Who chooses what you watch: you or Netflix? A look at recommendation engines

Recommendation engines have become an everyday tool for brands to enhance user experiences and drive customer engagement. Amazon, LinkedIn, and Spotify all leverage user preferences and patterns to harness the power to suggest relevant items, products, and content to users.

But how exactly do they use these engines to inform consumer decision-making? And to what extent can an algorithm go on to deliver a rotten tomato?

Personalization at its best: collaborative filtering

If the resulting consumer experience is a ‘personalized approach’, then it’s likely a brand has incorporated collaborative filtering in some form. This technique relies on examining attributes and characteristics of items or content to match them with user preferences.

Although Amazon’s landing page is fairly aesthetically uninspiring (a pointed choice given the rates of conversion), it excels at serving up products of genuine interest. It considers factors including genre, author, director or product attributes, providing tailored recommendations that suit individual tastes.

Yehuda Koren, one of Google’s top research scientists, puts the principle behind recommendation systems thus: “The most successful recommendation systems are those that understand users and their preferences better than users understand themselves.”

Filtering (when done well) will feel seamless and should add genuine delight to a consumer journey. ‘Customers also bought’ is a pertinent example of how filtering has become an almost universal feature across sectors, nudging shoppers to either a ‘treat yourself’ moment or further nod to explore the category, with a driver to click on items that previously might not have been considered.

This can be a big win with ‘nicher’ searches, encouraging users to discover lesser-known items or unique content that otherwise might never get seen.

Like all algorithms, filtering too has its frustrating limitations. It often lacks serendipity and imagination and users can get trapped in a ‘filter bubble’, with overexposure to similar recommendations and a lack of varied options. And feed in ‘too niche’ an interest, and the data may not exist to serve up recommendations that really resonate.

For ‘surprise and delight’ nods, collaborative filtering does the thinking for unassuming users. This approach provides recommendations based on by identifying users with similar tastes and preferences, and serving up content enjoyed by others who share similar interests.

Netflix and Spotify reign supreme in this field, analyzing data based on history, ratings and interaction patterns to create personalized offerings. This is rewarding for users when they are served something unexpected, but problematic where there’s a lack of data, or where popularity bias can sway users from discovering lesser-known gems.

Advanced search capabilities and consumer purchasing decisions

We find the real sweet spot when brands use advanced search capabilities in combination with AI and a natural tone of voice. Here, engines understand queries and deliver accurate results that feel on-brand, offering allyship to users. Less than a decade ago, without the specific name of a product, users were not able to find their search on Google. Fast-forward to 2023, and e-commerce sites have followed in the footsteps of the search giants by offering features like autocorrect, non-branded search, and synonym search to optimize the chance of a sale.

Take Ebay, for example, which has embraced advanced search, improving the experience for users while delighting them with new tools. How? By creating personalized recommendations based on past searches and behaviors via Google and Facebook, and shaping search results for new shoppers by prioritizing items available to purchase now (rather than bid). For existing users, the platform has brought in alerts and updates on an item’s availability they’re likely to want to bid on, as well as a ‘buy again’ option.

At True, we’ve been integrating clients’ sites with Algolia to improve the effectiveness of product search results. A recent example is our client, the brewery St Austell. Collaborative search has helped create more personalized product category pages on the wholesale site, based on customers’ behavior and interactions. This will be optimized shortly to a live (as-you-type) search, alongside using harvested user behavioral data to create more personalized promotional and product recommendations.

As technology continues to advance, recommendation engines and advanced search capabilities will play an increasingly vital role in shaping consumer preferences and driving business success. By understanding user queries and preferences, search engines can deliver real value, permitting consumers to make better-informed purchasing decisions. For brands, this means a more effective experience that fosters satisfaction and loyalty down the line.