Reducing Cognitive Stress In Retail

Digital retailers use data to study shoppers’ behaviour, predict their needs, and eliminate any friction points in the decision-making process. As a result, they provide shopping journeys that are very low on cognitive-stress. Now they are applying the same complex data models in their own physical stores. Different from what’s happening online; in physical retail current channels for in-store customer targeting lack relevancy and reach, diminishing profitability behind new industry entrants. 

The four main issues with current targeting models for physical stores are:

  1. Inefficiency: Store promotions and ads fail to reach more than 17% of their target audience.
  2. Expensive: CPG brands spend $178 Billion per annum in the US alone on in-store shopper marketing
  3. Passive: Promotions are planned long in advance, consider only a small subset of products and can’t be adjusted at run-time to meet live events.
  4. Outdated: In-store shoppers also shop online; they expect similar levels of customer experience and personalisation.

It is now more essential than ever for physical retailers to increase basket-value in-store by applying a real-time intelligent data model that consistently delivers unique in-store experiences, on-prem targeted campaigns, and highly personalised offerings. 

The reasons why it is essential are:

  1. Physical retailers can’t use competitor’s targeting data models (Amazon now is a competitor of Walmart and Tesco in the physical world).
  2. It is tough for each retailer to build a similar model alone because they will require massive alliances and partnership to share shoppers data and behaviour across the brand and store.
  3. It is impossible to succeed without it. Every day spent without physical retailers adopting an innovative targeting model, they are losing market share.

Intouch Mindful, as well as our 2 other products, (i.e., StoreHub and StoreAds), are operating on top of a unique technology that the founders of Intouch have developed using their knowledge and past experiences. We studied how personalisation is done in the online world and arrived at the conclusion that physical retailers require a different, revamped and reversed take on in-store targeted promotions. In brief, in the online world, retailers have control over the characteristics of the audience they are seeking to target with a specific product. It is easier to define the targeted audience in the online world because targeted promotions/ads can be executed across different websites, apps, and services.

However, physical retailers have no choice or say on who will walk into their store at any given moment. Physical retailer, in real-time, will have to find the right products to use for targeted ads to meet the personalised expectations of those who just walked into their store. This is why personalisation models that are making massive results online are not necessarily delivering on the same in physical retail. Online: retailers and brands find an audience for their products. Offline: retailers will have to find products for their real-time audience. While that sounds simple to make, the reality is that it is the most complicated task and require real innovation and a unique team to deliver on it – and we believe Intouch is in a unique position to do so.

Similar to how Google won the race to index the Internet, we are now racing to index the entire products range of physical retailing. Therefore, Intouch is currently developing an Artificial Neural Networks (ANNs) of pre-indexed and pre-profiled retail products. In this ANN, products are connected through each other, whereas each neuron is representing a unique product ID. The relationship between products is the link between two neurons. The strength of this relationship is weighted based on smart data mining techniques at the start, then re-evaluated based on in-store conversions. The model continuously gets trained from in-store conversions and learnings to ensure we deliver on better targeting efficiency.

Then we use image recognition models (a subfield of machine learning) to understand product images that are given to us by our customers in PIM files that are typically extracted from customer’s Product Information Management (PIM) system. We cross-reference the automatically generated tags of these images with the pre-indexed product tags our ANN contain. If a match is found, then this product’s pre-profiled attributes are pulled for usage; otherwise, the nonexistent product is flagged, and new attributes/profile is fetched.