Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Standard Operating Procedures

Offline Conversion Upload SOP

A standard procedure for uploading offline outcomes from the CRM to ad platforms: which outcomes, export, consent, format, upload, read the results, reconcile.
SOPadvanced9 min readUpdated

Owner, trigger and scope

Marketing operations owns the procedure: it prepares the file, runs the upload, reads the results and keeps the log. The paid media manager confirms which conversion actions each outcome feeds and which of them bidding relies on, and receives the note at the end of every upload. The CRM owner keeps the stages, the stored identifiers and the consent fields clean, and fixes the causes of rejected rows at source.

The trigger is the agreed upload day for outcomes recorded in the CRM, such as qualified leads, booked appointments or won deals, where no automated connection sends them. The day itself comes from the plan, not from this page: the measurement plan for lead generation decides which stages are imported, how often and with what value, and the procedure runs on whatever rhythm that plan sets.

Several neighbouring jobs belong elsewhere:

SituationWhere it is handled
Designing the connection: capturing identifiers at the form, choosing imports, enhanced conversions or a server routeConnecting the CRM to ads and analytics
Won deals sent automatically the moment a deal closesBuilding a closed-won workflow
Reconciling a closed month of platform conversions against salesReconciling Google Ads conversions with sales
Conversions or events stop recording altogetherTracking outage response SOP
Reporting the month's results to the businessMonthly reporting pack SOP

The procedure is written around Google Ads, whose help pages document offline conversion imports and the leads variant of Enhanced Conversions, but the same routine applies to any platform that accepts outcomes by file. The SOP library routes the other recurring measurement jobs.

Before you start: inputs and access

InputOwnerNotes
The list of outcomes to upload and the conversion action each one feedsPaid media managerWritten down once and changed only through the plan, never decided on the day
The CRM export of outcomes since the last uploadCRM ownerWith the stored identifiers, the stage-change times and any values
The consent status of each recordCRM ownerAs captured with the lead, not inferred later
The platform's import template and access to the accountMarketing operationsThe current template from the help centre and a login that can upload
The last reconciliation of platform conversions against CRM recordsMarketing operationsThe upload log and the most recent monthly comparison

Keep the guide to connecting the CRM to ads and analytics open. As it states from Google's help pages, an upload carries the click identifier or hashed customer data, the conversion name, the conversion time and an optional value and currency. The current file formats, import windows, time-zone rules and diagnostics screens live in Google Ads Help and change, so this procedure points at them rather than restating them.

The procedure

  1. Confirm which outcomes feed which conversion actions

    Before exporting anything, read the written list of outcomes and the Conversion Action each one feeds. The guide to conversion events across platforms explains why the event decides what the budget buys and why outcomes the platform cannot see, such as a sales-qualified lead, are the ones worth sending. If the list has changed since the last upload, the change goes through the paid media manager and the measurement plan first. If the account has old or overlapping import actions, cleaning up conversion actions comes before the next upload, not during it.

  2. Export the outcomes since the last upload with identifiers and times

    Export from the CRM every record that reached an agreed stage since the period the last upload covered, with its stored click identifier or the contact details the leads route matches on, the time of the stage change and the value where one is sent. The stages must be the ones the lead stages and pipeline guide asks a business to define, with exit criteria, so that a qualified lead means the same thing every week. The CRM connection guide explains where those identifiers come from; the export only collects them. Use the stage-change time, not the export time, because the upload has to say when the outcome happened.

  3. Remove records without the consent the upload needs

    Filter out every record whose stored consent state does not allow its data to be sent to the ad platforms. The CRM connection guide's consent section sets the principle: capture the consent state with the lead, check it before any outcome leaves, and send the minimum. Treat a missing consent value as a no, and count the removed rows so the log shows how many outcomes the upload could not carry. This is general practice, not legal advice; the platform's own requirements are in its help pages and the organisation's privacy advice decides the rest.

  4. Format the file to the import template

    Map the remaining rows onto the platform's current import template: the identifier or match key, the conversion action's exact name, the conversion time in the format and time zone the template asks for, and the value and currency where the action uses them. Leave out every field the upload does not need; the CRM connection guide is explicit that names, notes and job titles do not belong in an outcome upload. Check the file against the upload log to remove any record already accepted in an earlier upload, so that one outcome never counts twice.

  5. Upload and read the import results

    Upload the file and read the platform's import results before closing the session: how many rows were accepted, how many were rejected and the reasons given. The enhanced conversions guide explains the leads variant, where outcomes are matched by hashed data instead of a click identifier, and how to read the diagnostics on each conversion action. Save the results with the file. Remember that uploaded outcomes arrive later than tag conversions, so the platform's reports for recent days keep changing after each upload; the Conversion Lag between click and outcome is why.

  6. Fix rejected rows at the source, not in the file

    Group the rejected rows by reason and find the cause in the CRM or on the form, not in the spreadsheet. A missing identifier usually means a form or landing page stopped storing it; a wrong time usually means a field holds the wrong value or zone; an unknown action name means the list and the account have drifted apart. The tutorial on auditing and documenting CRM fields shows how to find what writes and reads each field, which is the quickest way to the cause. The CRM owner fixes it, and the corrected records join the next upload, marked in the log.

  7. Compare uploaded outcomes with the CRM count

    Compare the outcomes the platform recorded for the period with the CRM's own count of the same outcomes. They will not match exactly, because some rows are rejected, some fall outside the window and some leads never came from an ad. The tutorial on reconciling Google Ads conversions with sales lists those causes and sizes them for a closed month; this step only checks that the gap for this upload sits close to the gap the last reconciliation found. A gap that suddenly widens is the earliest sign of a broken capture, and if counts collapse altogether the tracking outage response SOP takes over.

  8. Record the upload and tell paid media what changed

    Write the upload into the log and add anything unusual to the measurement change log, so that a later reader knows why a week's imported conversions jumped or fell. Then send the paid media manager a short note: the period covered, the outcomes by action, the rows rejected and the comparison. Where bidding depends on the imported action, a late, short or oversized upload changes what the strategy learns, and the guide to bidding with few conversions explains why accounts with rare outcomes need that context before anyone moves a target. Value-based reading of the results can be checked with the ROAS calculator.

A solar installer's weekly upload (illustrative)

Illustrative example — not a real result

Rejected rows traced to a form, not to the file

Wexatrov Solar, an invented solar panel installer, uploads two outcomes every Monday: survey bookings, which feed the primary bidding action, and signed contracts with their value, which feed a secondary reporting action. One Monday the export holds 64 survey bookings and 9 signed contracts. Consent filtering removes 5 bookings. Of the 59 booking rows uploaded, 17 are rejected for a missing click identifier, against 2 or 3 in a normal week. Rather than deleting the rows or guessing identifiers, marketing operations groups them by landing page and finds that all 17 came through the quote page, which the web team had rebuilt the week before; the new form no longer stored the identifier in its hidden field. The CRM owner and the web team restore the field and test it with a real click. The 17 bookings cannot be recovered for that week, so the log records the gap, the cause and the fix, and the note to paid media says that the week's survey bookings in the account are understated and that bidding should be read with care until the next two uploads look normal. The monthly comparison later shows the gap closing once the form was repaired.

QA checks

Before the upload is closed, marketing operations confirms each check below. If one fails, fix the cause and re-run the step rather than patching the file.

  • Every row maps to a conversion action that exists and is meant to receive it.
  • No record is uploaded twice across uploads; the log and the CRM agree on what was already sent.
  • Rows without the required consent were removed before upload, and the number removed is logged.
  • Rejected rows have a cause fixed in the CRM, not a correction made only in the file.
  • Uploaded counts are compared with the CRM's own count for the same period, and any change in the gap is explained.

Hand-offs and records

The procedure hands on three things. The paid media manager receives the note after every upload, with anything that should change how bidding results are read that week. The CRM owner receives the grouped rejections and fixes their causes before the next upload. The monthly comparison and the period's upload log go to whoever prepares the monthly reporting pack, so the pack's lead and pipeline figures can be read against what the platforms were actually told.

The lasting record is an upload log entry: the date, the period covered, outcomes and rows by conversion action, rows rejected and why, the comparison with the CRM count and anything paid media was told. Keep the uploaded file and the import results with it. The log is also what makes duplicates preventable, because it is the only place that says which records have already left the business. The academy checkpoint on conversion windows, reconciliation and offline uploads tests which outcomes to upload, the consent checks before an upload and how to read the results against the CRM afterwards.

Common mistakes

Each of these teaches bidding something untrue, quietly.

  • Deciding the outcomes on the day. The list of outcomes and actions belongs to the plan; changing it in the middle of an upload changes what bidding optimises for without anyone agreeing to it.
  • Exporting by the date of the export. The upload must carry when the outcome happened, or it lands in the wrong period or outside the window.
  • Uploading records without consent. A missing or declined consent state stops the record, whatever else the CRM holds.
  • Fixing rejected rows in the spreadsheet. The same fault rejects the same rows next time; fix the form or the field.
  • Uploading the same outcome twice. Overlapping export periods and re-sent corrections double-count; the log prevents it.
  • Ignoring the gap until month end. A widening gap between uploaded and recorded outcomes is a broken capture that gets more expensive every upload.
  • Not telling paid media. A short or late upload changes what bidding sees; the people reading the results need to know.

Frequently asked questions

Why upload outcomes by file when the CRM could send them automatically?

Where an automated connection exists and works, it is usually the better route, and this procedure stops applying to the outcomes it sends. Many teams still upload by file because their CRM has no supported connection, because the outcome is recorded somewhere the connection cannot reach, or because they want a person to see the rows before they leave the business. The routine makes that manual route as dependable as an automated one.

Why does the platform never show exactly the number of outcomes we uploaded?

Because an upload only counts where it can be matched to an ad interaction under the platform's rules. Some rows are rejected, some fall outside the conversion window and some leads never came from an ad at all. A gap is normal; what matters is that it is understood, broadly steady from one upload to the next, and explained when it moves.

Should a corrected row be uploaded again after a rejection?

Only once the cause has been fixed in the CRM and the record has not already been accepted. Correcting the file by hand hides the fault that will reject the same kind of row next time, and uploading a record twice risks counting one outcome as two. Fix the source, then include the corrected record in the next upload and mark it in the log.

Terms used on this page

Tools to try

Frameworks

Where this page is used

Part of these learning paths

Also related

Sources

  1. Google Ads Help — Google Ads Help
  2. About enhanced conversions — Google Ads Help
  3. Google Analytics Help — Google Analytics Help

Related