New paper: Unpacking the Personalisation-Privacy Paradox in the Context of AI-Enabled Personalisation in the Physical Retail Environment

Once the stuff of fiction, it is now possible for you to enter a store and receive an offer which was personalised to your preferences, past purchases and even what’s happening around you. Unlike John Aderton in Minority Report, though, you don’t need to scan your eyes as you enter the store, in order to get that great offer. Rather, a beacon detects the presence of your phone in or around the store, communicates with various databases and then, delivers a personalised offer to your mobile phone. 

While such hyper-personalisation is already common in the online environment, due to the power of tracking technology and machine learning, there is no guarantee that customers will welcome it in the physical world. Indeed, research shows that attitudes towards personalisation vary significantly with the context in which it takes place; and there are various characteristics of the physical retail environment which could greatly impact on how customers experience and react to hyper-personalised offers. For instance:

• Consumers’ motivations vary for in-store vs online retail;
• The phone where the offer is delivered has a small screen with, possibly, a low-quality interface; and
• It is unclear whether privacy concerns would be more salient in the physical environment than the online one.

These differences mean that we can’t simply transpose to the physical retail environment what we know, already, about customer attitudes towards online personalisation. That’s why Ben Keegan, Maria Ryzhikh and I set out to investigate how physical retail customers experience and perceive this type of offers (we call them AI-enabled personalisation; AI-EP for short).

We set out to identify:

• What customers like vs dislike about AI-EP personalisation
• Whether they perceive any privacy threats (and, if so, how they deal with them)
• As a result of the above, when they are receptive to AI-EP

We conducted a case study analysis of a fashion retail app, in London, which included in-depth interviews with 18 female shoppers aged 18 to 30 years old. We chose this demographic group because, according to YouGov, they are twice as likely than men to agree that they spend a lot on clothes and to value immediate access to fashion items; and they are more likely than men and older women to shop at multiple retailers. They are also a key target for high street fashion retailers’ promotional efforts.

We found that firms that want to use AI-EP need to negotiate a number of opportunities and challenges – much like playing a game of snakes and ladders – if their investment in AI-EP is not to be wasted or, even worse, generate a consumer backlash.

We found that our respondents mostly want discounts on desired items, and they hope that the enhanced data collection will lead to improvement of the in-store experience. 

Unlike what happens in the online environment, shoppers don’t really look for inspiration from AI-EP. This is possibly because, overall, they see the AI as lacking the intuition and subjectivity needed to make good fashion recommendations. 

Participants resent interruptions from notifications because they feel that notifications deplete their phone’s battery and intrude on a leisurely activity. They also dislike generic or irrelevant offers, indicating low tolerance for trial and error.

Our respondents were generally happy to share functional data (e.g., clothes size, style preferences…). However, opinions were very divided regarding the use of social media and location data: While some didn’t mind, others displayed a visceral opposition to firms tracking and using information from their social media accounts as well as their location. Location data is essential for offer contextualisation. Without this information, firms lose a valuable opportunity to personalise offers in response to environmental factors such as the weather or crowd levels; as well as store specific ones, such as store lay-out or stock availability.

Overall, participants expressed a strong desire to control information held in the system and used for personalization; and to edit / delete information that they thought was outdated or irrelevant (for instance, related to gifts they had bought). Some participants also expected that the firm should be able to explain why they had received a particular recommendation (which would be very difficult for firms to do, given the opacity of machine learning algorithms).

These findings lead us to conclude that:

• Attempts to use AI-EP for customer acquisition are likely to be ineffective or even detrimental.
• The opportunity of AI-EP seems to be, mostly, in customer retention and, particularly, at the point of purchase (unlike online personalisation, which works well in the pre-purchase stage), in the form of enticing discounts. 
• Customers have very high expectations of AI-EP, as a result of their experiences with online personalisation. Though, it may be challenging for firms to deliver on that promise, because of the technical restrictions of in-store personalisation and of customers’ discomfort with location tracking. 

The findings from this study have just been published in the journal Information Systems Frontiers, The paper is entiled: “Snakes and Ladders: unpacking the personalisation-privacy paradox in the context of AI-Enabled Personalisation in the Physical Retail Environment”.

I really enjoyed working on this project with Maria and Ben. We want to take this further by researching:

• AI-EP in other contexts – Research in other empirical settings is needed before claims can be made about consumer perceptions and experiences of AI-EP, generally. 
• Perceptions of AI-EP benefits – It is valuable to identify which messages most clearly communicate the desired benefits valued by different types of customers and/or in different contexts.
• Impact of the various factors identified on AI-EP acceptance – Our qualitative approach enabled us to identify the important, relevant factors. However, we can not use it to quantify how the different factors identified in this study combine to amplify – or not – purchase intention when exposed to AI-EP.
• The digital-physical customer experience dynamic – One distinctive feature of AI-EP is that it takes place in-store. It would be valuable to understand how AI-EP could augment the access to onsite staff.

Let me know what you think about our paper. And if you want to facilitate the follow up studies, do not hesitate to leave a comment below, or get in contact with me.