Originally published on medium.com on 10/8/23.
Tapping into the anonymous user revenue opportunity
Most eCommerce personalization and recommendations are based on a user’s purchase history, site profile, and cookies. However, this misses out on a vast amount of the traffic coming from first-time and anonymous visitors. After all, without self-identification, how can you personalize?
Attempts to categorize anonymous users into broad cohorts based on location, traffic source, and other generic attributes provide some segmentation, but nowhere close to the personalization of known users. With minimal differentiation, anonymous shoppers are more likely to see irrelevant recommendations that fail to engage or convert.
Fingerprints enable granular personalization at an individual level to understand interests and provide hyper-relevant recommendations. Data reveals this segment represents massive untapped revenue potential:
- 96% of anonymous users get no marketing outreach- forgoing a huge growth opportunity.
- With just one channel targeting anonymous users, brands see purchase rates lift 5.3X for this high-value segment.
- Anonymous users are 58% more likely to buy within their first week compared to known users when engaged. Similarly they are 8.4X more likely to create accounts when messaged within their first month.
(86% of Retail Users Are Anonymous — What Does Your Anonymous User Base Look Like?, Braze, 2022)
The numbers paint a compelling picture. Anonymous users offer outsized revenue growth potential once engaged. Advanced fingerprinting provides the granular insight to craft personalized experiences and realize this potential.
This article explores how state-of-the-art fingerprinting and anti-spoofing techniques can help eCommerce merchants thrive in the new privacy-centric landscape and cater better to their anonymous users.
Robust fingerprint methods
Effective fingerprint services are challenging due to private browsing, multiple devices, browser updates etc. Advanced techniques are required for reliable fingerprints:
- Font Fingerprint analyzes installed fonts which combine uniquely to identify users. Robust across browser updates.
- Canvas Fingerprint draws shapes and extracts pixel data to generate hashes resistant to spoofing. Provides font-independent signal.
- WebRTC Methods can extract real device IP addresses which can then be correlated with other signals to improve accuracy
- Additional Methods — Advanced fingerprinting techniques leverage other browser and device APIs to generate signals that complement the above methods, including: AudioContext Fingerprint, Browser Plugin Enumeration, CSS Query-based Fingerprint, Battery Status Fingerprint, WebRTC Methods to extract true device IP address etc.
The key to effective fingerprint methods is generating reliable fingerprints resilient to private browsing modes, browser updates, mobile OS version changes and other dynamics. This requires going beyond simplistic techniques like logging IP addresses and User Agents.
Novel techniques are necessary for fingerprints robust enough to uniquely identify users over time.
Cat-and-mouse game of fingerprint spoofing
As businesses adopt advanced fingerprinting, users are also becoming aware of fingerprint tracking and take countermeasures:
Legitimate Privacy Seekers — Use VPNs, Tor and proxy services to mask IP addresses and geolocations. Leverage browser privacy modes and delete cookies/local storage to appear new. Install browser extensions that block access to fingerprinting APIs. Use privacy-focused browsers like Tor, Avast Private Browser , Brave, and DuckDuckGo that restrict APIs.
Fraudsters & Bots — Actively spoof fingerprint attributes like User Agent, Resolution, WebGL parameters. Use botnets and device emulators to mimic human fingerprints at scale. Inject scripts into pages to override fingerprint data returned to websites. Employ cutting-edge techniques like generative deep learning to dynamically spoof fingerprints.
For business analytics, eCommerce sites need advanced spoof detection techniques to exclude contaminated data.
Unmasking spoofing through novel techniques
Dedicated users and fraudsters are going beyond basic spoofing of individual attributes. This requires multi-layered spoof detection. Spoofing requires robust multi-layered detection such as:
- Behavioral Analysis — Monitoring user interaction patterns on site for signals like mouse movements, scroll velocity, click sequencing and typing rhythms. Correlating across visits, machine learning models can detect irregularities indicative of bots or humans attempting spoofing.
- Consistency Checks across browser, device, fonts and other data points. For example, if the User Agent indicates Chrome but the list of supported fonts aligns with Firefox, it suggests spoofing. Cross-referencing WebGL, Canvas, DOM and other attributes reveals discrepancies.
- Network Traffic Analysis — Inspecting network packet sizes, order, latency, and geolocation for alignment with fingerprints provides another layer of authentication. For example, mismatched region-specific IPs and languages indicate location spoofing. Unusual latency for claimed region also suggests spoofing.
- Fingerprint Stability Analysis — Naturally changing fingerprint attributes like battery level, GPU drivers, browser add-ons and font installs can be analyzed for unnatural changes that defy expected patterns. For instance, battery increasing from 20% to 80% within minutes of navigation signals potential spoofing. Temporal stability analysis identifies sudden anomalous changes.
- Supervised ML Classification — Large labeled fingerprint datasets can train classifiers like random forests to learn patterns of known spoofed fingerprints. The models can then reliably detect new spoofed fingerprints.
No one method is foolproof against sophisticated spoofing — combining multiple techniques provides high accuracy spoof detection.
PhotonIQ Fingerprint for browsers & devices
PhotonIQ Fingerprint, powered by AI and the Macrometa Global Data Network, handles the complexity of fingerprint generation, user identity tracking across sessions and spoof detection:
- Robust Fingerprint Engine — Generates device fingerprints on-the-fly using browser fonts, Canvas, WebGL and other advanced methods resistant to private browsing.
- Built-in Spoof Detection — Performs embedded behavioral, consistency, stability checks and ML classification for every fingerprint.
- Identity Mapping Across Devices — Updates fingerprints in real-time as users upgrade their browsers, go incognito mode etc. Enables persistent tracking.
- Real-time synchronization across devices utilizing Macrometa’s global edge network.
With PhotonIQ Fingerprint, eCommerce sites can focus on leveraging fingerprinting insights rather than building expertise in fast-evolving techniques.
Responsible fingerprint services: maintaining the privacy balance
Fingerprint services provide unprecedented visibility into previously anonymous visitors to understand their interests and serve hyper-relevant recommendations leading to more authentic engagements. However, businesses must ensure user privacy remains the top priority in leveraging fingerprinting data.
Maintaining transparency on how fingerprints are created and used, providing easy opt-outs, collecting only essential data, and focusing solely on enhancing user experiences are key.
With this privacy-centric PhotonIQ service, eCommerce businesses can tap into the potential of ethical fingerprint methods to delight every visitor while respecting their privacy.
Responsible use of fingerprints to create personalized experiences and fraud detection is a win-win for both businesses and users. As technology evolves, maintaining this balance through ethical practices will enable eCommerce sites to drive revenue while advancing consumer trust and privacy. Learn more today by chatting with an Enterprise Solution Architect.
