Offline Conversion Import for Long-Lag Sales Cycles

Google Ads will not accept an offline conversion whose click is more than 90 days old. A placement-to-start cycle in staffing runs 40 to 60 days. A district-level education contract runs longer. A crypto funnel that ends at a funded deposit after KYC review often sits at three weeks with a tail past sixty days. So the first question is never which tool uploads the file. It is what share of your closed records still carry a click identifier inside the window, and I measure that before quoting the rest of the work. I have declined projects where the answer was under 20 percent, because the fix belonged in lead capture and would have been fraudulent to bill as an import pipeline. Capture at the click: The identifier is written to the lead record at creation, not reconstructed later. gclid, plus gbraid and wbraid for iOS app traffic where no gclid exists, fbclid rendered into the fb.1.timestamp.fbclid form your Meta events already use, msclkid, ttclid, li_fat_id for staffing accounts that run LinkedIn. These live as fields on the candidate or lead object in the ATS or CRM, with the timestamp, because the timestamp is what decides whether the record is still eligible on the day it closes. Identity hygiene, where most match rates are lost: Phone numbers get normalised to E.164 before hashing, and in Mexican systems that means removing the legacy 1 that still sits after 52 in older records, which silently fails every match it touches. Email is trimmed, lowercased, and hashed with SHA-256 on our side of the boundary. RFC and CURP are not identifiers for this purpose and never leave your systems. Where records were typed by a recruiter into a free-text field, I say so and quantify it rather than hashing rubbish and reporting a low match rate as a platform problem. Consent as a stored field, not a banner: Every uploadable record carries a consent timestamp, the purpose it was given for, and the version of the notice in force at the time. No stored consent means no upload, with no exception and no override switch in the job. This is also the only version of this pipeline that survives a due diligence review, which matters more than usual in regulated funnels. The upload path: Google Ads API conversion upload for click-identified records, and the user data job route for enhanced conversions for leads where the click identifier is gone but hashed contact data exists. Meta receives offline events with the true event time. LinkedIn and TikTok where the spend justifies it. Jobs run on a schedule you can see, with failures surfaced to a channel rather than to a log nobody opens, and every job reports its match rate per source. Below roughly 60 percent I stop tuning the uploader and go back to collection, because that is where the loss actually is. When a record counts as closed: In staffing, a placement inside the replacement guarantee window is provisional revenue, so it uploads at guarantee expiry or at whatever stage your finance team recognises, decided with them rather than by me. In crypto, a registered account is not a conversion; a first funded deposit after KYC clearance is. In education, a signed contract without a cleared deposit is not one either. Not included: Media buying and campaign management. CRM implementation or data cleanup at scale. Buying you a CDP, which this work does not need. Lead scoring models. What I refuse: Raw email or phone sent to any platform. Uploading revenue that has not been collected. Uploading a record with no consent trail because a deadline is close. Who this is not for: Funnels that close inside a week, where enhanced conversions already covers you, and teams with no engineering access to their CRM's webhooks.