Google Ads Updates: What to Check in September 2026

A campaign can report more conversions without bringing in more orders. That distinction is the first thing I would keep in mind when reviewing Google's latest measurement updates.

In a B2B store, a purchase, an add-to-basket event and a request for advice mean different things. My paid media work sits alongside the website and catalogue, so I want reporting that helps explain those differences. A bigger number is useful only when I can say what it represents.

The Google Ads updates worth reviewing in September 2026

On 10 September, Google announced changes across its measurement tools: wider Data Manager integrations, additional diagnostics, a Data Strength Uplift metric and updates to Meridian, including general availability of GeoX. The new uplift metric describes additional conversions recovered through a first-party data setup. It should not automatically be read as additional sales caused by advertising. Google's measurement announcement explains the scope.

There was also an August Demand Gen update. Google described tests allowing viewers to begin messaging conversations from YouTube ads and general availability of multimodal video creation in Asset Studio. A test announcement is different from a feature being available in every account. Read the August Demand Gen release.

For a small or medium-sized e-commerce team, I would translate those announcements into three questions: is our order data reliable, can we distinguish recovered measurement from incremental demand, and do we have the landing pages and creative to use additional reach well?

Check what the campaigns are being asked to achieve

Before changing bids, open the conversion settings. For a campaign intended to sell products online, the purchase event will usually be the main commercial outcome. An add-to-basket event can be useful for diagnosis, but it is an earlier step.

Google distinguishes primary and secondary conversion actions. Primary actions are used for bidding when their standard goal is selected. Secondary actions are generally observational, with an important exception: actions in a custom goal can be used for bidding regardless of that setting. Check the campaign's actual goals as well as the labels. Google's conversion-action documentation sets out the distinction.

Then check for duplication. If two integrations report the same purchase as separate conversion actions and both are used, the account can tell a misleading story. Review transaction identifiers, currency, values and the relationship between the store and advertising reports. Do not change several measurement components at once and expect a clean before-and-after comparison.

Reconcile the advertising story with the order book

I would start with a small sample of completed orders and work backwards. Does the store show the expected total? Was the purchase recorded once? Is the reported value based on the same treatment of VAT, shipping and refunds as the comparison report?

The numbers do not have to match perfectly across systems. Attribution windows, consent choices and reporting dates can produce differences. The point is to understand the differences well enough to use the data. A reconciliation note is more valuable than quietly switching between whichever dashboard looks better.

Use new creative tools against a real brief

Faster production can help a team test more useful messages. It can also fill an account with material that looks different but says the same thing. For an industrial catalogue, I would brief a video around one recognisable need: the installation problem, the product in use, the key compatibility detail and the next step.

Horizontal and vertical versions should be checked in their own layouts. A specification that is readable on a desktop can disappear on a phone. Cropping a product demonstration may remove the very connection the buyer needs to see.

If a messaging placement becomes available, decide who will answer and what counts as a useful enquiry before switching it on. A low cost per conversation is not much help if messages go unanswered or mostly concern products the business does not supply.

A manageable test plan

1. Write the question. For example: can a new creative route generate more qualified enquiries for one product category?
2. Choose the outcome. Use completed orders or qualified leads, with earlier actions recorded separately.
3. Set boundaries. Define spend, eligible products, stock requirements and a review point that allows for the usual conversion delay.
4. Keep a change log. Record feed, pricing, website and measurement changes alongside campaign changes.
5. Review the commercial result. Consider order value, contribution, support workload and returns, not just platform efficiency.

Geo-experiments and marketing mix modelling can answer valuable questions, but I would not start there if basic purchase tracking is still uncertain. The sophistication of the tool should match the decision and the data available.

Do not leave the website out of the review

A relevant click still needs a usable page. Check the selected variant, stock, pack quantity, delivery information and basket behaviour. For longer trade orders, bulk ordering and clearer delivery information may deserve attention before another round of audience changes.

The practical benefit of the latest updates is the opportunity to improve the evidence behind decisions. I would use that opportunity to make the account easier to explain: what we spent, what customers did and what the business gained.

Common questions

Should I make add to basket a primary conversion?

Choose goals around the campaign's purpose. For direct online sales, I would normally keep completed purchases central and use basket activity to understand the journey.

Does better measurement mean better sales?

It can support better decisions, but recording an existing order more accurately is different from causing a new one.

Back to the journal