How an offline activity audience works
Plenty of sales never touch a website. Someone sees an ad, walks into the showroom and pays cash, or rings the clinic and books over the phone. Meta cannot see any of that, so you tell it: you export those transactions from your own records and upload them to an offline event set, either as a file or through a connected system.
Each row carries identifiers such as an email address or phone number, hashed before they leave your computer so the raw details are never sent. Meta hashes its own records the same way and matches the two. An offline activity audience is then the list of people who matched a chosen offline event within a chosen lookback window.
Two things follow from that. The audience is only as complete as your upload, and it only exists for people whose contact details you actually captured. A sale recorded as a walk-in with no name will never appear.
Why offline activity audiences matter
For most businesses in Nepal — dealerships, hospitals, education consultancies, anyone selling on cash on delivery — the transaction that matters happens away from the browser. Without uploading it, Meta optimises towards form fills and messages, which are only proxies. Feeding the real outcome back in changes what the system learns and what you can target.
The practical uses are ordinary and valuable: exclude people who have already bought so you stop paying to reach them, build a win-back list from customers who bought a long time ago, and seed lookalikes from people who paid rather than from people who merely enquired.
Common mistakes with offline activity audiences
The first is a low match rate caused by messy data. Phone numbers stored without a country code, names in one field, spaces and dashes left in — every inconsistency costs matches, and the audience ends up a fraction of the sales you uploaded.
The second is uploading once and calling it done. An offline event set that is fed after a launch and then forgotten produces an audience that ages out and disappears, usually just as someone starts relying on it.
The third is treating the upload as a measurement fix rather than a data responsibility. You are sending customer records to a third party, so consent, retention and who does the exporting all need deciding before the first file, not after a complaint.
How to act on it
Fix the source first. Capture a phone number or email at the point of sale, store it in one consistent format, and include a transaction date and value in the export. Hash and upload on a regular schedule rather than by hand, so the audience never goes stale.
Then use it deliberately: exclude recent buyers from acquisition, retarget the ones with a repeat-purchase cycle at the right moment, and use paid customers as the seed for your best lookalike. If you are already sending web conversions through the Meta Pixel and Conversions API, keep the naming consistent across both so a single sale is not counted twice. It is worth understanding offline conversions as a whole before you build audiences on top of them.
