Google Ads works for SaaS when the conversion signal you feed it reflects real pipeline value rather than raw form fills. Automated bidding optimizes toward whatever you tell it to value, so a poor conversion definition produces efficient acquisition of the wrong leads.
Most of the manual optimization work in Google Ads has been automated away. What remains is more consequential and less discussed: deciding what the machine should optimize toward.
Give it a conversion signal that counts every form fill equally and it will efficiently buy you the cheapest form fills available. It is doing exactly what you asked.
This guide covers the decisions automation did not remove: getting the conversion signal right before anything else, structuring the account so bidding can actually learn, choosing a bid strategy that matches how much you know, doing the negative keyword work nobody enjoys, deciding whether Performance Max belongs in a B2B account, and knowing when a second engine is worth the effort.
The conversion signal is the whole strategy
If a newsletter signup and a demo request both count as one conversion, smart bidding will chase newsletter signups, because they are cheaper and more plentiful. This is measurable: Dreamdata's 2026 benchmarks, attributing to closed-won revenue rather than form fills, put Google Search at 67% return on ad spend at the median and 138% for top performers, a gap driven largely by what the account was optimizing toward. Those values should come from the same sales reporting the rest of the business uses, and GA4 needs configuring to pass them back.
Fix this before touching anything else. Assign values that reflect real downstream worth, and import qualified opportunities back from the CRM so the algorithm optimizes toward pipeline rather than toward activity.
The feedback loop is the part teams skip because it crosses systems. Getting qualified opportunities and closed-won values back into Google means connecting the CRM to the ad account, usually through offline conversion import, and keeping the definitions consistent on both sides. It is unglamorous integration work, and it is the difference between an algorithm optimizing toward pipeline and one optimizing toward whatever is cheapest to collect.
Structure for signal, not for tidiness
Highly granular account structures made sense when humans set bids per keyword. Now they starve each campaign of the conversion volume automated bidding needs to learn.
There is a floor below which a campaign cannot learn: too few conversions a week and automated bidding never leaves the learning phase, so it behaves erratically and the noise gets blamed on the settings. Consolidating starved campaigns into a shared theme with enough volume often improves results more than any bid change, precisely because it finally gives the model something to learn from.
Consolidate into themes with enough conversions to be statistically meaningful. Fewer, better-fed campaigns beat many precisely-segmented ones that never exit the learning phase.
Which bidding strategy should you use?
Start with maximize conversions while you accumulate data, then move to target cost per acquisition once you know what an acquisition is genuinely worth. Target return on ad spend only makes sense when you can pass back real revenue values.
Change targets gradually. Large adjustments reset the learning period and cost you a week or two of performance each time, which is how accounts end up permanently in learning. Whichever strategy you choose, resist changing it weekly; each one needs a stable run of data to prove itself, and an account that switches strategies chasing last week's numbers never gives any of them long enough to work.
Should SaaS accounts use Performance Max?
Performance Max hands almost every lever to Google at once: budget flows across Search, Display, YouTube, and the rest based on where the algorithm predicts conversions. For ecommerce with a clean revenue signal that often works well. For B2B lead generation it is riskier, because the same weak conversion signal that misleads Search bidding misleads Performance Max across far more inventory, and you can see much less of where the money actually went.
If you run it, feed it the same pipeline-weighted signal you built for Search, add account and audience exclusions so it does not spend on your own brand or on obviously irrelevant placements, and watch its reporting closely. Treated as a supplement with a good signal and tight exclusions, it can find demand Search misses. Treated as a set-and-forget replacement for a considered account, it efficiently buys exactly the cheap conversions you told it to value.
The negative keyword work nobody does
Automated bidding does not fix irrelevant traffic; it just bids on it efficiently. The search terms report is where you see what your keywords actually matched, and mining it for negatives is the recurring maintenance that keeps broad match from quietly widening your spend into adjacent meanings you never intended to buy.
- Job seekers. Careers, salary, jobs, and internship variants waste budget in every SaaS account.
- Free and open source. Unless a free tier is your funnel, these searches rarely become customers.
- Tutorials and definitions. How-to and what-is queries are research, not purchase intent.
- Wrong-category matches. Broad match finds adjacent meanings of your terms with impressive creativity.
What about Microsoft Ads and other engines?
Microsoft Ads carries lower competition and lower click costs, with a smaller audience that skews toward enterprise and desktop. For B2B that mix is often favorable, and importing existing campaigns makes testing cheap.
Treat it as a supplement rather than a replacement. Volume will be a fraction of Google's, and the effort is best justified once Google is already working rather than as an escape from it.
- ✓ Automated bidding optimizes toward whatever you told it to value. Fix the conversion signal first.
- ✓ Over-segmented accounts starve campaigns of the conversion volume smart bidding needs.
- ✓ Change bid targets gradually. Large moves reset learning and cost weeks of performance.
- ✓ Performance Max needs a pipeline-weighted signal and tight exclusions, or it buys cheap conversions across more inventory.
- ✓ Microsoft Ads is a cheap supplement once Google works, not an escape from Google.
FAQ
Why are my Google Ads leads low quality?+
Almost always the conversion signal. If every form fill counts equally, smart bidding buys the cheapest ones. Assign values reflecting downstream worth and import qualified opportunities from your CRM.
How many campaigns should a SaaS Google Ads account have?+
Few enough that each accumulates meaningful conversion volume. Granular structures made sense under manual bidding and now prevent automated bidding from ever learning.
Should B2B SaaS use Performance Max?+
Only with the same pipeline-weighted conversion signal you built for Search, plus account and audience exclusions and close attention to reporting. Without those, it spreads a weak signal across more inventory than Search alone and hides where the budget went.
How do you stop wasting Google Ads budget on bad clicks?+
Work the search terms report regularly and build negatives from it. Automated bidding does not filter irrelevant traffic; it bids on it efficiently, so the negative keyword list is the maintenance that keeps broad match from widening your spend.
Is Microsoft Ads worth running for B2B SaaS?+
Often yes as a supplement. Lower competition and cheaper clicks, with an audience skewing enterprise and desktop. Expect a fraction of Google's volume and test by importing existing campaigns.
Sources
- [1]Google Search delivered 67% ROAS at the median and 138% for top performers, measured with data-driven attribution on closed-won deals. Dreamdata is an attribution vendor and its customer base skews to B2B advertisers. Dreamdata, LinkedIn Ads Benchmarks Report 2026, March 10, 2026.