Solved: ROAS Dropped After Adding Interest Targeting

Hi there,

Thanks for reaching out!

Happy to give you some of my initial thoughts and guidance on your situation. It's a really common thing to see these days, where broad targeting outperforms what we *think* should be a perfectly crafted interest-based audience. The answer isn't that interests are dead, but that the way most people use them is completely broken, and the Meta algorithm has gotten scarily good at its job, but only if you let it.

Let's get into why this is happening and how you can build a proper targeting strategy that actually works.

You need to stop paying Facebook to find non-customers...

Here’s the first uncomfortable truth about advertising on platforms like Meta. When you set a campaign objective to something vague like "Reach" or "Brand Awareness," you're giving the algorithm a very specific, and very unhelpful, command: "Find me the largest number of people for the lowest possible price."

The algorithm, being a very literal and efficient machine, does exactly what you asked. It scours your target audience and finds the users who are least likely to click, least likely to engage, and absolutely, positively least likely to ever pull out a credit card. Why? Because those users aren't in demand by other advertisers. Their attention is cheap. By choosing those objectives, you are actively paying the world's most powerful advertising machine to find you the worst possible audience for your product. It's a complete waste of money for any business that actually needs to make sales to survive.

So what happened in your case? You ran a campaign with a strong conversion objective (driving ROAS) and no interest targeting. You gave the algorithm one simple job: "Here's my pixel data, here are my ads, now go find me people who are most likely to buy my stuff, I don't care who they are." And it did its job brilliantly. It sifted through millions of user signals—behaviours, past purchases, website visits—far more effectively than you ever could with manual interest targeting.

When you then added interest targeting, you likely constrained the algorithm too much with the wrong kind of interests. You narrowed its field of vision and forced it to look for buyers in pools of people that weren't actually your buyers. This is the core of the problem, and it's not your fault; it's how most people are taught to think about targeting.

Your ICP is a Nightmare, Not a Demographic...

This brings us to the biggest reason why your interest targeting failed. Forget the sterile, demographic-based profile your last marketing hire made. "Companies in the finance sector with 50-200 employees" or "Women aged 25-40 who like yoga" tells you absolutely nothing of value. It leads to the generic, lazy interest targeting that made your ROAS plummet.

To stop burning cash, you must define your customer by their pain. You need to become an expert in their specific, urgent, expensive, career-threatening nightmare. Your customer isn't just a job title; she's a leader terrified of her best developers quitting out of frustration with a broken workflow. Your customer isn't just 'a small business owner'; he's an entrepreneur staring at a cash flow crisis, unable to make payroll next month. Your ICP isn't a person; it's a problem state.

Once you've isolated that nightmare, your entire approach to targeting changes. You stop looking for demographics and start looking for behaviours and affinities that are a direct consequence of that pain.

Let's look at a practical example. Say you sell a project management software for creative agencies. A bad, demographic-based ICP would be "Marketing Managers at agencies with 10-50 employees". The interests you'd pick would be rubbish like "Marketing", "Advertising Age", or "Ogilvy". These are way too broad and full of people who aren't your customer.

A good, pain-based ICP is: "An agency project lead who is constantly chasing creatives for updates, dealing with scope creep from clients, and using a messy combination of spreadsheets and emails to manage projects, making them look incompetent in front of their boss."

Now, what interests does that person have?

• They probably use competitor software they hate, like Asana, Monday.com, or Trello.
• They might follow project management influencers like Scott Berkun or read blogs like 'A List Apart'.
• They might be in Facebook Groups like 'Dribbble' or 'Creative Agency Owners'.

These are specific, tangible interests you can actually target. You are targeting the symptoms of their pain. This is the difference between guessing and a proper strategy. Tbh, it's wierd how few people actually do this work before spending thousands on ads.

I'd say you need a proper audience hierarchy...

So, does this mean you should abandon broad targeting? Not at all. It means you need to be much smarter about how you structure your entire ad account. When I audit client accounts, I almost always see a messy collection of campaigns with no clear strategy. The best results come from a clear, tiered approach that guides customers from initial awareness to purchase and beyond.

This is the audience prioritisation I'd usually implement, assuming the account has enough data. If not, we'd start at the top and work our way down as the data comes in.

ToFu (Top of Funnel - Cold Audiences): This is where you find new customers.

• -> Broad Targeting: Yes, keep it. For many accounts, this will be your workhorse. Give it a clear conversion objective (e.g., Purchase), great creative, and let the algorithm do its thing. This should be your benchmark.
• -> Detailed Targeting (The Right Way): This is where you test ad sets based on your pain-based ICP research. Create tightly themed ad sets around the tools, influencers, publications, and communities your ideal customer engages with. The goal here is to find pockets of users that might outperform your broad campaign.
• -> High-Value Lookalike Audiences: Once you have enough data (at least a few hundred purchases, ideally 1000+), create lookalike audiences from your best customers. Not just any customers, but your *highest value* customers. You can create a custom audience from a CSV list of customers with the highest lifetime value. A 1% lookalike of these people is often pure gold.

MoFu (Middle of Funnel - Warm Audiences): These people know who you are but haven't bought yet.

• -> Website Visitors (Last 30-90 days): Anyone who has visited your site but hasn't purchased. Exclude people who have already bought.
• -> Video Viewers: People who have watched a significant portion (e.g., 50% or more) of your video ads. This is a great indicator of interest.
• -> Social Engagers: People who have liked, commented on, or saved your posts on Facebook or Instagram. They're engaged with your brand but need a nudge.

BoFu (Bottom of Funnel - Hot Audiences): These people are on the verge of buying. This is your lowest-hanging fruit and should have the highest ROAS.

• -> Added to Cart (Last 7-14 days): People who put something in their basket but didn't complete the purchase. Hit them with ads reminding them what they left behind, maybe with a small incentive if that's part of your strategy.
• -> Initiated Checkout (Last 7-14 days): Even higher intent than an add to cart. Something stopped them at the final hurdle. Remind them to complete their purchase.

By structuring your campaigns this way, you create a full-funnel system. The ToFu campaigns feed new people into your MoFu and BoFu audiences, and you systematically convert them. You can allocate your budget based on performance, giving more to what works. Alot of the time, we see 80% of the budget going to ToFu and 20% to MoFu/BoFu retargeting, but it all depends on the account.

You'll need to know your numbers to scale...

Focusing on ROAS is good, but it can also be a trap. It tells you what happened yesterday, but it doesn't tell you how much you can actually afford to spend to acquire a customer today. The real question isn't "How high can my ROAS be?" but "How high a Customer Acquisition Cost (CAC) can I afford to acquire a truly great customer?" The answer to that lies in its counterpart: Lifetime Value (LTV).

If you don't know this number, you're flying blind. You might be turning off a campaign with a 2x ROAS that is actually profitable in the long run, or scaling a campaign with a 5x ROAS that is acquiring low-value, one-time buyers. Calculating your LTV is not that hard. You need three bits of info:

1. Average Revenue Per Account (ARPA): What do you make per customer, per month (or year, if it's a one-time purchase business, just use average order value)?
2. Gross Margin %: What's your profit margin on that revenue? Not just revenue, but actual profit.
3. Monthly Churn Rate: What percentage of customers do you lose each month? (If you have a one-off product, you can estimate repeat purchase rate instead).

The calculation is simple: LTV = (ARPA * Gross Margin %) / Monthly Churn Rate

Let's take an example for a subscription box business, like one we worked with that achieved a 1000% return on ad spend. Their numbers might look like this:

• ARPA: £40/month
• Gross Margin: 60%
• Monthly Churn: 10%

LTV = (£40 * 0.60) / 0.10
LTV = £24 / 0.10 = £240

So, each customer is worth £240 in gross margin over their lifetime. A healthy LTV:CAC ratio is generally considered to be 3:1. This means this business can afford to spend up to £80 to acquire a single customer (£240 / 3). Suddenly, a £60 CPA from a broad targeting campaign doesn't look so bad, does it? It looks like a bargain. This is the maths that unlocks aggressive, intelligent growth and frees you from the tyranny of short-term ROAS.

You probably should fix your offer...

The final piece of the puzzle, and arguably the most important, is your offer. The number one reason campaigns fail, even with perfect targeting, is a weak offer. An offer that doesn't provide enough value or doesn't have a clear audience with a need for that value is doomed from the start. It's a lack of demand.

A great offer makes the algorithm's job easier, especially with broad targeting. It acts as a filter. A powerful, specific offer will naturally repel people who aren't a good fit and attract your ideal customers like a magnet. The ad copy itself does the targeting for you.

You need a message they can't ignore. It needs to speak directly to the pain you identified earlier.

• For an eCommerce brand, you don't just sell a product; you sell a transformation. Instead of "Buy our new skincare serum," try using the Before-After-Bridge framework. "(Before) Tired of hiding your uneven skin tone under layers of makeup? (After) Imagine waking up to clear, glowing skin you're proud to show off. (Bridge) Our Vitamin C serum is the bridge that gets you there in 30 days."
• For a B2B service, you don't sell your service; you sell a solution to a costly problem. Use Problem-Agitate-Solve. "(Problem) Are your cash flow projections just a shot in the dark? (Agitate) Are you one bad month away from a payroll crisis while your competitors are confidently raising their next round? (Solve) Get expert financial strategy for a fraction of a full-time hire. We build dashboards that turn uncertainty into predictable growth."

A powerful offer, combined with a conversion-focused campaign, gives the algorithm clear instructions. It's not just looking for people who buy things; it's looking for people who respond to your specific message about a specific pain point. This is how you make broad targeting laser-focused without even using interest targets.

This is the main advice I have for you:

So, to bring it all together, here is what I would recommend you do. It's a process, and it takes some work upfront, but it's what separates the advertisers who get consistent results from those who are always guessing. I've detailed my main recommendations for you below:

Your experience wasn't an anomaly; it was a lesson from the Meta algorithm. It's telling you that its ability to find your customers is now often better than our manual ability to describe them with crude interest targeting. The solution isn't to abandon one for the other, but to build a more sophisticated system where you use Broad as a powerful discovery tool, you test hyper-relevant interests based on real customer psychology, and you capture every last bit of value with a robust retargeting funnel. All of this is underpinned by knowing your numbers and having an offer that truly resonates.

It can be alot to take in and even more to implement correctly, especially when you're busy running a business. Getting this stuff right—the deep customer research, the financial modelling with LTV, the complex campaign structures, and the constant testing—is a full-time job. It’s what separates businesses that struggle to get a consistent return from their ad spend from those that scale profitably and predictably.

If you'd like to go over your specific situation in more detail, we offer a free, no-obligation initial consultation where we can look at your ad account together and map out a more concrete strategy. It's often really helpful for business owners to get a second pair of expert eyes on their setup.

Regards,

Team @ Lukas Holschuh