How offline conversions work
Plenty of business never touches a checkout. Someone sees an ad, calls the number, visits the showroom, negotiates, and pays in cash or by bank transfer. The ad platform sees the click and then nothing. Offline conversion upload closes that loop: you send the platform a file, or connect your CRM, listing the outcomes that happened away from the site.
Each row carries the customer’s identifying details, the time the outcome happened, what it was, and usually its value. Those details are hashed — converted into unreadable strings — before being sent, and the platform hashes what it holds the same way and looks for matches. Rows that match a person who saw or clicked an ad within the attribution window become reported conversions. Rows that match nobody are simply discarded.
Google, Meta and LinkedIn all support some version of this, either as a manual file upload or as a live connection from a CRM.
Why offline conversions matter
They correct the wrong lesson. Without them, a lead generation account learns from form fills, and the campaigns that produce the greatest number of forms look best. Feed back which enquiries actually became customers and the picture usually reorders itself, because the cheapest leads are often the least likely to buy.
This matters unusually much in Nepal, where a great deal of buying is finished on the phone, over Viber or WhatsApp, or in person, and where card payment online is not the default it is elsewhere. A business here that judges advertising only on what completes on the website is judging a fraction of its trade.
They also unlock value-based bidding for service businesses, since an uploaded deal value lets the platform pursue revenue rather than enquiry volume.
Where it goes wrong
Match rate is the first hurdle, and it is usually a data problem rather than a platform problem. Phone numbers stored without a country code, email addresses collected inconsistently, names in one field, and details normalised differently from how the platform expects them all reduce matches.
The second failure is timing. Every platform has a window beyond which an outcome can no longer be attributed, so a business that exports its sales once a quarter will find much of the file too old to count.
The third is uploading the wrong stage. Sending every deal marked “won” in a CRM that nobody keeps updated teaches the platform from fiction.
What to do about it
Start by capturing a usable identifier at the point of enquiry — an email address or a phone number with its country code — and store it in a consistent format. Then decide which CRM stage represents a real outcome, and get the sales team to keep that stage honest, because everything downstream depends on it.
Upload frequently rather than in occasional large batches, and connect the CRM directly if the tool allows it. Include deal value where you have it. Check reported match rate after each upload and treat a poor one as a data-hygiene task. If enquiries are still landing in a spreadsheet nobody trusts, fix that first — connecting the CRM properly is what makes uploads worth doing, and a clean tracking and analytics setup is what lets you check the result.