Google Ads Offline Conversion Uploads in Malta: The 7-Day Attribution Gate in 2026

If you run Google Ads offline conversion uploads in Malta, there is a seven-day deadline sitting inside your account that nobody sent you an email about. Conversions uploaded more than seven days after the event still show up in your Campaigns page columns — but they never reach the data-driven attribution model, and they never teach Smart Bidding anything. Two totals, one account, same set of leads.

The rule was quietly documented in the Google Ads Help Center and surfaced publicly on 22 August 2026 by consultant Adriaan Dekker, who credited Hana Kobzova with the find. There was no release note. For any Malta business running Google Ads campaigns where the sale closes days or weeks after the click — solar installers, law firms, financial services, recruitment, B2B services, high-ticket home improvement — this is the single most consequential measurement detail of the year.

What Google Actually Changed (And What It Didn't)

Nothing about your campaign settings changed. What changed is that Google wrote down a constraint that had been operating invisibly.

The daily attribution modelling engine — Google calls it Search funnels — processes conversions that occurred within the last seven days. An offline conversion uploaded more than seven days after the initial event is, in Google's own wording, "bypassed by the attribution modelling calculations."

Standard reporting operates without that restriction. Upload a conversion three weeks late and the Campaigns page will retroactively record it against the right campaign, the right ad group, the right keyword. The Model Comparison report will act as though it never happened.

Neither number is wrong. They are counting under different rules, and most advertisers have spent years assuming they were counting the same thing.

A Worked Example

A prospect in Sliema clicks your Search ad on 1 September. Your sales team qualifies them on the 8th. The deal closes on the 12th. Your CRM sync pushes the conversion to Google on the 15th.

• Campaigns page: one conversion, attributed to 1 September.
• Model Comparison / data-driven attribution: nothing. The event is fourteen days old at upload.
• Smart Bidding: learns nothing from your best lead of the month.

Multiply that across a quarter and you have a bidding algorithm optimising towards the leads that close fast rather than the leads that close big. In practice that usually means Google steers spend towards low-intent, high-volume enquiries and away from the considered-purchase traffic that actually pays your salaries.

The Wording Has an Inconsistency Worth Noting

Google's documentation describes the gate two different ways in the same passage. One statement keys the rule to the age of the conversion (the engine only processes conversions that occurred in the last seven days). The other keys it to upload latency (a conversion uploaded more than seven days after the event is bypassed).

Those tests diverge. A conversion that happened fourteen days ago but was uploaded two days after it happened passes the latency test and fails the event-window test. Google does not say which applies. Until it does, the safe operating assumption for Malta advertisers is the stricter reading: if the event is older than seven days, treat it as invisible to the model. That also means historical backfills — the "let's import the last six months from the CRM" project — carry no modelling weight at all.

Three More Reasons Your Reports Never Reconcile

The upload gate is not the only divergence between attribution reports and standard columns. Google's documentation now names four structural differences, and it is worth knowing all of them before your next client review or board meeting.

1. Timing convention

Attribution reports count by time of conversion. The Campaigns page counts by the time of the ad query that preceded the click. Adding "by conversion time" columns to the Campaigns page aligns the two.

2. Network coverage

The Model Comparison engine explicitly excludes conversions from the Search Partner Network, Gmail and App campaigns. Standard columns include them by default. If you have ever tried to assess Search Partner return on spend from an attribution report, you were working from a view that structurally cannot contain the data.

3. Campaign and format coverage

Attribution reports cover Search, Shopping, Video, Display (excluding pay-for-conversions) and Demand Gen. App and Hotel campaigns are absent. Video format coverage is narrower too — masthead, outstream and in-feed video appear in Campaigns but not in attribution.

Four independent sources of divergence, before anyone has even chosen an attribution model. This is why the reconciliation exercise you have been putting off has never worked.

What This Means for Malta Advertisers Specifically

Malta's paid media market has three characteristics that make the seven-day gate bite harder here than in larger markets.

Small volumes make model accuracy fragile. A typical Maltese advertiser is working with tens of conversions per month, not thousands. Data-driven attribution and Smart Bidding are already operating close to the edge of statistical usefulness at that scale. Removing every conversion with a sales cycle longer than a week doesn't degrade the model — it can starve it entirely. If your lead generation engine produces 30 qualified leads a month and 20 of them close after day seven, your bidding algorithm is learning from ten data points.

Manual sales processes create upload latency by default. Most Maltese SMEs do not have a Salesforce-to-Google Data Manager API pipeline firing hourly. They have a sales manager who exports a spreadsheet at month end, or a HubSpot workflow set to a weekly cadence, or an accountant who confirms which deals actually invoiced. Month-end batching is the single most common offline conversion pattern in this market, and it guarantees that most uploads land outside the window.

Considered-purchase categories dominate. The sectors driving Malta's search spend — property, iGaming B2B, financial and corporate services, renewables, healthcare, education, professional services — all have sales cycles measured in weeks. These are precisely the businesses that adopted offline conversion imports to escape lead-volume optimisation, and precisely the businesses whose data is now being filtered out of the model they were trying to improve.

There is also a compliance layer. EU advertisers already run offline uploads through Consent Mode v2 and GDPR-compliant hashing, which adds its own processing friction. If your consent-gated data is also arriving late, you are compounding two forms of signal loss on the same conversion.

The Fix: A Seven-Day Upload Discipline

You cannot change Google's window. You can change how fast your data gets there. Here is the practical sequence we work through with clients on paid media engagements.

Step 1 — Measure your actual upload latency

Before anything else, find out what your current median lag is. Pull your last 90 days of offline uploads and compare, per record, the conversion timestamp against the upload timestamp. Most advertisers who have never checked discover a median somewhere between 12 and 35 days. That number is your problem statement.

Step 2 — Move to daily, automated uploads

Manual CSV imports and monthly batches have to go. The Data Manager API is now Google's unified ingestion point, and since 15 June 2026 new adopters can no longer use the older Google Ads API path for offline conversion imports. If you are still on a spreadsheet workflow, this migration is overdue regardless of the attribution question. Daily is the target; hourly is better and rarely harder once the pipeline exists.

Step 3 — Upload at the earliest defensible milestone

This is the part that requires a commercial decision rather than a technical one. If your true revenue event is "invoice paid" and that happens on day 45, you will never make the window. So define an intermediate conversion that happens inside seven days and correlates with revenue: qualified discovery call held, proposal issued, site survey booked, application approved. Upload that as your primary Smart Bidding signal with a value that reflects its historical close rate, and keep the final sale as a secondary, reporting-only conversion.

This is not a workaround. It is better practice than optimising to a 45-day event even without the seven-day gate, because it gives the algorithm a faster feedback loop.

Step 4 — Use enhanced conversions for leads properly

Enhanced conversions for web and leads collapsed into a single toggle over 2026. Hashed first-party identifiers captured at form submission give Google a match key that does not depend on the GCLID surviving in your CRM. Get this configured correctly and a meaningful share of your leads become matchable immediately rather than at upload time.

Step 5 — Stop reconciling the irreconcilable

Rebuild your reporting so that attribution reports and standard columns are used for different jobs. Standard columns for commercial reporting — what did we spend, what did we get. Attribution reports for path and assist analysis only. Never present the two side by side to a stakeholder without explaining the four structural reasons they differ.

Step 6 — Re-baseline your bidding

If you have been running Target CPA or Target ROAS on partially-modelled data, your targets are calibrated to a filtered reality. After you fix upload latency, expect a period of instability as the model ingests conversions it previously never saw. Widen your tolerance for two to three weeks, then re-set targets against the new baseline. A structured growth strategy should treat this as a planned transition, not an emergency.

What About Meta and Other Platforms?

Meta's Conversions API has its own latency expectations — Meta recommends server events within minutes and treats events older than seven days as unusable for most optimisation purposes, so the same discipline applies. LinkedIn's conversions API allows a 90-day lookback for uploads, which makes it comparatively forgiving for long B2B cycles. TikTok runs a similar server-side events API, though it is worth noting that TikTok Ads are still not available to advertisers in Malta; a rollout is expected during 2026 but has not landed, so Maltese businesses cannot yet buy TikTok inventory directly.

The broader point stands across every platform: in 2026, measurement systems reward advertisers who close the loop quickly, and quietly penalise those who don't.

Frequently Asked Questions

Does the seven-day rule mean my late conversions are deleted?

No. Late uploads are recorded in full in standard Google Ads reporting columns and attributed retroactively to the correct campaign, ad group and keyword. They are excluded only from the data-driven attribution modelling engine and, by extension, from the signal that trains Smart Bidding. Your revenue reporting stays accurate; your optimisation does not.

How do I check my current offline conversion upload latency?

Export your offline conversion records for the last 90 days from your CRM or upload log and calculate the difference between the conversion date and the upload date for each row. Take the median rather than the average, since a few very old backfilled records will skew a mean badly. Anything above seven days means a portion of your data is not reaching the model. In Google Ads itself, Tools and Settings then Conversions then Uploads shows recent upload history and diagnostics.

Should Malta advertisers stop using offline conversion imports altogether?

No — offline imports remain the only way to tell Google which leads were actually worth having, and abandoning them sends you straight back to optimising for raw form fills. The correct response is to shorten the loop: upload daily via the Data Manager API and optimise towards an earlier qualified-lead milestone rather than the final invoice. Turning imports off would cost you far more than the seven-day gate does.

Does this affect Performance Max campaigns in Malta?

Yes. Performance Max relies on Smart Bidding, which is fed by data-driven attribution output, so the same upload latency constraint applies. Performance Max is additionally labelled as Cross-Network inside attribution reports, which makes channel-level reconciliation harder again. If you are running Performance Max with offline conversion values, upload discipline matters more here than in any other campaign type.

Is a seven-day upload window realistic for a B2B sales cycle in Malta?

Not for the final sale — a two-to-six week cycle is normal for Maltese B2B and professional services. That is why the practical answer is to move the optimisation target earlier rather than to try to close deals faster. Pick a milestone that reliably occurs within seven days of the click, assign it a value based on its historical conversion-to-revenue rate, and let that be what Smart Bidding chases. Keep closed-won revenue as a reporting conversion so your commercial numbers stay honest.

The Takeaway

Google did not change a setting in August 2026. It documented a constraint that has been shaping bidding decisions in Maltese accounts for a long time without anyone being told. The advertisers who benefit from this disclosure are the ones who treat it as an operations problem — pipeline speed, milestone definition, upload cadence — rather than a reporting curiosity.

Audit your latency this week. If the median is above seven days, you are not measuring your Google Ads account. You are measuring the fast half of it.