On 15 June 2026 Google announced three changes to how campaigns spend budget and chase targets. The announcement packaged them together, but from an account manager’s seat they are three very different things: two are features you turn on, and the third is a system change that turned itself on.
That distinction matters, because the most consequential of the three is the one nobody had to enable.
The three changes, in order of impact
| Change | Status | Date | What it asks of you |
|---|---|---|---|
| Tighter delivery against bid targets | Live for everyone | 17 August 2026 | Nothing to enable — but review the impact |
| Smart Bidding Exploration | Global on PMax without a feed, beta on PMax with a feed and Shopping | From 15 June 2026 | Opt in and set a ROAS tolerance |
| Promotion mode | Beta on Search and Performance Max | From 15 June 2026 | Configure per event |
- 15 June 2026 The package is announced: Smart Bidding Exploration extended, promotion mode in beta, advance notice of the delivery change.
- 6 July 2026 The bid target adjustment tool becomes available, so accounts can recalibrate tCPA and tROAS before the change lands.
- 17 August 2026 The delivery change takes effect across all accounts.
The 17 August change: why your targets suddenly bite
The most technical item in the package affects budget-limited campaigns running Target CPA or Target ROAS. Historically the target in those campaigns behaved more like a direction than a constraint: when the budget ran out before the day did, the system could deliver at a cost per acquisition noticeably distant from the number in the settings. Since 17 August 2026 the behaviour is more literal — campaigns deliver closer to the target you stated.
That reads like an improvement, and narrowly it is: a target that means what it says is easier to govern. But what it does to your numbers depends entirely on where the campaign sat before.
- Campaign delivering above its stated tCPA. Cost per acquisition falls toward the target, and volume almost always falls with it. If the target was written optimistically and only "worked" because the system was overshooting it, that campaign is about to partially switch itself off.
- Campaign delivering below its stated tCPA. The system now has room to push: expect more volume and a CPA climbing toward the number you wrote. If that number was a safety margin rather than a true break-even, you are about to pay for the difference.
- Campaign not limited by budget. Minimal impact — the change is specifically about the case where budget is the bottleneck.
Your target is no longer an aspiration. The number you typed months ago as an "ideal goal" is now the number the system genuinely converges on. The first move is not to change bids: it is to reread every tCPA and tROAS and ask whether it is the value your margin can carry, or a leftover from a test six months ago.
Reading the impact without mistaking it for seasonality
Late August is the worst moment of the year to isolate an effect: the back-to-work rebound moves demand in almost every vertical. The honest comparison is not "before and after 17 August" but year over year on the same weeks, cross-referenced with impression share lost to budget. If budget-lost share holds steady and CPA falls, the change is doing what it says. If budget-lost share grows while CPA falls, the campaign is simply retreating from the expensive auctions — which may be exactly right, or a real loss, depending on what those auctions are worth to you.
Smart Bidding Exploration: buying demand you cannot currently see
Smart Bidding Exploration introduces a concept worth understanding properly, because it recurs everywhere in 2026-era Google Ads: tolerance. Instead of telling the system "never go below this ROAS", you tell it "on part of the spend you may accept a lower ROAS, if that is what it takes to explore queries we are not reaching today".
Google reports that campaigns using it see on average +18% unique converting query categories and +19% conversions. The first number is the more interesting one: it does not just say "more conversions", it says "conversions from kinds of searches you were not already covering". That is precisely the growth a mature account struggles to find with the classic levers.
With the June announcement the feature was rolled out globally to Performance Max campaigns without a product feed, and opened in beta to Performance Max with a feed and to Shopping campaigns.
The two conditions that make exploration safe
Widening the search surface only works if two things are in order — the same two that make any automation trustworthy:
- Conversion value has to be true. If you are sending Google cart value instead of margin, or treating every lead as equal, exploration will find more volume of the wrong kind. We wrote a separate piece on moving offline conversions to the Data Manager API: in lead generation this is the prerequisite, not a later optimisation.
- Brand boundaries have to be explicit. The more the system explores, the more you need brand exclusions and campaign-level negative lists. The same controls that govern AI Max in Search campaigns apply here.
Promotion mode: the right lever for peaks
Promotion mode, in beta on Search and Performance Max, solves a problem as old as Smart Bidding itself: what to do during Black Friday, a launch or a sale week. The common practice — raise the budget, loosen tROAS by hand, then remember to put both back — has two flaws. First, every manual target edit is a signal the system has to absorb, and it absorbs it exactly when you have the least time to wait. Second, sooner or later somebody forgets to change the values back.
With promotion mode both adjustments — ROAS tolerance and extra daily budget — are scheduled against a defined window and expire on their own, leaving the campaign’s baseline settings untouched.
Calendar item, now. If you sell online, the window that matters is November. Configuring promotion mode in late October means testing it for the first time during the peak. The moment to try it is a minor September event, where a mistake is cheap.
When it makes sense, and when it does not
| Account situation | Smart Bidding Exploration | Promotion mode |
|---|---|---|
| Mature account, brand demand saturated, growth flat | Yes — this is the scenario it was built for | Only for real events |
| New account, few conversions per month | No — it needs stable volume to learn from | No |
| Ecommerce with widely different margins per product | Only after you send margin as value | Yes, with tolerance set on promotional margin |
| Lead generation without offline conversions | No — you risk buying worse leads at volume | Largely irrelevant |
| Chronically budget-limited campaign | Fix the budget first, then explore | Yes, this is the use case |
The next-30-days checklist
- Reread every target. For each campaign on tCPA or tROAS, write next to the current value the value your margin actually supports. If the two differ, the 17 August change is already costing you volume or already capping it.
- Segment by budget limitation. Filter campaigns by budget status and impression share lost to budget: only the limited ones are touched by the change.
- Compare year over year, not month over month. Separate the effect from the back-to-work rebound before you declare an impact.
- Verify conversion value before enabling any tolerance. Margin instead of revenue for ecommerce; differentiated value by funnel stage for lead generation.
- Test promotion mode on a small event, before it matters.
- Refresh brand exclusions in every campaign where you turn exploration on.
The thread running through all three is the same: Google keeps moving control away from bids and toward signals. The number you write as a target counts for more than it used to, the value you declare per conversion counts for more than it used to, and the boundaries you set for the system count for more than they used to. It is the same logic we traced through Performance Max controls and reporting.
If you want to know where your account sits against these three changes, the free audit starts exactly here: targets, budget limitation and the quality of your conversion value. You can also see how we work and which services cover this ground.