Have you ever wondered why many companies opted to meet their paper requirements through Dunder Mifflin Paper Company even though it was a small-scale company and offered more expensive products? It was because of the personal touch that Micheal, Dwight, Jim (Jim? James? Jimothy?) and the rest of the salespersons gave to their customers. With digitalisation and the rise of e-commerce, somewhere we all miss the personal recommendations shopkeepers used to provide us with, saying “Aap ko yeh zyada pasand aayega”. The key to making online services and shopping ‘feel like home’ (while being at home) is using Big Data in various capacities. By recording each click in the database, big data comprises of the smallest of the move made by a customer when surfing the website. This information is then analysed to find patterns in consumer behaviour and subsequently individualistic suggestions are made to better the user experience and increase revenue.
Various multi-billion-dollar companies and unicorns utilise Big Data in different ways to stay at the top of their game. Goods based sites such as Amazon use the information on what items a consumer views, adds to cart, purchases, gives reviews/ comments etc. to personalise recommendations of different goods. Intuitively, the main motive behind this is for them to get consumers to buy more goods at a go, or in technical terms, to increase Average order value (AOV). On the other side of the spectrum are companies that provide online services or softwares, Netflix being one of them. It effectively uses the taste and consumption habits of individuals to suggest different titles and thumbnails in order to increase viewership. Let us expand more on how Netflix uses it to its advantage.
Even though Stuart from TBBT was a very eccentric character, one of his characteristics stood out from the rest - his vast knowledge of comic books. Remember when he recommended Penny, Amy and Bernadette the comic books to get started with and they were hooked on to them, discussing and debating on it for hours? Netflix was started with the same motive, an on-demand delivery service with personalised movie recommendations. Now that they have gone all virtual, their revenue stream is solely dependent on subscribers paying for their subscription plans, so the company must keep its viewers hooked to the vast content provided on their site.
Apart from the very interactive and relatable social media posts that we comment and often share, the thumbnail of a movie, TV series or documentary plays a significant role in determining if a user chooses to click and start watching the show. This follows from that fact that humans are incredibly visual creatures, and hence have a bias towards visually appealing content. The company estimates that if you don’t find anything exciting to watch a total of 60-90 seconds, you would close the app and resort to other activities. Now, this would have been easier if the viewers knew exactly what they wanted to watch - just type the title in the search bar and start viewing. But, 80% of the viewership comes from the personalised recommendation engine rather than the search engine. This entails the fact that the company needs the best possible technology to put forth the thumbnails of content as accurately for each individual as it gets.
Aesthetic Visual Analysis or AVA is the algorithm that helps the company find the best thumbnails. The process searches for the best images that could be a potential thumbnail from the millions of frames present in hours long of video content. Just like a fingerprint, metadata captures the different characteristics of each frame. These frames are then analysed on three parameters - visual, contextual and compositional.
● The Visual parameter deals with the contrast, colour and brightness. For instance, consider the scenes that portrayed the time period when Hannah Baker was alive - they were bright and sunny, showcasing a happy and joyful moment for Clay. Whereas, the activities that took place post Hannah’s death had a gloomy blue filter.
● The Contextual criteria accounts for object detection, motion and shot angles; in the case of Stranger Things, most of us would like to see Eleven, Will or Mike on the thumbnail, rather than a side character such as Bob or Mrs. Wheeler.
● The Compositional matter of the frame deals with visual principles in design, cinematography and photography; an extremely zoomed-in or zoomed-out picture may not be very effective in communicating the theme of the show.
Henceforth, A/B testing is executed on the thumbnails that pass this check. Very simply put, in A/B testing users like you and me are shown the different covers of the same content, and the covers that have the most interactions are then considered the best for specific target groups. Along with this, based on the individual’s engagement with previous titles, the thumbnails would keep changing. A classic example would be the Riverdale thumbnail being dim and dark, featuring Jughead if the user is a fan of thriller while a lively and bright one featuring Archie, Betty and Veronica if they are into romance.
This was just the part where the company picks out the right thumbnails. Adding on to this, Netflix directs you to titles according to what the people around are watching. Does this indicate that your friend who stays two buildings away and prefers watching horror films would get the same recommendations as you, who is more into the stand-up comedy specials? No. This is only a way of measuring the country’s culture and local taste. Layering on top of this, an individual would be receiving suggestions according to the genre they prefer. In other words, the recommendations that you, a resident of Mumbai would get would be in line to what the users in South Asia play along with other stand up comedy viewers from across the world. All this is just what one company is doing to cater to its audience. The company spends a good amount of money and effort on updating its recommendation software, the algorithms that would learn your taste and suggest the perfect title from the company’s catalogue. But could this also mean that though strong analysis and algorithms, one day the search bar would be gone for the good?
With such extensive data testing and analytics done, Netflix has valued its engines at a whopping $1 billion.
We’ve all heard the phrase ‘Customer is King’. Having a loyal customer base is one of the top objectives of each and every business, not only a multi-national company but even the Kirana store right next to your house. Alongside trust, bettering the user experience is another very important factor in a world that’s being taken over by the internet, digital media and e-commerce. Data is certainly the way forward for businesses. Unicorns, start-ups and now even local businesses are getting in the game of making their online customer experiences more personalised than generic. Now that this seems to be the new normal, the businesses not using it need to buck up their game!
Anvi Agarwal
References :
4. https://medium.com/swlh/the-importance-of-predictive-analytics-for-e-commerce-stores-d7ef0ce2d32e
Great Post. Nicely written.
SO COOL !! very informative