DSA: Boosting ROAS 1.2x for B2B in 2026

Dynamic Search Ads (DSA) can be a goldmine for expanding your reach, since they automatically build headlines and pick landing pages right from your website’s content. But getting the most coverage out of DSA isn’t as simple as flipping a switch. It takes a smart, hands-on approach to targeting, negatives, and constant tweaking. This teardown looks at a recent campaign we ran, showing you how we dissected its performance and used a few strategic changes to seriously increase its reach and how efficiently it brought in conversions.

Campaign Overview: The “Digital Horizon” Project

We had a client, a mid-sized B2B company that sells cloud-based project management software, who wanted to launch what we called the “Digital Horizon” DSA campaign back in Q3 2025. Their goal was simple: cast a wider net to find new leads, especially for those long-tail searches that their standard exact and phrase match campaigns were probably missing. The big challenge, like with any DSA campaign, was getting that extra coverage without tanking the quality of the leads or letting costs get out of control.

The campaign was slated to run for three months, from July 1, 2025, to September 30, 2025, and we had a total budget of $45,000 to work with. Our main targets were a Cost Per Lead (CPL) below $60 and a Return On Ad Spend (ROAS) of at least 1.5x. Based on our initial math, we thought we could get around 2.5 million impressions and pull in 1,500 conversions.

Initial Setup and Strategy

We structured the campaign with just one ad group and pointed it at the client’s entire website as the dynamic ad target. For bidding, we went with “Maximize Conversions” but set a target CPL of $70 to give the algorithm a guardrail while it was learning. To avoid burning money right out of the gate, we pre-loaded a negative keyword list with a bunch of generic terms and competitor brand names. We just used a standard text ad format, letting Google Ads do its thing by generating headlines and display URLs from whatever landing page it decided to use.

Starting with this “catch-all” method was a conscious choice, as it’s the fastest way to see what new search query patterns are out there. It’s a broad approach, for sure, but it gave us a baseline of data that we could use later for much tighter, more effective targeting. The idea was that the website had enough content that the algorithm would have plenty to work with to find good searches.

Performance Analysis: Q3 2025

Right away in the first month, impressions shot up and actually beat our forecast. But the conversion rate was lower than we wanted which meant our CPL was way over target. This is the classic DSA problem in a nutshell: you get tons of coverage, but often you pay for it with a lack of precision.

The initial CPL of $39.18 was actually better than our $60 goal, but that was mostly because we were getting a huge number of cheap clicks. The real story was the first month’s ROAS of 1.1x. It told us that even though people were converting, the value of those conversions wasn’t high enough to justify the ad spend. This was a clear signal that we needed to get more control over which website pages Google was indexing and how it matched them to searches.

Optimization Steps and Adjustments

Our strategy changed quite a bit over those three months as we constantly dug into the search term reports and conversion data. The objective stayed the same: get the most valuable coverage possible while cutting out the irrelevant spend.

Week 1-4: Initial Negative Keyword Expansion

The first four weeks were all about aggressive negative keyword expansion. We were living in the search term report, looking for all the generic, informational, and competitor-related queries that were getting clicks but zero conversions. For example, search terms like “what is project management,” “free project management tools,” and “asana alternatives” got added to the negative list immediately. This kind of proactive cleanup is something you absolutely have to do with any DSA campaign. If you don’t, you’re just paying Google to educate people on topics that have nothing to do with your product. That initial work cut our irrelevant impressions by 18% and nudged the CTR up from 3.20% to 3.30% by the end of the first month.

Week 5-8: Segmenting Dynamic Ad Targets

It was obvious that targeting the entire website was just too broad, so we decided to restructure our dynamic ad targets. Instead of using the “All Webpages” setting, we created specific targets using URL categories. For instance, we set up one target that only included URL contains "/solutions/", a second for URL contains "/features/", and a third for URL contains "/pricing/". This gave us the ability to adjust our ad copy and bids for each section. We bid up on the “/pricing/” category because we knew those searchers were much closer to making a decision. For the “/solutions/” category, we kept the bids a bit lower but used broader ad descriptions. This segmentation was a big reason our conversion rate jumped from 1.06% to 1.42% in month two, simply because the ads were a much better match for what people were looking for. An eMarketer report on ad spending trends confirms that highly segmented campaigns almost always beat broad ones, with conversion metrics often improving by double-digit percentages.

Getting this granular with our targeting is also a core part of the strategy for getting better audience insight, which you need to drive real performance improvements.

Week 9-12: Audience Layering and Bid Adjustments

In the last month of the quarter, we started doing audience layering. We applied a +15% bid adjustment for anyone on our remarketing lists who had been to a product page but hadn’t converted. We also went after “in-market” audiences for “Project Management Software” and “Business Software” with a +10% bid adjustment. This move combined the wide reach of DSA with the targeted focus of audience data, giving us a huge lift in ROAS that jumped from 1.4x in month two all the way to 2.1x in month three. Our conversion rate also saw its biggest jump, hitting 1.85%. It just goes to show you that even with all the ad automation, the human element of knowing your audience is what makes the difference.

We also set up a rule that would automatically pause any dynamic ad targets that got a lot of impressions but zero conversions over a 30-day window. This auto-exclusion helped keep the campaign clean and made sure our budget was going to the parts of the site that were actually working.

What Worked and What Didn’t

What Worked:

• Constant Negative Keyword Pruning: Continuously adding negative keywords was the single best thing we did. It stopped us from wasting money and made our leads much higher quality. The campaign probably would have failed without it.
• URL Category Targeting: Moving away from targeting the whole website and focusing on specific URL categories made a huge difference in our conversion rate and efficiency. It let the DSA engine focus on high-value pages instead of getting lost in blog posts or the “about us” section.
• Audience Layering: Using DSA with remarketing and in-market audiences was a fantastic way to improve our ROAS. It put a layer of user intent on top of the broad keyword discovery process.
• Automated Exclusion Rules: Setting up rules to auto-pause underperforming ad targets was a big time-saver and kept the campaign running clean without constant manual checks.

What Didn’t Work (or required adjustment):

• The Initial Broad Targeting: While it helped us find new queries fast, targeting the whole website from the start led to a ton of junk impressions and clicks in the first month. This is rarely a good strategy for long-term efficiency.
• Relying on “Maximize Conversions” without a CPL Cap: At first, we let “Maximize Conversions” run wild without a target CPL, and while we got conversions, some of them were pretty expensive. Adding that target CPL was what got our costs back in line. It’s a common mistake to think smart bidding can solve everything on its own. It still needs your parameters.
• Ignoring the Search Term Report: If we had forgotten to check the search term report regularly, this campaign would have been a disaster. That report is your only real window into how people are finding your dynamic ads, and checking it is not optional for DSA success.

Lessons Learned and Future Implications

The “Digital Horizon” campaign really proved that while dynamic search ads are great for getting more coverage and finding new search queries, you only unlock their real power with constant, hands-on optimization. It’s not a “set it and forget it” tool. The initial broad strategy worked for discovery, but it was the later work, the segmentation, negative keyword management, and audience layering, that really drove the campaign’s efficiency and pushed up the ROAS.

Moving forward, we’ll push to start DSA campaigns with more specific dynamic ad targets right from the beginning, maybe focusing just on key product or service categories instead of the whole website. This should help cut down on the initial wasted spend on irrelevant searches. We’ll also make it a priority to integrate first-party data for audience targeting from day one, using what we know about existing customers to guide our bid adjustments. The focus has to be on continuous monitoring and quick changes, treating DSA like a living system, not just a static part of the account.

There’s no question that DSA can fill the gaps that your traditional keyword campaigns leave behind. But if there isn’t a strategic person guiding it, it can turn into a very expensive way to get a lot of visibility with very little to show for it. The trick is to find the right balance between the machine’s automation and a smart human’s intervention.

Just remember that DSA is an amazing tool for discovery, but it takes constant work to turn that discovery into profit. You have to cast a wide net, but you also have to make sure that net is fine enough to only catch the fish you actually want.