by Erin Lutenski, Head of Marketing, Decentriq
Canadian brands invested roughly $11 billion in advertising in 2024, with growth projected to continue through 2026. Much of that spend is going into artificial intelligence (AI)-powered personalization, loyalty program, and campaign optimization. Fewer than 30 per cent of AI leaders say their chief executives are satisfied with those returns, according to Gartner’s 2025 Hype Cycle for Artificial Intelligence. The data those AI systems run on is, in most Canadian marketing organizations, fragmented and cut off from the partner signals that would complete the picture.
The data silo problem Canadian marketers know well
In most marketing organizations, different systems hold different slices of customer data, and no single team ever sees the full picture. Loyalty platforms hold behavioural history, CRM (customer relationship management) systems hold contact and transaction records, media platforms hold campaign exposure, and ecommerce holds purchase signals. The customer who just bought in-store is invisible to the email team running a win-back campaign that same evening.
82 per cent of enterprises report that data silos actively disrupt their critical workflows, and 68 per cent of all enterprise data is never analyzed, according to IBM research. For Canadian brands, there’s an added layer: the Personal Information Protection and Electronic Documents Act (PIPEDA) governs how organizations collect, use and disclose personal information in commercial activity. Sharing customer data across systems or partners outside a customer’s original consent introduces both operational risk and compliance exposure.
Canadian marketers have growing volumes of first-party data but can’t activate it in ways that reflect who their customers are. Deloitte Digital and Meta research published in August 2025 found that 80 per cent of consumers are more likely to make a purchase when brands offer personalized experiences. Consistently falling short on that expectation comes down to data infrastructure.
AI raises the stakes
There’s a persistent belief in marketing circles that AI will eventually solve the data problem: that machine learning will synthesize disparate signals into a coherent picture regardless of the underlying infrastructure. There are a few reasons why this is unrealistic.
Fifty-seven per cent of organizations estimate their data is not AI-ready, according to the Gartner report. Gartner also predicts organizations will abandon 60 per cent of AI projects that lack AI-ready data before the end of 2026.
Neither figure is a prediction about model quality or compute costs. Instead, both are predictions about data foundations — foundations that are being outstripped by capability.
The reason is structural. AI models improve with data completeness, consistency, and continuity. When a Canadian retailer trains a personalization model on CRM data that excludes in-store transaction signals, the model’s recommendations reflect the data to which it has access instead of how the customer behaves.
The problem goes deeper than speed. Fragmented data shapes what AI learns. So for example, a loyalty model trained on siloed inputs learns siloed patterns: segments that miss crossover behaviour, attribution that misses offline conversion, and personalization that fails at the moments that matter most. Investing in more sophisticated models on top of fragmented foundations makes things worse, not better. The AI gets increasingly good at optimizing toward a picture of the customer that’s missing half the data.
The instinct is to treat this as an internal problem with an internal solution: consolidate the data, upgrade the warehouse, invest in a better CDP. That’s necessary, but it only addresses part of the problem. And for many Canadian brands, not even the most important part.
The case for data collaboration
Consolidation is the conventional response to data fragmentation: move everything into a data warehouse, a customer data platform (CDP), or a data lake. For many organizations, that’s necessary and worthwhile. But it addresses only half the problem.
Competing on personalization at scale requires data that no single organization entirely owns. A financial services brand needs purchase-signal data that only retailers hold; a retailer building a loyalty program needs media exposure data from publishers; an advertiser measuring cross-channel attribution needs platform signals that third parties control.
No amount of internal consolidation changes that. The data that would complete the picture sits on the other side of an organizational boundary. Collaboration infrastructure addresses what consolidation misses. The primary vehicle for that is the data clean room.
What is a data clean room?
A data clean room is a secure environment where two or more organizations can analyze combined datasets without either party exposing raw data.
Raw data stays with its owner; only the agreed results leave the environment. That’s what lets brands run joint audience analysis, build lookalike audiences, and measure cross-channel attribution, without the legal back-and-forth of conventional data sharing.
Why privacy-first infrastructure opens doors
A common objection is that privacy-first approaches limit what marketers can do with data. The opposite is true. Most cross-company data partnerships stall not for regulatory reasons but because of trust: no organization will hand over its customer data to a partner or platform without certainty about how it will be used.
Traditional data-sharing agreements rely on contractual controls and post-hoc auditing. Collaboration infrastructure enforces the rules before data moves anywhere: unapproved analyses cannot run, and neither party can access the other’s raw records.
For Canadian brands, this matters beyond strategy. PIPEDA sets the national standard for commercial data use, and Quebec’s Law 25 establishes stricter provincial requirements, with its most significant privacy obligations in force since September 2023. Both require organizations to handle personal data responsibly; Law 25 goes further by requiring demonstrable technical safeguards. Technology-enforced controls at the computation level meet that standard. Contracts alone cannot.
From one-off campaigns to always-on partnerships
Most data partnerships today are episodic. A brand and a publisher agree on a campaign-specific data match, generate results, close the collaboration, and watch the insight decay. By the time the next campaign planning cycle begins, the audience model is months out of date.
Always-on infrastructure changes that. Both parties establish governance up front: what analyses can run, how often the data refreshes, etc. The partnership then runs continuously without a new negotiation or legal review for each campaign cycle. Audience segments stay current, attribution models retrain on fresh cross-channel signals, and loyalty data enriches with partner transaction data in near real time.
When IKEA Austria combined its first-party CRM data with audience data from Austrian digital marketplace willhaben via Decentriq’s data clean room, the results were measurable within days. The mechanics are directly applicable to Canadian brand-publisher partnerships. The campaign delivered a 20-30 per cent reduction in cost per visit and a 10 per cent average increase in return on ad spend, per AIM Group, with no raw customer data exchanged.
A persistent, governed data partnership builds knowledge that compounds over time; a series of disconnected campaigns doesn’t. For Canadian brands, the opportunity is the same, scaled to local market dynamics: retailers and financial services firms sitting on complementary first-party datasets, publishers holding exposure signals that advertisers need, loyalty program enriched by partner transaction data. The partnerships are logical. What’s been missing is infrastructure that makes them safe enough to actually execute.
The infrastructure question Canadian leaders need to answer
For Canadian marketing leaders, the question is no longer whether to invest in AI-powered personalization. That decision has been made, and the spending is committed. What remains open is what data foundation that investment will run on.
Brands that keep treating data fragmentation as a technology problem they’ll get to later are compounding a debt that grows with every AI initiative they launch. Those that address the infrastructure first will close the gap between AI investment and AI value. That means a dual move: consolidating what they hold and accessing what partners hold. Both depend on infrastructure that meets PIPEDA and Law 25 requirements and earns partner trust.
Leaders who shift the question from “how do we fix our data warehouse?” to “which partner holds the data that, combined with ours, would change what we can learn about our customers?” will find the AI returns that most organizations are still waiting for. That’s a shift from internal IT projects to strategic data partnerships, which is where the unrealized value sits.
Erin Lutenski is Head of Marketing at Decentriq, where she leads content and product marketing across data collaboration and privacy-preserving technology. She has spent her career translating complex technical subjects into practical insight for marketing and data audiences.
