Google Ads Says Target Edits Skip Learning, Which Makes Big Edits Riskier
Google Ads product manager Tal Akabas said on the October 15, 2026 episode of Google's Ads Decoded podcast that changing a target or budget on an established campaign does not trigger a learning period. Smart Bidding instead scales bids "directly in proportion" to the change. That removes the old excuse for timid 15% steps, but it also means a big edit hits your CPA immediately.
That second part is the bit I keep coming back to. A learning period was annoying, but it was also a buffer. If the system just does what you told it, faster, then a bad target is a bad target on day one.
The claim comes from PPC Land's writeup of the 40-plus-minute episode, where Akabas talked with Ads Product Liaison Ginny Marvin about bidding and budget controls. His exact words were that "there isn't actually a learning period" when targets or budgets change on a steady-state campaign. New campaigns are different; there, he said, "some learning period can take place."
What Akabas said, and what he carefully didn't
The quote is narrower than the headline version that will circulate on LinkedIn this week. Three limits are worth holding onto.
First, it applies to established campaigns. Anything new, recently reactivated, or restructured is still on the old rules.
Second, it covers target and budget edits only. Adding or removing campaigns from a portfolio, merging ad groups, swapping bid strategies: none of that is covered, and Google's own help page on learning periods still lists "a setting for the bid strategy was changed" and "campaigns, ad groups, or keywords have been added to or removed from the bid strategy" as reasons for a Learning status. Whether a target value counts as a "setting" there is honestly a bit ambiguous. The help page also says ad group target changes "may trigger" Learning on Shopping campaigns, which sits a little awkwardly next to the podcast claim.
Third, and this is the one that matters, PPC Land points out that Akabas did not say how large a target change can be before proportional scaling stops behaving. Halving a tCPA is technically "a target change." I'd be surprised if the system treats a 50% cut the same way it treats a 5% nudge, and nobody at Google has said it does.
So: no learning label. That doesn't mean no volatility.
Where the 15-20% rule actually came from
Most of us learned the step rule from practitioner content, not from Google. Navah Hopkins, writing in Search Engine Journal's Ask A PPC column in December 2025, listed "changing the budget more than 15% in a 7-day period" as a learning trigger, along with "changing TCPA/TROAS goals (especially if they are large changes)." She described learning as anything from 48 hours to two to four weeks. That's a reasonable summary of what the industry believed, and to be fair to Hopkins, she was writing a platform-agnostic piece as a Microsoft employee, so it was never meant to be Google gospel.
Google's line has been drifting for a while. In a March 2026 Ads Decoded episode, product manager Kristina Park said the system can handle direct target adjustments when the business needs them, and that changing targets every few minutes is "probably suboptimal," according to PPC Land's coverage of that episode. Her advice was to wait at least one conversion cycle before adjusting again. That's a very different instruction from "never move more than 15% a week."
The October episode basically finishes that thought. Google is telling advertisers the step rule is a habit, not a requirement.
The August change makes proportional scaling bite harder
Here's where I think most teams will read this wrong. On August 17, 2026, Google started rolling out a change where budget-limited tCPA and tROAS campaigns now optimize toward the target instead of beating it, across Search, Shopping, Performance Max, Demand Gen and Travel. Google's example was a $10 tCPA campaign delivering at $5. Under the new logic, delivery moves toward $10.
Put that next to "bids scale in proportion to your target changes" and the target stops being a ceiling you hope to stay under. It becomes a dial the system will turn to almost exactly where you set it. Measured's analysis, quoted by PPC Land, described the August change as converting "a setting many advertisers treated as a ceiling into a direct instruction."
The target used to be a wish. Since August, on budget-capped campaigns, it's closer to an order.
An analogy that seems to fit: the old system was a car with a laggy throttle, so you learned to press gently and wait. Google just told you the lag is gone. Pressing gently is still fine. Stomping on it is now a lot more literal than it used to be.
A target edit that won't wreck your week
None of this means you should start swinging targets around. It means you can stop splitting one decision into five tiny ones over five weeks, which (honestly) was never a great use of anyone's time in Q4.
The workflow I'd use on an established campaign:
- Decide the real target first. Work backward from margin or payback, not from last month's CPA. If the number you actually want is 25% away from today's target, the old rule says take it in two steps over two weeks. On a steady-state campaign, Google's position now supports taking it in one.
- Size the move against conversion volume, not a percentage rule. Google's help page says calibration takes "a few conversion cycles (1-2 typically)" and can be faster with more data. My rough line: if a campaign gets fewer than about 30 conversions a month, I'd still cap single moves around 20%, because there's not much signal for the proportional math to work from. Above that, I'd trust the bigger move.
- Judge it over 1-2 conversion cycles, then stop touching it. If your typical lag from click to conversion is 10 days, that's a 10-20 day read. Not three days. Park's "wait at least one conversion cycle" advice from March still applies even if the Learning label never shows up.
- Annotate the change. Pre- and post-August data already reflect different bidding rules. If you also change targets without notes, you'll have no idea in January which change did what.
For budget changes specifically, Akabas's claim mostly matters for uncapped or lightly capped campaigns that you want to scale for Black Friday. I'd still raise budget the day before a known demand spike rather than the morning of, mostly because pacing needs room to work, not because of learning. (If you're also tempted to layer seasonality adjustments on top, the data on Black Friday seasonality adjustments is worth reading first. Akabas himself called them "a relatively more advanced tool" in the same episode.)
The audit nobody runs: under-adjusted targets
The quieter cost of the step rule is the target changes that never happened. It's a pretty common pattern: the tCPA was set at launch, the business economics shifted, and nobody moved it because the last time someone did, the campaign went "Learning" and a client asked questions.
A 10-minute check:
- Pull every tCPA and tROAS campaign with its current target, 30-day actual CPA or ROAS, and the date the target was last edited (change history will tell you).
- Flag anything where the target hasn't moved in 90+ days and actual performance sits more than 20% away from it in either direction.
- For each flagged campaign, write down the target your margins actually support. If it's different, that's your list.
If you're budget-capped and beating target, remember the August change: Google is already walking that campaign toward the number you set. Either lower the target to something you actually want, or accept that the "efficiency" you were seeing is going away. And if channel mix is part of the problem, the new channel prioritization controls Akabas also discussed are probably a better lever than another target tweak.
Where I think this goes by spring
My guess, and it is a guess: by the end of Q2 2027, Google's help documentation will explicitly say target value changes on established campaigns don't cause a Learning status, and at least half of agency onboarding decks will still teach the 20% step rule anyway. Folk rules outlive the systems they were built for. They usually do.
One slightly uncomfortable thought, and I'm not sure what to do with it. A lot of account management cadence (weekly calls, "we're in learning, let's give it time") was built around the learning period as a reason to wait. If the wait mostly isn't real anymore for established campaigns, some of that cadence was really just about managing nerves. Not the worst thing to be good at, to be fair.
Anyway, I don't think the takeaway is "change targets freely now." It's closer to: make the decision you were going to make in five steps, make it once, and then actually give it the full conversion cycle before you second-guess it.
