The luxury advertising industry has a measurement problem. Most brands optimize for ROAS (Return on Ad Spend) as their north star metric. It's clean, it's simple, and it's almost certainly giving you the wrong answer.
ROAS measures the revenue generated per advertising dollar spent. Sounds logical. But in luxury, where a single customer might browse for three months before making a €5,000 purchase, attributing that sale to the last ad they clicked is absurd. The Instagram ad they saw in January planted the seed. The Google search ad in February kept them engaged. The TikTok video in March brought them back. ROAS credits only one of those touchpoints and ignores the rest.
After managing digital advertising budgets of tens of millions of euros annually across every major platform (Google, Bing, Baidu, Naver, Meta, TikTok, Snapchat, Pinterest, LinkedIn, Reddit), I can tell you with certainty: the brands that win in luxury advertising aren't optimizing for ROAS. They're building predictive models.
The Propensity-to-Purchase Model
The concept is straightforward, even if the execution isn't. Instead of targeting customers based on demographics or past purchases, we built an AI/ML model that scored every user action on the site according to its predictive value for purchase intent.
Not all actions are equal. Viewing a product page might get a score of 5. Adding to wishlist: 25. Adding to cart: 50. Viewing the shipping policy page: 35 (surprisingly strong signal, because it means someone is seriously considering buying). Visiting the returns page: 15. Each micro-conversion received a weighted value based on historical correlation with actual purchases.
These propensity scores were then fed directly into Google's advertising platform as custom signals, allowing the campaign algorithms to target users based on predicted purchase intent rather than broad demographic or interest categories.
The result, documented by Google in their Think with Google AI Excellence series: the predictive model outperformed the control group on campaign efficiency (Source: Google Think with Google, "LuisaViaRoma: l'intelligenza artificiale in un e-commerce di moda").
Why Traditional Targeting Fails for Luxury
Standard digital advertising segments audiences by age, gender, location, interests, and past behavior. A 35-year-old woman in Dubai who follows luxury fashion accounts on Instagram becomes a target. This is better than nothing, but it's crude.
The problem with demographic targeting in luxury is that luxury customers don't conform to demographic patterns the way mass-market consumers do. NIMA Digital's experience across 25+ years in luxury digital shows that a 22-year-old graduate student and a 55-year-old executive can have identical purchasing behavior on a luxury site, with identical lifetime value. Demographics tell you almost nothing.
Behavioral signals tell you everything. But not just any behavioral signals. The right ones, weighted correctly, analyzed in sequence.
The TikTok Symphony Breakthrough
In 2024, we became the first Italian brand leadership team to deploy TikTok Symphony AI avatars for luxury advertising. The results were dramatic:
- •+71% click-through rate compared to standard creative
- •-11% cost per acquisition
- •80% of all clicks were driven by AI-generated avatar content
(Source: TikTok for Business case study)
These numbers need context. TikTok's audience skews younger than most luxury brand marketers are comfortable with. The conventional wisdom in luxury is that TikTok is for brand awareness, not conversion. The data says otherwise, if your creative is right.
The AI avatars worked because they solved a specific production bottleneck. High-quality luxury video creative traditionally requires expensive shoots, talent booking, and weeks of production. TikTok's algorithm demands fresh creative every 7-10 days. The math doesn't work with traditional production.
AI avatars bridged that gap: brand-appropriate creative at the velocity TikTok demands. The 80% click share from AI avatars wasn't because they were gimmicky. It was because we could test 10x more creative variations and let the algorithm find winners faster.
Smartly.io and the Creative Optimization Loop
Running alongside TikTok Symphony, our deployment of Smartly.io for Video Shopping Ads delivered:
- •-30% CPA (cost per acquisition)
- •+65% ROAS
- •+8.3% CTR
(Source: Smartly.io case study)
Smartly.io's value in luxury advertising isn't just automation. It's the creative optimization loop. The platform tests variations of ad creative (headlines, product images, video cuts, call-to-action placement) at a scale no human team can match, then allocates budget toward the winning combinations in near-real-time.
For luxury brands, this raises a legitimate concern: does automated creative optimization compromise brand integrity? The answer depends entirely on how you set up the guardrails.
NIMA Digital's approach was to define strict brand guidelines within Smartly.io (approved fonts, color palettes, imagery styles, copy tone) and then let the AI optimize within those constraints. The machine handles the "what converts better" question. The brand team handles the "what's on-brand" question. Neither overrules the other.
The POAS Revolution: Profit Over Revenue
ROAS tells you revenue per ad dollar. It doesn't tell you profit.
A €200 item with 60% margin generates €120 profit on a sale. A €500 item with 20% margin generates €100 profit. ROAS optimization will push budget toward the €500 item. Profit optimization will correctly allocate more budget to the €200 item.
At NIMA Digital, we pioneered the shift from ROAS to POAS (Profit on Ad Spend) optimization for luxury e-commerce. This required integrating margin data from the product catalog into the advertising platform's optimization signals. A non-trivial data engineering task, but one that fundamentally changed how budget was allocated.
Attribution: The Unsolved Problem
I won't pretend attribution is solved. It isn't. Every attribution model is a simplification of reality. But some simplifications are less wrong than others.
NIMA Digital uses Fospha for Marketing Mix Modeling, a probabilistic approach that estimates each channel's true contribution based on statistical modeling rather than deterministic click tracking. It's not perfect, but it corrects the most egregious errors of last-click attribution: the chronic undervaluation of upper-funnel channels (social, video, display) and the overvaluation of lower-funnel channels (branded search, retargeting).
For luxury brands with long purchase cycles and high average order values, accurate attribution isn't a nice-to-have. It's the difference between investing in the channels that actually drive growth and doubling down on channels that merely capture existing demand.
Platform-by-Platform: Where Luxury Budgets Should Go in 2026
Based on managing advertising across every major platform for luxury e-commerce, here's NIMA Digital's perspective on platform allocation for luxury brands targeting MENA and European markets:
| Platform | Role in Luxury | Budget Share (indicative) |
|----------|---------------|--------------------------|
| Google (Search + Shopping) | Demand capture, high-intent | 30-40% |
| Meta (Instagram + Facebook) | Brand building, retargeting | 20-25% |
| TikTok | Creative testing, younger audience | 10-15% |
| Snapchat | Gulf markets, discovery | 5-10% |
| Pinterest | Inspiration, long-term intent | 5% |
| Programmatic (Criteo, etc.) | Retargeting, performance | 10-15% |
| Emerging (Baidu, Naver) | Market-specific expansion | Variable |
The Team Required
Running predictive AI advertising at this level is not a one-person operation. The team we coordinated included 50+ professionals across multiple disciplines. For brands building this capability, the minimum viable team needs:
- •A data scientist or ML engineer who can build and maintain predictive models
- •A media buyer per major platform who understands luxury creative requirements
- •A creative strategist who can brief AI tools while maintaining brand integrity
- •An attribution analyst working with MMM or MTA tools
- •A feed/catalog specialist managing product data across platforms
NIMA Digital fills this gap for luxury brands through its curated network model, providing senior-level expertise without the overhead of building a 50-person in-house team.