Recommendations For The Next Generation Of Ecommerce

Tugdual Grall, Technical Evangelist, MapR Technologies discusses data-driven recommendation engines in retail


A positive customer experience is not something that should be reserved only for the shop floor. With online channels becoming increasingly important, retailers are facing the challenge of differentiating themselves in an already saturated market.

As a result, many forward thinking companies are looking at how implementing more personalised and innovative technology will not only attract customers but also retain their loyalty. 86% of web-savvy shoppers who’ve experienced personalisation technology believe it had impacted their purchasing decisions, according to the Infosys Rethinking Retail Survey.

But it’s not just ‘one or the other’ with customers. 69% of online shoppers feel that consistency across both online and offline stores is of great importance. So, retailers need to consider how they can harness data and recommendation systems to offer a more seamless integrated experience.

While capturing and using data is important to retailers, it’s also important to consider the consumer thought process. Pew Research recently conducted a survey which showed consumers worry considerably about the security of their information online. More than 90% reported feeling like they’d lost control over how their personal information is being used online.

So whilst recommendation engines can help shoppers find what they want, it is essential that sufficient consideration is given to how these systems are implemented.

The mechanics behind the engine

With personalisation technology is increasingly adopted by big online brands, few of us haven’t experienced it. For example, LinkedIn’s “People You May Know” feature may help you find a former colleague, or you might listen to music recommended by Spotify, or shop for items off Amazon’s "Frequently Bought Together" section.

All these examples demonstrate the core value of recommendation engines – providing potential options that best meet an individual’s specific needs. By incorporating algorithms and machine learning from past customer data, companies can develop recommendation systems that can make precise predictions about user preferences.

From financial services to media, many industries drive at least a portion of their business through a recommendation engine. But it is in the retail industry, in the context of an online retail or ecommerce operation, that recommendation engines are arguably most effective.

A more tailored and personalised shopping experience increases the probability of a retail costumer returning – driving future sales and building loyalty to the brand in question. For the merchant, the information generated from the recommendation engine can help increase upsell and cross-sell rates, whilst also reducing churn.

With the enormous quantity of data collected from each online transaction – from clickstream, and mobile data, to past transactions and behavioural data – retail merchants hold the key to revolutionising the industry. Leveraging new big data technologies, such as Hadoop, enables businesses to mine in the volume and variety of this data to discover patterns and outliers to optimise their sales.

Retailers can use this analysis to predict what the best product or service that they should offer an individual customer. Such offers need not solely be based on what products that specific customer has bought, but also what other similar customers have purchased —providing “next best offer” opportunities.

Ready. Set. Recommend.

Machine learning is transcending the research arena into the world of business, allowing companies to harness this technology through recommendation engines. With its increased accessibility, retailers can now easy begin incorporating it into their business strategy.

When getting started with a recommendation engine, retailers should consider the following steps:

  1. Digesting item meta-data: Important for identifying items or products for recommendation.
  2. Digesting log files containing user history behaviour.
  3. Analysing users’ behaviour to create new meta-data which can be funneled back into the search engine to support future recommendations.
  4. Enabling users to engage with a search index as it gets populated with item meta-data and user behaviour meta-data. This then generates new user history which feeds back into the system as part of a closed loop mechanism, improving future recommendations.
  5. Boost the system’s accuracy by finding alternative behavioural data that produces better recommendations.

This approach not only provides the benefits of recommendations and personalisation, but does so without exposing private consumer information.

Ecommerce 2.0

It is estimated that online shoppers will spend a whopping £52.25 billion this year, up from £44.97 billion in 2014. This translates into a great opportunity for retailers to serve up their own portion of sales growth.

We’re entering the next generation of ecommerce, and retailers are looking to sophisticated methods —like recommendation engines—to capitalise on the significant volumes of market and customer preference data.

And as consumers, we can also enjoy the benefits of these sophisticated recommendation engines to tailor and optimise our own experience. How we engage with online resources is becoming more refined and relevant: whether it’s a simple “to read more” tag, or Google Maps adjusting landmarks based on your request, such as, showing other cafés in the same area as the coffee shop you searched for.

Technology is promising an exciting future for the retail industry with a seamless interaction between machine and shopper, guiding customers through their purchasing journey. It won’t just be the volume and quality of data that will determine which retailers achieve this, but also the infrastructure they deploy to leverage it. Businesses have the opportunity to not only increase product discovery and sales, but also build long-lasting relationships with their customers through creating each one’s own individual story with that brand.


Read next:

Working At The Boundaries Of Aesthetics And Inference