Intent-Based Catalog Segmentation for 40% ROAS Lift

A 40% ROAS Lift Wasn’t Built for You — But the Playbook Can Be

Meta’s Advantage+ Shopping Campaigns delivered a documented 40% improvement in ROAS for retail advertisers using product-set optimization. The problem? If you sell software, financial services, or B2B solutions, you don’t have a product catalog in the traditional sense. You have services, tiers, use cases — things that don’t fit neatly into a feed. But the underlying mechanics — intent-based catalog segmentation and dynamic creative matching — are vertical-agnostic. You just need to rebuild them differently.

What Meta Actually Did (and Why It Worked)

Let’s demystify the retail media playbook before we port it elsewhere. Meta’s product-set optimization works by breaking a retailer’s full catalog into subsets — think “running shoes under $120” or “winter jackets for women” — and then dynamically serving the right creative to the right user based on behavioral signals. The algorithm doesn’t show every product to every person. It matches intent signals to curated subsets, then wraps the winning product in a dynamically assembled creative unit.

Three things make this work:

• Signal ingestion: The system reads purchase behavior, browse history, cart activity, and lookalike patterns to infer intent.
• Catalog segmentation: Products are grouped into logical sets that correspond to different buyer states — discovery, consideration, conversion.
• Dynamic creative assembly: Ad copy, imagery, and CTAs are composed on-the-fly to match the inferred intent and the selected product set.

Retailers have a structural advantage here: their “catalog” is already structured data with SKUs, prices, images, and categories. But the logic isn’t retail-specific. It’s intent-to-offer matching. And that’s something every vertical can build.

Rebuilding the “Catalog” When You Don’t Sell Products

The first challenge for non-retail verticals is defining what your catalog even is. In SaaS, it might be feature sets mapped to personas. In financial services, it could be product tiers — checking accounts, investment portfolios, insurance bundles. In education, it’s programs, certifications, course tracks. The key is structuring your offerings as discrete, targetable units with distinct value propositions.

Here’s the framework we’ve seen work across verticals:

Rebuilding the "Catalog" When You Don't Sell Products

1

Audit Your Offer Architecture:

List every distinct thing you sell or promote. Don’t think in terms of your org chart — think in terms of what a buyer would compare side by side. A cybersecurity firm doesn’t have one product; it has endpoint protection, cloud security, compliance solutions, and managed detection. Each is a "product set."

2

Map Offers to Intent Tiers:

Categorize each offer by the buyer intent it satisfies. Some offers answer early-stage questions ("What is zero-trust architecture?"). Others serve high-intent buyers ready to evaluate vendors. This mapping is your segmentation layer.

3

Create Structured Data Feeds:

Build a feed — even a simple spreadsheet initially — that includes offer name, description, target persona, intent tier, primary benefit, supporting proof point, and associated creative assets. This becomes your pseudo-catalog that platforms can ingest.

4

Define Signal-to-Segment Rules:

Specify which behavioral signals (page visits, content downloads, search queries, community mentions) map to which catalog segments. Someone reading comparison content gets served your competitive differentiator set. Someone engaging with thought leadership gets the awareness set.

Dynamic Creative Matching Without a Product Feed

Meta’s dynamic creative optimization (DCO) pulls product images, prices, and descriptions directly from a catalog feed. Non-retail advertisers need to simulate this. The good news: the tools exist. The bad news: most teams don’t connect them properly.

Meta’s Advantage+ creative now supports custom catalog structures beyond traditional retail feeds. You can upload service-based catalogs with custom columns. Google’s Performance Max campaigns offer similar flexibility through asset groups. The platform infrastructure has caught up — the gap is in how teams structure their inputs.

What dynamic creative matching looks like in practice for a B2B SaaS company: a prospect who visited your pricing page and downloaded a security whitepaper sees an ad featuring your enterprise security tier with a “Request a Demo” CTA and a customer proof point from a similar-sized company. Meanwhile, someone who only read a blog post sees a thought-leadership-style ad with an ungated resource offer. Same campaign. Different catalog segment. Different creative assembly.

This isn’t hypothetical. Companies using Intercept to capture intent signals across platforms are already feeding those signals back into their media buying to power exactly this kind of segmentation. When you know what someone is actively researching — not just that they visited a page — the creative matching becomes dramatically more precise.

Where Intent Data Makes or Breaks the Model

Here’s what separates a mediocre implementation from one that actually moves ROAS: the quality and specificity of your intent signals. Most teams rely on first-party behavioral data — site visits, email opens, form fills. That’s table stakes. The real lift comes from capturing intent signals before someone hits your site.

Think about it. By the time a prospect visits your pricing page, they’ve already done 70% of their research elsewhere — on Reddit, in LinkedIn discussions, through peer recommendations, on review sites like G2 or Capterra. If your catalog segmentation only fires based on on-site behavior, you’re optimizing the last 30% of the journey.

Intent-based lead generation platforms solve this by monitoring buying signals across the open web. When someone asks “best alternatives to [your competitor]” on Reddit or discusses a pain point your product solves on LinkedIn, that’s a signal. Feeding those signals into your catalog segmentation layer means your dynamic creative fires earlier, with more context, at a moment when the buyer is actively forming preferences. Our insights on intent data go deeper into how this signal capture works at scale.

A Practical Implementation Roadmap

Enough theory. Here’s how to actually build this for a non-retail vertical, step by step:

One critical note: don’t skip the attribution layer. If you’re pulling intent signals from off-site sources, you need a way to connect those signals to downstream conversions. Understanding agentic referral attribution becomes essential when your signal sources span multiple platforms and touchpoints.

1

Build Your Service Catalog Feed:

Structure your offerings into 5-15 distinct "product sets." Each set needs a unique value prop, target persona, and at least three creative variations (headline, description, image/video, CTA). Use Google Sheets or a lightweight PIM — don’t overthink the tooling.

2

Instrument Your Intent Signals:

Layer first-party data (site behavior, CRM engagement) with third-party intent data. Platforms like Intercept capture real-time buying signals from community discussions, forums, and social platforms. Map each signal type to a catalog segment.

3

Configure Platform-Side Segmentation:

In Meta, create custom catalogs with your service feed. In Google Ads, build asset groups within Performance Max that correspond to your catalog segments. On LinkedIn, use matched audiences layered with intent-based exclusions and inclusions.

4

Set Up Dynamic Creative Rules:

Define which creative elements (headline, image, CTA, proof point) change based on the catalog segment being served. Start simple — even two variations per segment is better than a single static creative for everyone.

5

Measure by Segment, Not by Campaign:

The whole point is granularity. Track ROAS, CPA, and conversion rates at the catalog-segment level. You’ll quickly discover that two or three segments drive disproportionate returns — double down there.

6

Iterate on Signal-to-Segment Mapping:

This is where ongoing optimization lives. As you gather data, refine which signals map to which segments. A signal you assumed indicated high intent might actually correlate with tire-kicking. Let the data reshape the mapping quarterly.

Why This Works Better Than Broad Targeting

Statista reports that global digital ad spend will exceed $870 billion in 2027, yet average conversion rates remain stubbornly low — under 4% for most industries. The problem isn’t reach. It’s relevance. Broad targeting with generic creative is the media buying equivalent of shouting into a crowd.

Product-set optimization — whether you call it that or “intent-based catalog segmentation” — works because it reintroduces specificity at scale. You’re not personalizing one-to-one (that’s expensive and fragile). You’re personalizing one-to-segment, which is sustainable and measurable. Meta proved the model in retail. The opportunity now is for every other vertical to adopt the same architecture with intent data as the foundation.

The playbook is clear: structure your offers like a catalog, capture intent signals early and broadly, match segments to creative dynamically, and measure at the segment level. Teams that do this in the next two quarters will have a compounding advantage over those still running static campaigns with broad audiences.

Start by building your service catalog feed this week. Everything else follows from there.

FAQs

What is product-set optimization and how does it apply outside of retail?

Product-set optimization is Meta’s approach to grouping catalog items into subsets and dynamically serving the most relevant set to each user based on behavioral signals. Outside retail, you replicate this by structuring your services, tiers, or solutions as discrete “product sets” in a custom catalog feed, then using intent signals to determine which set each prospect sees.

How do you build a catalog feed for a service-based or B2B business?

Create a structured data feed — even a spreadsheet works initially — listing each distinct offer with its name, description, target persona, intent tier, primary benefit, proof point, and associated creative assets. Upload this as a custom catalog in Meta or structure it as asset groups in Google Performance Max.

What intent signals should non-retail advertisers use for catalog segmentation?

Combine first-party signals (site visits, content downloads, CRM engagement) with third-party intent data from community discussions, review sites, forums, and social platforms. Signals like competitor comparison searches, pain-point discussions, and category-specific questions on Reddit or LinkedIn indicate active buying intent and should map to specific catalog segments.

Can dynamic creative optimization work without a traditional product feed?

Yes. Meta’s Advantage+ creative and Google’s Performance Max both support custom catalog structures beyond retail feeds. You define creative components — headlines, descriptions, images, CTAs — for each service segment, and the platform assembles the optimal combination based on user signals and segment mapping.

How do you measure the success of intent-based catalog segmentation?

Track ROAS, CPA, and conversion rates at the individual catalog-segment level rather than at the campaign level. This granularity reveals which intent-to-segment mappings drive the strongest returns, allowing you to reallocate budget toward high-performing segments and refine underperforming ones.