Choosing a video analytics partner hinges on four key factors: the solution’s ability to integrate with your existing camera infrastructure, its GDPR compliance, its scalability across multiple sites, and the actual operational value measured by your teams. Before shortlisting technology partners, these 9 questions will help you make an objective comparison rather than relying solely on sales pitches.
| # | Question | Evaluation Criteria |
|---|---|---|
| 1 | What business use case does the solution actually cover? | Business Fit |
| 2 | Does it integrate with my existing camera fleet and systems? | Integration |
| 3 | How reliable are the AI models on my own video streams? | Integration |
| 4 | How does the solution ensure GDPR compliance? | Compliance |
| 5 | Are the algorithms based on biometric data? | Compliance |
| 6 | Can the solution be deployed across hundreds of cameras? | Scalability |
| 7 | What support is provided during and after deployment? | Scalability |
| 8 | What operational value do existing customers see? | Business value |
| 9 | What is the business model and the expected return? | Business value |
The first mistake in a video analytics project is choosing a technology before defining the business need. A good partner doesn’t start by asking, “What AI are we going to buy?” but rather, “What operational problem do we want to solve?”—such as reducing wait times, measuring foot traffic, understanding the customer journey, or streamlining logistics. Ask your partner to link each proposed feature to a measurable business objective—not just to a technical detection capability.
Replacing an entire camera fleet is rarely realistic, either budget-wise or operationally. A video analytics solution must work with the infrastructure already in place: RTSP streams, ONVIF protocols, existing VMS, and monitoring systems. Compatibility with sensors already in use allows you to leverage an existing investment rather than requiring a new one. In this regard, integration often plays a more significant role in a project’s success than the sophistication of the algorithm itself.
Not all computer vision models are created equal, and a demonstration using a generic use case offers no guarantee of performance in your actual environment. Ask specific questions: What is the accuracy rate? How are false positives measured? Were the models trained under conditions comparable to yours (camera angles, lighting, traffic density)? Always insist on testing the solution on your own video streams before committing to it.
Video streams may contain personal data, which makes GDPR compliance a key factor in selecting a partner—especially in retail and transportation, where the public is directly filmed. A reputable partner must be able to present a clear data governance policy, principles of “privacy by design” and “privacy by default,” and ideally a data protection impact assessment (DPIA) already conducted on deployments comparable to yours.
This is a game-changer, both legally and operationally. Some solutions rely on biometric or facial recognition approaches, which significantly increase compliance challenges. Others favor non-biometric methods, where a person is treated as a simple, anonymized silhouette, without individual identification or tracking. This approach makes it possible, for example, to automatically exclude staff from a foot traffic count based on distinctive features (badge, uniform) without ever resorting to facial recognition.
A proof of concept on three cameras and a deployment across several hundred sites are two entirely different challenges. This is where partners who are truly capable of supporting a project’s growth stand out. Nhood, for example, uses XXII’s CORE platform across more than 300 cameras:
“CORE provides us with key insights in real time—visitor paths, hotspots, and wait times—across more than 300 cameras.” — Simon Chopin, Chief Data Officer, Nhood
Always ask how many sites and cameras the partner already has in production, and how large-scale updates are managed.
A video analytics project doesn’t end with installation. The quality of project management—configuration, team training, support, and evolving use cases—determines the solution’s actual adoption in the field. A reliable partner must be able to detail its deployment process, typical timelines, and level of support after the system goes live, rather than stopping at the sales phase.
Beyond the features, the real question is: what decisions has a company been able to make thanks to this data? At Decathlon, video analytics has made it possible to rethink the allocation of in-store teams:
“CORE allows us to optimize schedules and refocus our teams on customer service.” Director of Innovation at a major sports retailer
Ask for client references in your specific industry and quantifiable results—not just theoretical use cases.
Video analytics solutions are generally structured around licenses per stream, per location, or per feature, with annual or multi-year subscription models. The right benchmark isn’t the license price alone, but the expected return: reduced wait times, better team allocation, and increased conversion rates. Have this return quantified based on a pilot project before committing to a large-scale deployment.
A good video analytics partner is judged by its ability to integrate with existing systems, its rigor regarding GDPR compliance and anonymization, its proven ability to deploy at scale, and the concrete operational value its customers derive from the solution. These 9 questions will help you build an objective shortlist tailored to the real challenges facing the retail and logistics industries.