am billing sync, and reaches billing at 8 am on the third day. That's 46 hours, and any invoice billing sends in that window goes to an address the customer has already replaced. Every run in that chain succeeded.
A long interval also widens the window in which both systems can change the same field before either one sees the other's value. So running rarely doesn't remove conflicts. It mostly delays them, and leaves them to be resolved by whichever sync happens to run next.
For most landscapes working on live, regularly updated data, a sync every couple of minutes is a reasonable place to land, and it’s usually the shortest interval worth setting. It keeps data close to current without using up capacity the way a sync every few seconds can. Many syncs don’t need live data at all and run perfectly well every few hours or once a day.
That interval holds up best when each run pushes only the records that changed since the last one. In the example above, if five of the 500 records changed in a two-minute window, the run needs one request instead of 50. At 720 runs a day, that's 720 requests, or about 21,600 a month, comfortably inside a Team plan's allowance and leaving plenty of the base's capacity for other systems.
How to build a sync that detects changes is covered in change detection for Airtable integrations. For sustained high volume, InAir's guide to operational resilience in Airtable covers batching updates into requests of up to 10 records, and routing traffic through a queue that paces writes under the rate limit. Payload size is a separate question from frequency. Webhook Payload Optimization covers when trimming the fields in a payload matters, which is mainly for very wide records or heavy field types sent outbound.
This check finds the places where frequency is quietly causing stale data.
The fix for a "too slow" row isn't always a shorter interval. Sometimes it's an event-triggered sync for that one field, or a path that skips an intermediate system. The check tells you where the gap is. It's also one of the five inputs in the Integration Load Scorecard.
The interval on a sync is easy to set and easy to forget. At one end, it uses up capacity and multiplies overwrites. At the other, it leaves systems working from old data. Neither end shows up as an error to the people relying on the data, which is why the question is worth asking on purpose: how often should this run, what does each run touch, and how old can the data get before someone is working from the wrong value?
If you want to find where frequency is quietly costing you across a live landscape, and fix it without disrupting the syncs already running, that's the work InAir does. Schedule a Discovery Call with InAir to find out what your integrations are actually carrying.