TLDR;
- Most B2B SaaS founders assume you must exclusively target the US and optimise for deep-funnel events (like Purchases) to get quality users, believing top-of-funnel registrations are a waste of cash.
- Data from an $11k campaign driving 4,622 software registrations shows US-only targeting costs up to 124x more ($38.57 CPA) than Worldwide targeting ($0.31 CPA) for the exact same conversion event, while optimising for Purchases fails entirely due to the platform's 7-day attribution window.
- To survive in high-friction niches, you have to run a wide geographic net to feed the algorithm cheap data at the top of the funnel (TOFU), then aggressively qualify those users with retargeting (MOFU/BOFU) and lookalike audiences seeded from your best existing customers.
- Check out the interactive CPA geographic modeler, funnel flowcharts, and targeting frameworks below to map this math to your own SaaS economics.
I see a lot of founders in B2B SaaS—especially in high-friction or regulated spaces like software for the gambling niche or eCommerce seller tools—burning through their ad budgets in a matter of weeks. They set up their Meta or Google Ads accounts, target "US Only", optimise straight for "Purchases", and then wonder why they’re paying $150 a click with zero conversions to show for it.
Tbh, the advice you usually read on forums is mostly to blame. People tell you to restrict your targeting to the most competitive countries and the deepest funnel events because that is where the quality resides. It sounds logical. But the ad platform algorithms do not care about your logic. They care about data liquidity. If you choke the system, it will charge you a massive premium just to deliver an impression.
To show you exactly how this plays out, I am going to walk you through a forensic analysis of a specific campaign we ran for a B2B SaaS client. This was a software tool designed to help Amazon sellers manage their feedback and review requests. It is a very specific B2B proposition. Over the course of the campaigns, we spent $11,013.42 and drove 4,622 software registrations.
That averages out to $2.38 per registration. But that average is a complete lie. It hides the brutal reality of how much certain targeting choices will actually cost you. Let us break open the raw data and see what is really going on under the hood of a B2B software campaign.
The Dataset and Provenance
Before we get into the numbers, let us be clear about what we are looking at. This is not some aggregated industry benchmark report. This is raw export data from a single Meta Ads account for one specific B2B SaaS client. The data covers 12 distinct campaigns spanning different stages of the funnel (Top of Funnel, Middle of Funnel, Bottom of Funnel), testing various geographic tiers and optimization events.
We are dealing with 490,084 total impressions, $11,013.42 in ad spend, and a highly segmented campaign structure. I have not invented any of these numbers. What you are about to see is the unvarnished math of how an ad account actually behaves when you start pulling different levers.
The Geographic Penalty: Why US-Only Targeting Will Bleed You Dry
The most stubborn myth in B2B SaaS advertising is that you have to target the United States exclusively if you want customers who will actually pay. Fair play, the US has high purchasing power. But let us look at what the algorithm charges you for that privilege.
I isolated our Top of Funnel (TOFU) campaigns that were all optimising for the exact same event: "Complete Registration". Here is the actual cost breakdown based on the geographic targeting:
- TOFU (US Only): We spent $925.60 and got 24 registrations. That is a Cost Per Registration of $38.57.
- TOFU (Developed Countries): We spent $1,337.42 and got 232 registrations. That is a Cost Per Registration of $5.76.
- TOFU (Worldwide): We spent $336.75 and got 1,083 registrations. That is a Cost Per Registration of $0.31.
Think about that math for a second. Targeting the US exclusively cost us 6.7 times more than targeting a broader list of developed countries (which includes places like the UK, Germany, Australia, etc.). Worse, it cost us 124 times more than targeting worldwide.
If you are running a B2B SaaS in a tough niche—like gambling software—and you insist on targeting only the US, you are fighting a massive algorithmic penalty. You are bidding in the most saturated auction on the planet.
Geographic Cost Per Registration Penalty
Cost disparity across TOFU campaigns optimising for the exact same "Complete Registration" event.
Data source: Single B2B SaaS account export, $11k total spend.
Yes, a lot of those $0.31 worldwide registrations will not convert into paying users. If you do not care about quality, you can run worldwide and just accept the junk. But the pragmatic play is to use a blended approach. We usually start with the countries where most of your existing paying customers are located. If you want to expand to get cheaper data into the pixel, you do it carefully. You run separate campaigns for developed countries versus developing countries. Never mix them in the same ad set, or the platform will just dump all your budget into the cheapest, lowest-intent regions.
The Optimization Trap: Why "Purchases" Breaks the Machine
The second major mistake founders make is telling the ad platform to optimise directly for "Purchases" when they have a free trial model.
In this dataset, we ran a Bottom of Funnel (BOFU) campaign optimising for Purchases. It spent $205.49 and generated 36 registrations at $5.71 each. But here is the catch: the Meta ad platform operates on a 7-day click attribution window. If your SaaS has a 14-day or 30-day free trial, the purchase happens outside the algorithm's memory. The system cannot see the conversion, so it thinks the ads are failing. It stops delivering, costs spike, and the campaign dies.
Look at how the costs compare when we optimise for earlier events that happen within that 7-day window:
- MOFU (Retargeting) - Complete Registration: $987.93 spend, 1,122 registrations at $0.88 each.
- BOFU (Retargeting) - Trials: $314.68 spend, 156 registrations at $2.02 each.
Algorithmic Attribution Drop-off
You must optimise for an event that happens quickly. For software, that usually means optimising for the initial registration or trial start, and then relying on your email nurturing, in-app onboarding, and BOFU retargeting ads to push them towards the actual purchase.
The Illusion of Detailed B2B Targeting
When you are selling niche software, you naturally want to use highly specific detailed targeting. In this dataset, the client wanted to reach Amazon sellers, so we tested detailed interests like online marketplace, drop shipping, Shopify, and Amazon Marketplace.
I drilled down into the ad set data for the developed countries complete registration campaign. This campaign generated 232 registrations in total. Here is exactly where those registrations came from:
- Lookalike 1% of Complete Registrations: 150 registrations
- Lookalike 1% of Purchases: 38 registrations
- Detailed Targeting (Small Business Owners): 27 registrations
- Lookalike 1% of Website Visitors: 13 registrations
- Lookalike 1% of Trials: 4 registrations
- Online Retail Interests / Marketplace Interests: 0 registrations (turned off due to underperformance)
Look at that spread. The Lookalike audiences based on hard pixel data accounted for 205 out of 232 registrations (88%). The highly logical, detailed interest targeting completely bombed.
Why? Because interests are messy. Just because someone likes the Amazon page does not mean they own an Amazon business; they might just buy household goods on Prime. But a Lookalike audience seeded from your actual paying customers gives the platform a concrete mathematical pattern to match.
Audience Targeting Reliability Hierarchy
Tier 1: High-Intent Lookalikes (The Workhorse)
LLAs seeded from bottom-of-funnel events (Purchases, Trials, Registrations). In our data, these drove 88% of conversions. They require a warmed-up pixel but deliver the most reliable CPAs.
Tier 2: Broad Demographic Proxies (The Backup)
Targeting generic B2B markers like small business owners or business page admins. These drove about 11% of volume. Good for early testing before you have enough pixel data for LLAs.
Tier 3: Detailed Niche Interests (The Trap)
Targeting specific software names or industry jargon. Often too small or too polluted with consumers. In this dataset, they failed to generate meaningful conversions and were turned off.
If you are launching a SaaS in a tricky niche, do not waste weeks trying to find the magic interest button. Use broad proxies to get initial traffic, and as soon as you have 100+ events, shift your budget aggressively into Lookalikes of those exact users.
The Winner-Takes-All Creative Reality
We tested seven different ad variations in this account. If you listen to basic marketing advice, you would expect a somewhat even spread of results, maybe a 60/40 split between the top few ads.
That is not how digital advertising works. It is a brutal, winner-takes-all environment. I pulled the ad-level breakdown for 1,046 of the registrations. Here is the distribution:
- Ad 1 ("Unlock amazon success"): 633 registrations (60.5%)
- Ad 2 ("Elevate your Amazon brand"): 349 registrations (33.3%)
- Ad 3 ("Boost your Amazon sales"): 62 registrations (5.9%)
- Ad 4 ("Email and alerts"): 2 registrations (0.1%)
- Ads 5, 6, 7: 0 registrations
Just two ads generated almost 94% of the volume. The algorithm identified early on which creatives were capturing attention and getting clicks at a cheaper rate, and it relentlessly funnelled all the budget into them.
This is why you absolutely must split-test creatives continuously. If we had only launched the email and alerts ad, this entire campaign would have been a catastrophic failure. You cannot guess what the market wants; you have to let the math decide.
The Math of Creative Concentration
Comparing the theoretical expectation of ad testing versus the empirical reality of algorithmic budget concentration.
If you only test one or two angles, you are highly likely to miss the outlier creative that makes the unit economics viable.
Modelling Your Own B2B Acquisition Strategy
When you look at this data—the massive cost difference between countries, the failure of deep-funnel optimization events, and the reliance on lookalike audiences—it paints a very clear picture of how you have to run B2B campaigns today.
If you are operating a SaaS in the gambling niche, for instance, you already have massive friction. Ad platforms are incredibly restrictive about gambling-related content, and the B2B side (selling software to operators) is a tiny subset of the broader market. If you try to force the algorithm to find US casino owners and optimise for purchases, your account will stall immediately.
Instead, you need a strategy that feeds the machine. You have to broaden your geographic targeting to get cheap registrations and pixel fires. You have to give away a free trial or a low-friction tool to generate top-of-funnel events. Then, you let your middle and bottom of funnel retargeting campaigns (which run on much smaller budgets, as seen in our $314 BOFU Trial spend) do the heavy lifting of qualifying those leads.
To help you visualise how geographic blending impacts your overall Cost Per Registration, I have built a simulator below based on the actual CPA ratios we observed in this dataset.
Geographic Blended CPA Simulator
Adjust the slider to see how allocating budget away from US-only traffic towards a blended Worldwide/Developed mix impacts your top-of-funnel registration costs.
The Cynical Truth About Scaling B2B SaaS
At the end of this campaign, the client had to cut their budget. Why? Because acquiring 4,622 users is only half the battle. If those users are not converting into paying customers on the backend, or if the product has a leaky bucket, you cannot scale the ad spend. As you push budget higher, your CPAs will naturally inflate because you exhaust the lowest-hanging fruit in your audiences.
You cannot fix a broken business model inside the Ads Manager. But if your software is solid, your economics are sound, and you are tired of setting cash on fire trying to force the algorithm to do things it hates, you need to change your architecture.
Stop fighting the machine. Give it broad targets, give it pixel data it can actually track, and let your landing pages do the selling. If you are struggling to make this math work for your own SaaS, we offer a free initial consultation where we will pull apart your ad account and see exactly where the algorithm is taxing you. Sort the tracking, structure the funnel properly, and the results usually take care of themselves.
Cite this Research & Data
If referencing this research, data, or strategy in your publication, please use the attribution below.
Lukas Holschuh
Founder, Growth & Advertising Consultant
Great campaigns fail without expertise. Lukas and his team provide the missing strategy, optimizing your entire advertising funnel—from ad creatives and copy to landing page design.
Backed by a proven track record across SaaS, eLearning, and eCommerce, they don't just run ads; they engineer systems that convert. A data-driven partnership focused on tangible revenue growth.