TLDR;
- Many software founders assume that broad interest-based targeting is the safest way to find new users, but the data suggests it often leads to higher costs and lower intent.
- In our investigation of 4,622 registrations, Lookalike Audiences (LLAs) based on down-funnel events—specifically past purchases and registrations—consistently outperformed general interest targeting for B2B software signups.
- The data proves that a worldwide LLA strategy can drive signups for as little as $0.31 each, provided the creative handles the heavy lifting of qualification.
- See the interactive audience comparison database and the LTV-to-CAC scenario modeler below to see how these metrics shift your profitability.
I’ve seen a lot of B2B software founders bin their entire ad budget because they’ve been told that "detailed interest targeting" is the only way to reach their niche. They spend weeks meticulously picking out interests like "Cloud Computing" or "SaaS" and then wonder why their CPA is through the roof. It’s a common trap. You think you're being precise, but you're actually just paying Facebook a premium to show your ads to people who might have once liked a post about a competitor three years ago.
I wanted to look at this properly, using a specific dataset from a campaign we ran for a B2B software product called FeedbackWhiz. For context, this is a tool for Amazon sellers to manage reviews and feedback. It’s a classic B2B SaaS setup. We managed to pull in 4,622 registrations at an average cost of $2.38. But the real insight isn't in that top-line number; it’s in how different targeting sets fought it out for those signups. I’ve crunched the numbers on lookalike audiences versus those "detailed interests" everyone loves, and the results are pretty blunt.
Performance Snapshot: 4,622 Registrations
This snapshot summarizes the high-level metrics derived from the account audit, focusing on the efficiency of the software registration funnel across a $11,013.42 total spend.
How do Lookalikes actually stack up against Interest targeting in a real-world B2B environment?
When I started digging into the ad set level data, a very clear pattern emerged. We weren't just guessing; we ran head-to-head tests. On one side, we had "Detailed Targeting" like "Small Business Owners" and specific "Online Retail Interests." On the other, we had 1% Lookalike Audiences (LLAs) based on three different quality markers: Website Visitors, Completed Registrations, and Purchases.
If you've been on any marketing forum recently, you'd probably hear that "interest targeting is dead" or "LLAs don't work post-iOS 14." I’d say that’s a bit of a generalisation. In this account, the "Small Business Owners" interest group actually pulled its weight, bringing in 27 registrations. However, it was the LLAs that provided the scale. The LLA based on "Complete Registrations" brought in 150 signups—more than five times the volume of the interest-based set in the same period.
The logic here is simple: detailed interests rely on what Facebook *thinks* someone is interested in based on their browsing. An LLA based on a "Complete Registration" pixel event relies on what Facebook *knows* a user has actually done. It's the difference between targeting someone who likes looking at gym clothes and someone who’s actually just walked into the gym. For software, where the barrier to entry is a signup, the "actual behavior" data of an LLA is almost always going to win over the "assumed interest" of a category.
The CPA Showdown: Interest vs. Lookalike
Comparison of Cost Per Registration (CPR) across different targeting methodologies within the TOFU campaigns. Lower bars indicate better efficiency.
Data source: Internal campaign audit of LH-TOFU segments. Lookalikes were consistently 60-80% more efficient than broad interest categories.
Is location the hidden factor that's actually driving these registration costs?
While the LLA vs Interest debate is important, my research found a much more massive lever in the data: Geography. If you look at the raw numbers, the "LH - TOFU (US) - Complete Registration" campaign had a cost per registration of $38.57. Compare that to the "LH - TOFU (Worldwide) - Complete Registration" campaign, which sat at $0.31.
That is a 124x difference in cost. You’d think I’d be shouting from the rooftops to just move everything to worldwide, but I’d say you've got to be careful. Worldwide traffic is often the "junk food" of the advertising world. It’s cheap, and it makes your charts look amazing, but if those 1,000 signups from developing countries don’t turn into a single paying customer, you’ve just spent $310 on an empty metric.
However, for a SaaS product, these worldwide campaigns serve a tactical purpose. They feed the pixel. By getting cheap registrations (even if the trial-to-paid conversion rate is low), we are teaching the Facebook algorithm what a "registrant" looks like. We can then use that data to build better LLAs for the US and "Top Countries" (UK, DE, FR, IT, ES). In our data, the "Developed Countries" registrations came in at $5.76. This is the "sweet spot"—much cheaper than the US but much higher quality than the worldwide average.
Geographic CPA & Scale Modeler
Model the impact of shifting your budget between high-cost (US) and high-volume (Worldwide) targeting based on the FeedbackWhiz dataset.
What happens when the "Detailed Targeting" actually works—but can't scale?
I noticed something interesting with the "Small Business Owners" interest group. It didn't have the massive registration volume of the LLAs, but it was incredibly steady. It’s the kind of audience that acts as a baseline. For software, this is your "safe bet." You know these people are in business, you know they likely have an Amazon account, and you know they need a solution.
However, "safe bets" are rarely "scale bets." Detailed interest targeting is limited by how many people explicitly show interest in those specific tags. In our dataset, the "Online Retail Interests" and "Marketplace Interests" ad sets had to be turned off because they simply couldn't keep up with the efficiency of the lookalikes. They were "dodgy" from the start—pulling in clicks that didn't convert into trials.
This brings me to a core principle: LLAs are better at finding the "un-obvious" customer. If I target "Shopify" as an interest, I’cmathing every other person selling a "get rich quick" dropshipping course. But if I use an LLA of people who have actually *purchased* FeedbackWhiz, Facebook might find someone who doesn't list "Shopify" or "Amazon" as an interest, but behaves exactly like my best customer. That’s where the real profit is made.
Targeting Efficiency Matrix
This table maps the targeting options tested in the campaign against their primary performance characteristics and our recommendation for software scaling.
| Targeting Type | Best Result (CPA) | Volume Potential | Strategic Recommendation |
|---|---|---|---|
| LLA 1% Complete Reg | $2.01 | High | Primary scaling audience for TOFU. |
| Small Business Owners (Interest) | $5.80 | Moderate | Stable baseline for US targeting. |
| LLA 1% Purchases | $3.40 | Low/Medium | Highest quality intent; use for BOFU. |
| Online Retail / Marketplace | $12.00+ | Low | Bin it; too broad for software B2B. |
Can we talk about the Elephant in the room: Creative is the real "Targeting"?
I’d be doing you a disservice if I just talked about audiences. In the FeedbackWhiz campaign, we didn't just test audiences; we hammered the creatives. We found that the ad copy "Automate your Amazon feedback" brought in 1,046 registrations. The next best, "Unlock amazon success," brought in 633.
Why does this matter for targeting? Because the creative acts as the ultimate filter. If my ad says "Automate your feedback," only someone who *actually has a feedback problem* is going to click it. This "pre-qualification" happens before the user even lands on the page.
If you're using LLAs, you're giving the algorithm a general direction. But the *ad itself* is what stops the right person from scrolling. I’ve seen campaigns where the audience targeting was "Broad" (no interests, no LLAs) and it still beat the narrow interests because the ad copy was so specific to a nightmare the customer was facing. For software, that nightmare is usually manual, repetitive work. "Automate" is the magic word because it promises to kill that nightmare.
The Software Conversion Funnel
Wait, are we over-optimising for the "Signup" and forgetting the "Purchase"?
There is a dangerous point in every SaaS campaign where you become obsessed with the Cost Per Registration. In our case, $2.38 is fantastic. But as I said to the founder of FeedbackWhiz, registrations are a lead metric, not a revenue metric.
The data shows we ran "LH - BOFU - Purchase" and "LH - BOFU - Trials" campaigns. This is where the real business happens. In B2B software, your trial period is often longer than the 7-day attribution window Meta uses. This is a nightmare for the algorithm. It means Facebook might not "see" the conversion that happens 14 days after the click.
This is why lookalike audiences are superior to interests for the long game. An LLA of *Purchasers*—even if the sample is small (we had 38 registrations from this LLA specifically at the time of the export)—tells Facebook the profile of someone who is willing to pay. Broad interests just find you "lookers." LLAs find you "buyers." If you're only optimising for signups because they're cheap, you're building a house of cards. You need to allocate at least 20% of your budget to BOFU (Bottom of Funnel) campaigns targeting the actual conversion event, even if the CPA looks "expensive" compared to a registration.
SaaS Unit Economics & Allowable CAC
Determine how much you can actually afford to pay for a registration based on your software's lifetime value.
So, what’s the verdict on LLAs vs. Interests for 2026?
I’ve run quite a few campaigns for B2B SaaS, and the conclusion from this investigation is right on track with what I’m seeing elsewhere. For software signups, detailed interest targeting is a starting point, not a destination. If you're still relying on "Business Page Admins" or "E-commerce" interests after your first 1,000 signups, you're leaving money on the table.
The hierarchy of targeting for software should always be:
- LLAs of Purchasers (The Gold Standard)
- LLAs of Registrants (The Scale Driver)
- Specific Interests (The Baseline)
- Broad Targeting (The Future - once the pixel is seasoned)
In our research, the LLA 1% based on Registrations delivered the best balance of scale and cost ($2.01). If you're struggling to scale your software, I’d bet my last pound that you're being too narrow with your interests or too scared to let the LLAs run worldwide to feed your pixel.
Of course, all of this targeting is useless if your landing page looks like it was built in 2005. For FeedbackWhiz, the campaign's success was as much about the conversion-focused landing page as it was about the ads. If people land on your site and can't find the "Free Trial" button within three seconds, you might as well just set your budget on fire.
If you’re feeling like your current campaigns have plateaued, or if you’re sick of seeing your CPA climb every time you try to scale, we should probably chat. I’ve seen these patterns dozens of times across SaaS, eCommerce, and eLearning. It’s usually a simple fix of campaign structure and a bit of a shift in targeting philosophy.
I’d like to invite you to book in a free strategy consultation. We’ll jump on a call, I’ll take a proper look at your Meta or Google Ads account, and we’ll figure out where the leakage is happening. No corporate fluff, just a straight-up look at the data and a plan to get your acquisition costs sorted. I look forward to speaking with you.
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.