The ROAS drop that happens when you scale Google Ads is predictable and, in most cases, preventable. Here is why it happens, the budget rules we use with every account, and the expansion sequence that keeps profitable campaigns profitable as spend grows.
By Jack Goldsmith, Founder & Performance Marketer, Social Surge · 1 September 2026
How do you scale Google Ads without losing ROAS?
Increase budgets in increments of no more than 15-20% every 14 days, confirm the account has at least 30-50 conversions per month before scaling, and expand horizontally (new campaigns, new product groups) rather than simply pouring more money into existing campaigns. Most ROAS drops on scale are caused by Smart Bidding leaving its learned state or by query expansion into lower-intent search terms, both of which structured, gradual scaling avoids.
Understanding why the drop happens makes it easier to prevent. There are three distinct mechanisms, and they typically stack on top of each other when budget increases are too aggressive.
Bid algorithm disruption. Google's Smart Bidding operates on a rolling conversion window, typically around 30 days, to calibrate its bid decisions. When you increase a daily budget significantly overnight, the algorithm exits its learned state and begins competing in auctions it previously skipped. Those auctions tend to have lower conversion rates. Google's Smart Bidding documentation notes that large changes to campaigns in learning mode should be evaluated carefully because the algorithm needs time to recalibrate. A gradual increment prevents this disruption.
Query expansion. As you push for higher impression share, your ads surface for a broader set of search terms. The terms you were not showing for before are, on average, lower intent than the ones you were already winning. This is not a targeting failure; it is a natural consequence of auction mechanics. The fix is proactive: review search term reports after every budget increment and add negatives before the next one.
Audience saturation and diminishing returns. Your highest-intent buyers, the people who convert fastest and most cheaply, are a finite pool. Once you have reached most of them, additional spend reaches progressively less-engaged audiences. In the accounts we manage, this effect typically becomes visible when impression share for exact and phrase match terms exceeds 80-85% of available auctions.
Most store owners default to vertical scaling: take a campaign that is working and increase its budget. It is the obvious move, but it is also where diminishing returns hit fastest. You are asking the same campaign to find more buyers in a pool that is getting shallower.
Horizontal scaling means growing by adding new campaigns, new product groups, new geographies (within the UK) or new keyword themes, rather than inflating existing campaigns. Each new campaign starts with a focused conversion signal and its own learning period, keeping quality higher than spreading a single campaign thin across a wider audience.
In practice, the most resilient scaling plans combine both approaches: a modest vertical increment on proven campaigns (staying within the 15-20% rule) running alongside a parallel horizontal expansion into the next product category or keyword segment. Neither approach in isolation is as robust as both together.
This is the most practical piece of scaling discipline we apply. No matter how confident you are in the account, do not increase a campaign's daily budget by more than 15-20% in a single adjustment. Then wait a minimum of 14 days before the next change.
The 14-day window corresponds to Google's Smart Bidding learning period. If ROAS stays stable or improves over that fortnight, the campaign has absorbed the increase and is ready for another increment. If ROAS deteriorates, you have caught the problem while it is still small, rather than after a large overspend.
One exception worth noting: if a campaign is budget-capped every single day (check "Impression Share Lost: Budget" in your campaigns view), the existing budget is acting as a ceiling on the algorithm. In that situation a larger increment is sometimes warranted, because the campaign already has the conversion data to support higher spend. Even then, we typically cap individual increases at 30-35% to avoid triggering a full learning reset.
A common mistake is keeping the tROAS target too high when increasing budget. The algorithm restricts spend to protect the target, so you end up with a bigger daily budget that does not actually get spent. You pay nothing extra, but you also grow nothing.
When scaling, consider lowering the tROAS target by 10-15% temporarily to give the algorithm room to find additional volume. Counterintuitively, actual ROAS often stays above the old target because the algorithm now has space to bid on high-intent signals it was previously ignoring to hit a tighter number. Once the new spend level is stable over two to three weeks, you can nudge the target back up gradually.
Portfolio bid strategies are worth considering at this stage. If you have several structurally similar campaigns (for example, Standard Shopping split by product category), a portfolio budget lets the algorithm redistribute spend intelligently between them without you managing individual budgets manually. This tends to smooth out the volatility that comes with scaling several campaigns at once.
Scaling an account that is not ready amplifies problems rather than results. These are the signals we check before recommending any budget increase.
When an account passes the readiness checks, here is the sequence we follow.
Step 1: Audit for wasted spend before adding budget. Run a search term report, review placement exclusions and check for keyword duplication across campaigns. Money saved by cutting waste improves signal quality for Smart Bidding and typically delivers a better return than the same amount added as new budget.
Step 2: Separate branded search into its own campaign. Branded terms convert at high ROAS and relatively low CPC. Running them alongside non-brand in a single campaign dilutes performance reporting and can cause the algorithm to over-bid non-brand terms to protect the blended average. A dedicated branded campaign lets each segment optimise cleanly, and lifting the branded ROAS pulls the overall account average up as a by-product.
Step 3: Expand horizontally into the next product group. If your current Shopping or Search campaign covers your top-selling products, build a parallel campaign for the next tier. Give it its own budget and a slightly looser tROAS target to allow for the learning period. Let it accumulate conversion data independently before drawing conclusions about performance.
Step 4: Increment non-brand budgets by 15-20%. Wait 14 days. Review search term reports, add negatives where needed and check impression share lost metrics. If ROAS is stable, increment again. If ROAS has dropped, diagnose before moving further: the most common culprit is query expansion into broad match terms, which is fixable with a negative keyword pass.
This is not a fast process. Accounts that scale from £3,000 to £10,000 per month in 60 days almost always experience the ROAS drop that scares store owners off scaling entirely. Accounts that follow the incremental sequence typically hold ROAS within 5-10% of their pre-scale baseline, in our experience.
If you want an independent view of where your account sits before a scaling push, a free PPC audit covers wasted-spend analysis, bid strategy review and a clear growth sequence tailored to your account. Or see our Google Ads management service for how we structure scaling as part of ongoing management.
No more than 15-20% per adjustment, then wait 14 days before the next increment. Google's Smart Bidding algorithms operate on a rolling conversion window; large overnight increases push the algorithm out of its learned state, which typically causes a temporary ROAS drop.
Three main reasons: the bid algorithm exits its learned state and starts bidding into lower-quality auctions; query expansion brings in broader, lower-intent search terms as impression share grows; and the best-converting audiences are finite, so additional reach tends to mean lower-quality traffic on average.
We recommend a minimum of 30-50 conversions in the past 30 days before attempting significant scaling. Below this threshold, Smart Bidding has insufficient data to optimise reliably and scaling amplifies noise rather than signal.
Horizontal scaling means growing by adding new campaigns, product groups, match types or geographies rather than simply increasing the budget of existing campaigns. It protects ROAS because each new campaign starts with a focused conversion signal rather than diluting the existing one.
Book a free PPC audit, full account review, wasted-spend analysis and a clear growth plan, delivered in plain English.
Book Your Free Audit