The rise of mobile and digital technologies has enabled consumers to search for any piece of information at the touch of their fingertips with near instant results. The speed and scale at which we can find information has not kept pace with our ability to process it, such that people are being overwhelmed with choices for nearly anything, like streaming content, news, products, services and experiences. In response, consumers have become shrewder, more discerning, and have ‘evolved’ shorter attention spans to filter out the noise that they’re continuously bombarded with. As a result, brands have had to explore tactics to personalise their interactions and propositions to engage with consumers on a one-to-one basis.
One approach to create greater consumer engagement is by combining customer data (e.g., demographic, psychographic, interest and behavioural) with techniques that enable ads, products and offers to be tailored to specific customers. An alternative approach is to map relationships between different products based on shared (e.g., two different brands of brown leather boots) or complementary characteristics (e.g., recommending a conditioner to complement a shampoo).
Recommendation systems are increasingly becoming an integral part of brand-user interactions and the consumer journey as the competition for buyer attention, time and money hits unprecedented levels. Whilst social media, entertainment, retail and e-commerce brands such as Facebook, YouTube, Netflix and Amazon may have pioneered this capability at scale, any business can benefit from using recommendation systems. By providing relevant and personalised suggestions for users, brands reduce friction and support users throughout their decision-making process which has a host of benefits from increased sales and revenue to trust and brand loyalty.
Challenges with manual product recommendation:
Why has artificial intelligence (AI) and machine learning (ML) become increasingly popular techniques for generating user recommendations? To understand that we need to have a look at what’s involved in manually making product recommendations.
Manual product recommendation can come with some difficulties, making the process complicated, confusing, and overall difficult to work with. While recommendations offer businesses great opportunities for growth and development, problems with manual recommendations can hinder this:
- One of the biggest problems with manual recommendations is the workload involved. To make a recommendation personalised, a wide variety of data is needed to determine whether a product is relevant to a particular user. This large manual workload can harm productivity and efficiency, as employees would need to dedicate large amounts of time and effort to do this. This can also affect a business's scalability, as a growing disparity between employees and users makes this service more difficult to offer.
- Manual recommendations can also be difficult in terms of quality. Without regular updates for each user, manual recommendations become static, inflexible, and based upon fixed rules. These static recommendations can become irrelevant over time as both users and products available change. Without considering new data for a recommendation, engagement is easily lost along with successful interactions and sales.
- Human error also plays a role in the problems of manual recommendations. Whether its bias in interpretation to simple oversight, the recommendations made by humans can be prone to errors. This can make recommendations less relevant, meaning that engagement (and by extension, successful sales) is lost, but could also cause offense, harming the public image of a business and impacting customer retention.
These issues can lead to some undesirable consequences. Without useful recommendations, the cost of search remains high: users must rely on their own research to discover and evaluate different propositions, potentially resulting in analysis paralysis, loss of interest, post-purchase dissonance and brand disconnect. Without a positive and personal experience, users may have less loyalty to a brand and have a greater likelihood of being swayed by a compelling competitor proposition. This increases customer churn, which impacts revenue, growth and profitability. This is crucial, as in a competitive market customer retention is key to growth.
By overcoming these challenges, businesses could see immediate results of increased customer satisfaction and user engagement. Furthermore, businesses would see long-term results such as higher customer retention and increased profits. With attention at a premium, better recommendations are critical to get ahead of the competition and deliver effective and high-quality services for users.
Different Approaches to Automated Recommendation Systems:
Recommendation systems work by analysing data on individual consumers, consumer segments and product relationships to automatically provide suggestions likely to be of interest to users. These data sets can be further enhanced with data on user interactions, such as whether a user purchased a recommended item or marked a film as ‘not interested’, which can be used as feedback to further refine the recommendations to user interests. Properly implemented, recommendation systems can be as useful – if not more – to a business than the customers that use it, by allowing businesses to leverage their data to create customer value, increase sales, revenue and drive growth.
The recommendations we receive as users are the result of different filtering techniques applied to data. This filtering can be done in several different ways and the specific technique will depend on the type of recommendation you aim to provide, user involvement in generating recommendations, and the type of data available:
- Collaborative filtering compares data of similar user’s preferences and generates recommendations based on what other similar users have found preferable. This is often seen as ‘customers like you also brought’ within online retail.
- Content-based filtering uses data that outlines similar attributes between content to offer recommendations, such as showcasing similar products when a user looks at a particular product.
- Demographic-based filtering categorises preferences based on user’s demographical data. This method can utilise traits within specific demographic classes and regions to offer recommendations that suit a vast majority of a particular group. This filtering is often used for advertisements related to a user’s local area.
- Utility-based filtering provides recommendations based on a calculation of the utility or ‘usefulness’ of the recommendation to each user. This utility can be quantified differently by the individual needs of the system to determine a user's preference for an item. This method can consider non-product attributes within the recommendation, such as product availability and reliability of vendor, allowing a user to get the best possible product for their needs.
- Knowledge-based filtering works by asking the user particular questions and generating recommendations using their responses as guidelines. This can be seen in real estate, where users input their preference of price, number of rooms, location etc to find relevant property suggestions.
Importantly, more than one method of filtering can be used at once and different techniques may perform better for different industries, offers or customers. A hybrid filtering system utilises two or more types of data filtering to tailor recommendations to a user.
The Business Benefits of Recommendation Systems:
Using recommendation systems can greatly benefit businesses that use them. These systems work to increase traffic and engage users, thereby increasing opportunities for successful transactions. Brands may also see greater customer satisfaction, leading to higher customer retention. This growing user base, along with the fact that order size and value increase as a result recommendation system, leads to increased profitability. These systems can also offer detailed market insight, helping businesses know the most popular and widely sought-after products to adjust similar items accordingly.
There are many ways that this technology can augment business. Entertainment platforms use recommendation systems to increase use-time on platform, driving up engagement. E-commerce stores utilise this technology to match users to products from mis-spelt or discontinued item searched, so that engagement is not lost, and sales can still be made.
Not only that, but using an AI solution within a business comes with particular benefits of these tools in general:
- Higher precision and quality: AI filters through large amount of data effortlessly, enabling recommendations to be based on larger amounts of data for more relevant and personalised offers.
- Improved efficiency: AI can work faster than humans at filtering through data with a lower margin of error, ensuring that these recommendations are available when needed.
- Technology retention: These tools can react dynamically to new data input for finer, more detailed decisions. The longer an algorithm is used and the more data it is exposed to, the greater performance in generating personalised recommendations.
- Scalability: AI systems have a high adaptability to growth and increased demand. This allows for businesses to grow with a larger user base without compromising on quality of service.
Successful use-cases
Recommendation systems have many use-cases that boast their success:
- Netflix credits 75% of watched content to its recommendation system, skyrocketing user engagement with the platform. They also reported the system has helped reduce customer churn by ‘several percentage points’, estimating that this saves the company a value of over $1 billion every year.
- YouTube reported that 70% of videos watched from the home screen are users’ personalised recommendations rather than simple recommendations.
- Google’s news engine reported that personalised recommendation led to a 38% CTR increase compared to simple popular item recommendation on their platform.
- eBay saw an 89% increase of ‘add to Wishlist’ actions after introducing a similar items recommendation list on users' page after lost auctions.
- Nike’s Training Club (NTC) mobile training app saw a 62% engagement increase after switching to a collaborative filtering model to recommend exercise plans over simple recommendations.
- More broadly, a 2018 study showed that 93% of businesses with an advanced personalisation strategy experienced increased revenue that year.
Summary:
There is no doubt that recommendation systems have helped many businesses drive productivity and growth, by increasing customer satisfaction - leading to higher engagement and loyalty. This benefits businesses immensely, both within the public image and the profitability of the service offered. Whatever the industry, a strong personalisation strategy can help any business elevate their platform and offer something great to its users. Personalisation is a rewarding long-term strategy for brands that is focused on creating an enriching experience and value for the user, which is then reciprocated in turn. Brands have the incredible chance to get ahead of the competition by overcoming pain-points with this technology. There will be winners and losers in competition, and with recommendation systems enhancing user business interaction, it is the businesses that invest in technology for their users benefit that will see the benefit back on themselves.
Written by Joseph Myler and Clayton Black