Executive Summary: Key Legal and Evidentiary Issues

• Appellants challenged a certification judge's refusal to certify a multi-jurisdictional privacy class action against Cadillac Fairview and related mall operators.
   
• Central to the dispute was whether cameras installed in mall wayfinding directories converted visitors' facial images into biometric or personal information.
   
• Both parties contested the admissibility and weight of expert evidence, including reports from a computer programmer, a privacy expert, and a computing science professor.
   
• Quebec-specific claims raised questions about whether "compensable damage" needed to be pleaded to certify claims under the Civil Code of Québec and the Québec Charter.
   
• Judges considered whether the identifiable class and common issues criteria under the Class Proceedings Act had been satisfied.
   
• Ultimately, the appellants' decision not to appeal the certification judge's finding on preferable procedure proved determinative, since all five statutory certification criteria must be met conjunctively.
   

Facts of the case

Joshua Cleaver and Curtis Kieres appealed a decision of a British Columbia Supreme Court certification judge who declined to certify their proposed multi-jurisdictional class action against The Cadillac Fairview Corporation Limited and several related corporate respondents. The underlying claim arose from a pilot project Cadillac Fairview ran between May 30 and July 24, 2018, using cameras installed in mall wayfinding directories at shopping centres in British Columbia, Alberta, Manitoba, Ontario, and Québec. The cameras used video analytics software supplied by MappedIn Inc. to estimate visitor age and gender. Each detected face was converted into a string of 128 numbers, generating five million records containing time, location, camera ID, estimated age, gender probabilities, and an embedding number [the judgment separately refers to "over five million records" when discussing the full dataset]. The project ended after media reports suggested facial recognition technology was in use. Privacy commissioners from Canada, Alberta, and BC subsequently investigated and released a report on October 28, 2020, concluding the software's operation amounted to collection of personal information under applicable privacy statutes, though no further regulatory action was taken because Cadillac Fairview had already disabled the software and deleted the data. The appellants filed their underlying claim on November 19, 2020, alleging intrusion upon seclusion, statutory privacy breaches, negligence, and various breaches of Québec law. A three-day certification hearing followed in September 2024.

Policy and legislative provisions at issue

The certification application turned on section 4(1) of the Class Proceedings Act, R.S.B.C. 1996, c. 50, which sets out five mandatory requirements for certifying a class proceeding, including an identifiable class of two or more persons, common issues among class members, and a class proceeding being the preferable procedure for resolving those issues. Because these criteria operate conjunctively, all five must be satisfied before certification can be granted. For the Québec-based claims, the appellants invoked provisions of the Civil Code of Québec, C.Q.L.R., c. C.C.Q.-1991, and the Québec Charter of Human Rights and Freedoms, C.Q.L.R. c. C-12, arguing the certification judge wrongly required proof of "compensable damage" rather than a simple injury, and misapplied the standard for punitive damages under the Québec Charter.

Reasoning and analysis

Justice Mayer, writing for the court, found the certification judge had correctly admitted the Office of the Privacy Commissioner report, though only for the fact of its existence rather than the truth of its conclusions. The judge below had also properly admitted expert evidence from Jason Frankovitz and John Wunderlich while excluding speculative portions, and had appropriately considered the responding expert evidence from Dr. Richard Zhang, who concluded it was not possible to identify individuals from the embedding numbers generated by the software. On the merits, the Court of Appeal agreed with the certification judge that the appellants had failed to establish any basis in fact that facial images or biometric information had actually been recorded and used to identify individuals — a finding fatal to the identifiable class and common issues criteria. Justice Mayer further held that even if the appellants had succeeded on those points, the appeal could not succeed because they had not challenged the certification judge's separate finding that a class proceeding was not the preferable procedure for resolving the dispute. Since section 4(1) of the CPA requires all five certification criteria to be met, the unchallenged finding on preferable procedure was, on its own, sufficient to defeat certification regardless of the outcome on the other grounds of appeal.

Ruling and overall outcome

The Court of Appeal for British Columbia dismissed the appeal. The respondents were the successful party. The court did not identify any monetary award, costs order, or damages figure in its reasons; the decision addressed only the certification question and did not quantify any amount owed by either side.