Digital Assistants a brand s best friend? By: David Wright, Kantar Large internet companies are putting an awful amount of effort into building or improving Digital Assistants (DA). Facebook launched M, Baidu released Duer, Amazon launched Echo and Microsoft launched Cortana. Meanwhile Apple's Siri now can be activated by saying "Hey Siri!" (No click required, they copied "Ok Google" sigh!), and has also been built into Apple TV. Apple has also recently acquired UK AI start-up VocalIQ, with the possible goal of making Siri more conversational. Google has taken a different strategy, on mobile, standard search is becoming more and more like interacting with a DA. Google has also avoided the often sassy personification offered by SIRI (ask SIRI what is zero divided by zero? ). Beyond, the giants, some amazing start-ups are emerging, such as x.ai, who have created an assistant Amy that will schedule meetings for you (no more calendar ping pong), democratizing PA s so everyone can have one not just the boss. This is the one I really want to use. X.ai s visionary founder Dennis R. Mortensen, describes his assistant as a vertical AI, it does one job well. M, SIRI, and Cortana are therefore horizontal AI s, as they do numerous jobs, but mostly in a question/answer format. DA's are the current AI battleground, and therefore it's a very busy and competitive space! Kantar Behavioural Data SIRI usage To get an idea of how people are using DA s in practise I found a group of more than 3000 panellists on our US mobile behavioural panel, who have used SIRI in the past year. Whilst we cannot measure, when a panellist has made an internal device call, like setting an alarm, we can evaluate searches that require SIRI to connect to one of its content partners (i.e. Bing for web searches, Wikipedia for factual information, Wolfram Alpha if calculations are required). We were able to capture, almost 70K of these SIRI connections, and classify their purpose. What we found was that 63% of the searches from SIRI resulted in a Bing search for information. The next largest category was Maps/Location at 23% of searches.
Siri Searches Itunes (find music) Maps/Location Bing Searches Wolfram Calculations For 10% of SIRI searches processed by Bing, a photo is also displayed, and I was able to capture and analyse these. Often photos are displayed if you search for something factual, particularly if that fact is resident on Wikipedia. The bonus being that Wikipedia s companion site Wikimedia, has a ready supply of Creative Commons licensed photos. History or Maths 3% Famous/Celeb 4% Flags 0% Siri Photos Corporate Logos 2% Music 1% Nature 6% Movies 18% Sports Logos 18% Sports Faces 48% For the photos I captured sports data figured prominently. The largest category was that of portrait photos of sportsmen. It is very likely this would have been for searches relating to player stats and information and that service has been heavily integrated into SIRI. Sports logos and Movie Posters were the next largest photo categories.
The nature category was also interesting, sharks, spiders and snakes featured prominently! However, as interesting as the photos are the only represented 10% of the Bing calls, and only give an indication of the things people search for, and the visuals they are used to experiencing.
Kantar Behavioural Data Google Mobile Search As stated above, it feels like Google Mobile search is becoming more like a DA. So, using our mobile behavioural panel, which we isolated searches that contained, How? (3% of search on our panel) and Near..? (1% of search). These are the sorts of searches that suit the rapid fire, useful and practical questions a person would ask a DA, when they are in a mobile moment of need. Near..? Searches containing Near..?, unsurprisingly related strongly to hotels and restaurants/shopping. Top keywords for Near..? searches
TV FIX MONEY WORK REMOVE CAR ANDROID PLAY PHONE FREE HAIR IPHONE COOK RESET APP XBOX CARD WATER CLEAN UNLOCK GALAXY CALORIES BABY DOG HOME INSTALL ONLINE BECOME GOOGLE FACEBOOK OPEN WORTH BOX GROW MUSIC OIL CONNECT MOVIES PAY WINDOWS How..? How questions are an interesting category and much promoted by Google s YouTube, re the creation of How to videos. Interestingly, TV was the most searched item, in relation to How..?. Particularly apparent were technical questions such as: How do I connect YouTube from Phone to TV?, Phone to TV connectivity was a significant trend in the data. How do i edit my contacts on my Samsung galaxy? What was striking was the number of technical questions relating to mobile devices that are being fed through search engines. It did make me wonder whether device manufacturers should be on-boarding this information into the device, so that they manage the customer relationship, as opposed to a search engine indexing tech support log sites. Top keywords for How..? searches Why is this important? At a simplistic level, accurate and speedy typing on a mobile screen is troublesome (and on a smartwatch impossible). The current vision is that a DA could become a central port of call, that seamlessly manages variety of tasks. For instance, "Book me tickets to Star Wars after 7 pm tonight, and arrange a taxi to get me there". The future vision, is that the DA will take your request, and in microseconds its AI will weigh the information it finds (from various API's), against your detailed profile information (and context - where you are, what time it is, what your interests are, what your personality type is, your mood etc..), and present only the most useful response. No more long lists of blue links, no more SEO or SEM (in its current form). Brands, need to start thinking about how this, rapidly approaching reality, can be leveraged to their advantage.
Usability Testing We recently completed a usability test, where a group of people were given a set of tasks that required the use of Siri or Google Voice: Location based questions - like where is the nearest Costa, or Aldi Supermarket; Information searches - like find me the best smartphone deals, or baked chicken recipes. Weather and time questions. The tapes revealed all sorts of interesting insight such as: "If I do a location search I want a map in the result (Siri didn't always do this)!" "I like it when Siri replies to me (via Voice), not when it gives me a list of tiny links" "I want it more conversational, like if SIRI asks me additional questions, to help it get me the best information" "I only use Siri at home, I'd be too embarrassed when I was shopping" "If I ask for the best restaurant near me, it gives all restaurants no matter how bad the review (which it also shows in the results). Just give me the ones with good reviews you have this information!" "If I ask for a baked chicken recipe, or how to remove a stain, why not link straight to a 'how to' video on YouTube?". "I need to adjust the way I talk, so Google understands me". In essence people want a DA to be conversational, to bring back the most useful response and to not send back links! What does this mean for brands? A first step could be at least creating a DA strategy. This could be based on your existing SEM knowledge, but might require you to reimagine the path to your brand via a customer talking to a DA. It might be that you need to ensure your data is available to the aggregation/comparison/review sites that the DA uses to find its information. Perhaps you need a content strategy (like the "How to" videos) that is relevant to your brand and the problems your customers might be trying to solve. A second step could be creating your own DA for your valued customers. Essentially an AI based CRM system, that allows you to be there for your customers 24/7. This could be via your app, website or your via Social Media (oh and please make sure you store all your Social Media/CRM conversations with your customers - a data goldmine!). I would follow Facebook M's lead here. They have a team of people answering the questions, but in a structured way, that lets AI routines, listen and emulate. So at the very least ensure your CRM data is stored in full, and get some advice on how it could be analyzed (with an aim to training an AI system).
The third step, is exploring the best ways to distribute your AI/DA. If you are Nike, and someone has a question about Nike shoes, should you be working with Apple or Google to ensure the queries will find your AI or the helpful content you have created? For large brands, it is definitely worth asking! Currently DA's are closed ecosystems, which can make it difficult for brands to be involved. But in future, it is likely that the will be opened up, allowing for third party apps. So instead of saying "Ok Google"... you might be saying "Ok Nike" if you have a specific request about running shoes, or fitness. It's about conversation It is interesting that Facebook's DA, M, is to be integrated into its messenger App, and that the conversational interface is text (not voice). Really it shouldn't matter how people want to communicate. It might be via video, or a photo. What is certain that conversation must be based on trust, and will need to reflect the values of your brand. Utilized well, I believe DA interfaces could in some cases reduce churn, and increase brand affinity.