Our current customer-listening capabilities are hampered by the poor quality of the signal. Rather like a submariner has to interpret the ocean activity around them through the pings and burps of Sonar, so we have had to intuit or guess our customers’ intentions from the crude information of cookies, visit, transactions and returns. However, the rise of voice interfaces in the home, the office and via mobile phones give us a theoretical rich stream in order to listen, analyse and learn about their needs, attitudes and behaviours.
Amazon’s Alexa, Apple’s Siri, and Google’s inventively-named Google are each featuring more highly in customers’ lives. Lights are dimmed, products are ordered, locations sought, travel booked, music is played and messages are sent. From this rich seam of conversational data – the questions, formulations, exchanges and variation over time – there is a growing list of data I’d love to be able to access. I’m imagining a nuclear family with voice-control points throughout the house, but focused upon the perennial situation of a teenager nagging a harassed parent for a new must-have ‘consumer non-durable’ product and I’ve listed some of the data I’d like an AI bot to be able to send me.
● Politeness Amplitude and Vector. Language-processing could assess the level of politeness used for the first-recorded request. Over the inevitable course of nagging or repetition we could establish the Politeness Vector, the directional movement of politeness or rudeness, along with the level of politeness at the point of purchase by the parent. This would be the optimal Politeness Conversion Level.
● Pack Effectiveness. This metric would look at how children in a family work together to persuade parents to buy. We would look at the number of voices, the conformance of the arguments, the regularity of ‘echoing’ arguments. At the point of purchase we would be able to establish the likely effectiveness of nagging based upon number, congruence and pack behaviour.
● Argument Triangulation. This metric would consider the range, variety and expression of arguments deployed. From a simple repetition of “I want it, I want it” (zero triangulation) to a variety of positions (I want it, I need it, my friends already have it, it will help my homework, I can pay half, you can get Gran to pay…), we can correlate the effectiveness of flexible positions, along with extracting key phrases to use in marketing and product descriptions.
● Nag-to-buy Periodicity. This tracks the duration between first nag and purchase, tracking too the frequency of the nags in between. The greater the number of nags the more they are seen as influential, whereas a long nag-free gap might indicate that the purchase delay is due to price, planning or pay cycles.
● Budget Power Factor. Voice recognition will allow us to see ‘inside’ the family – age, gender, of course, but also who typically approves a purchase. Whether it’s from a permissioning, budgeting or consensual position, each family member will be ranked for their role in the change from “idea” to “purchase”. Textual analysis will also allow us to see trigger-concepts and persuasion points in order to tailor marketing communications.
● Average Nag Value. This would reference all requests and purchasable items to their current cost and consider the “Monetary Nag Load” and how that changes over time, by individual and occasion. When linked to actual purchases we could calculate the ‘Capitulation Rate’ (being the proportion of mentioned products actually purchased).
● Propensity to Succumb. This is calculated from the arguments being deployed (see Argument Triangulation), the parental discussion, and the historical trigger-concepts, to create a probability that the budget-holder will approve the purchase. Based on this value the retailer can determine their promotional and discount strategy.
This information could be shared via a new markup language – a new cousin of XML and HTML, perhaps called “NagML”? Ideally, Amazon, Google and Apple could collaborate on a new standard for exchange, and thereafter we could further correlate with insights from social media.
With birthdays coming up I will have further occasion to consider the data opportunity from always-on listening devices matched with AI. If you have time why not drop me a note, at Ian@internetretailing.net, with your ideal data points and metrics and I’ll summarise later this summer – ready for the peak season onslaught!
This piece first appeared in the May 2018 edition of InternetRetailing magazine. To explore the issue further, click here.
Image courtesy of InternetRetailing Media