Artificial Intelligence in Market Research

Good morning and welcome to our Smarties webinar on artificial intelligence in market research, I will share with you today two case studies of what we do and insights consulting with artificial intelligence, and I will also give a little introduction on the status of artificial intelligence. If you go to a conference nowadays can be a marketing conference, a business conference or a research conference. You hear a lot about deep learning. Machine learning, pattern, recognition, automation, artificial intelligence. All these words are really

Buzzing today – and you could ask yourself the question: is this: the next hype lets do a little investigation. There are two characteristics of a hype. The first one is that we using different words to say the same or the same words to say different things is this. Also happening in this fields, looking at all the words that you saw on the second slide, is this also the case? I think so, because all the words that you saw could be placed under the umbrella

Term artificial intelligence, so all those different words almost linked to the bigger concept of artificial intelligence. They are all a part of what we could call. Ai and artificial intelligence is actually the broader concept of machines being able to carry out tasks in a way that we would consider smart. So where are we today with AI? Well, actually, we could say that there are three types of artificial intelligence. First of all, we have narrow AI. Secondly, we have general AI and last but not least, we have super AI

Narrow AI means that a machine or a smart system can do one specific task very well. For example, the people at google google deepmind who were playing a little game of go against a human super champion. Thats narrow AI. The machine is able to understand, go as a game very very well. Another example of narrow AI is Amy Ingram Amy Ingram is a personal assistant developed by X, dot, AI, a start-up and Amy is able to book your meetings, so she will have a dialogue with the people that want to book a

Meeting with you, she will look into your calendar and she will arrange the meeting she will, even if needed, order some food or a good coffee for during the meeting, so that’s narrow, AI. The smart machine helps us and is doing one specific task very well. For us and within narrow AI, you see systems that are pure automation, its a very rule-based, but you also see systems that are a little bit more advanced and then our self learning Amy, for example, is

very much rule-based she knows your preferences and based on your preferences she will book the meeting for you the game of go by googles deepmind department that’s very much self learning the system has taught itself how to become lets say a world champion in playing go so that’s narrow AI then we have generally I what is that well its a machine who can do different tasks very well and the machine also makes itself smarter so the.

machine becomes better at doing all those different tasks over time where are we in terms of general AI well I think we only see the first signs of this despite what you might read in the news and especially on the technology websites 99% of the AI systems today are narrow AI so generally I we are just seeing the first signs of that and then the third level is super AI that’s when we reach what they call singularity that’s when machines become smarter than.

humans machines even become smarter than all humans on the planet you dont need to fear super AI today because scientists predict that it will take us at least two decades before super AI will be there and before the technology will have advanced in that direction so what you need to remember is that today most of the AI systems 99 percent of the AI systems are narrow AI they are built and trained to do one specific task very very well so we investigated the first dimension of what could be a hype the.

words that were using the terminology that we are using secondly the hype is also happening when lots of people are talking about something but most people actually dont know what theyre talking about and only a handful of people has really experience with the new thing is this also happening with artifice intelligence I think yes because if you look at the latest grid study and if you.

look at what were doing with artificial intelligence today in market research there are lots of people talking about it but only a handful of people is really doing something and most people read about it in the news but havent thought very carefully about what it could mean for the market research profession and what it could do to the market research industry because if you look at the numbers on this slide its only 23% that today already thinks that AI could be a game-changer for market research but if you look at what trend.

Watchers futurists management gurus say about AI they are all convinced together with all the scientists who are working on artificial intelligence that it will be a game-changer for every single profession that it will be a game-changer for almost every single industry but we as researchers seem to think that we are immune to this change so I think we need to level up our game and we need to start to imagine what.

artificial intelligence could do for us as researchers a second thing that we see when we talk about AI to people is that a lot of people have fear for it and that’s normal because if you look at the history of Technology humanity has always had fear for new technologies we were a little bit afraid of steam engines we were afraid when electricity was coming around the corner I think some people in the audience will still remember the days that internet digital was coming and it also lots of people didnt know what to think about it.

didnt know what to do about underestimated the power it would have and some even had fear for it and that’s the same with artificial intelligence but I think that we dont need to be afraid of it because artificial intelligence can actually be a positive thing for most industries and especially for the market research industry because AI can help us to get rid of the monkey.

jumps of the repetitive tasks that we still do on a day-to-day basis artificial intelligence can take away those tasks and automate them can do them forests can do – maybe even better and artificial intelligence systems can also help us to do things that were just not possible before can help us to augment our own human intelligence so I think we need to embrace artificial intelligence it can really be a good thing for us and if you look at my explanation about the difference between narrow AI on the one hand and general.

and super AI on the other hand and that most of it today is still narrow AI we could say that we dont need to be afraid of the fact that machines will take over our jobs in the next 10 to 20 years though machines AI systems will become our little helpers they will help us to do things faster to do things better to do things that were just not possible before meaning that we as humans can focus on empathy can focus on soft skills can.

focus on generating added value by using the skills that were really good at as humans using the skills that are unique to humans so its artificial intelligence a hype I would say yes it is because we use different words to do to say the same or we use the same word to say different things and we should say that artificial intelligence is the.

bra concept where technological advancement is creating machines that can help us as humans to do things better faster and different and secondly yes its a hype because we are talking about it but we are not making action and today I think the time has come that we should look at artificial intelligence and look at what it could do for our industry for our business and especially within market research a profession where there are still a lot of repetitive tasks and where there’s still a lot of room for.

improvement we should look into what can a AI systems do to make market research better to do things that were just not possible before if the time has come to dive into what is it what can can it do for us in to experiment to build systems little narrow AI systems that can help us to do different things to do real-life tests with those systems to.

learn from what goes well and what goes wrong and to share our experiences with the rest of the industry and that’s exactly what I want to do with you today I want to share with you 2 AI systems that we have built and we have tested I will share with you or learnings and I hope that by doing that we can make little by little and together as an industry progress and we can make our industry future-proof and we can make our industry a better one by applying AI.

systems so I will share with you two case studies the first one is a study where its a back-end system a back-end system that is helping our moderators on our research communities our consumer consulting boards to manage the health of their community the second case study will be around activating insights within an organization where we have created a chatbot called Galvan who helps lets say the corporate market researcher to bring insights to as many people as.

possible within the organization so its an insight activation chatbot why have we chosen those two domains within market research well because we think that those two fields or those two kind of systems and platforms are part of the future of market research we have applied AI to research communities or consumer consulting boards because communities of any form are the future of research or are a big part of the.

future of research if you look at market research spent today 5% of the annual global budgets that is spent on market research already goes to community systems if you look at predictions for the next 10 years amongst others by ray pointer he says that in 10 years from now 70% of all market research done will be done of community platforms so communities are definitely the research platforms of the future that’s why we think that if we apply AI to a methodology we should apply it to the.

methodology that will be most popular in the future so that’s why we have chosen to build a back-end system for our moderators to help them manage their community and their the health of their community secondly we applied ai to insight activation why because based on studies that we have done ourselves but are you also start done by others we see that the biggest unmet needs today of corporate researchers is not getting in better.

insights but its activating the insights that they already have into the organization in a better way making more impact with the insights that they already have so that’s why we created an AI agent a smart chat bot that can help corporate market researchers to activate insights within the organization well now share those two case studies with you I will explain what we have created and it will also tell you what.

our learnings and conclusions are I will introduce each of the case studies by referring first to a movie each time that probably both of that probably most of you will will know or are familiar with so lets start with the first case study proactive community management a back-end dashboards for community managers to manage the health of their community and I want to refer here to add the movie Minority Report in the world of Tom Cruise in Minority Report murder future murder can be seen future.

murder can be predicted and is also prevented if we take this kind of vision and we apply it to research communities or consumer consulting boards we could say that it would be fantastic if we could look into the future could see which participants will be still good participants in the future will they still participate frequently enough will it still be the case that what they say is insightful and that we cannot only predict it we only look into the future and see what will happen there.

but that also today we cant prevent the future from happening so that we dont have participants that will not participate anymore so that we dont have participants that will participate but dont say anything insightful or meaningful so that actually we always have a very healthy productive and insightful community so we have created a system to do that a back-end system a back-end dashboard that can help our community managers and moderators with that because the system that we have created uses two months of activity data on one specific community to predict for that specific community.

what will happen with every single participant in terms of the frequency of participation and the quality of participation the quality of what they say so based on what every member has done in the last two months we predict what every member will do in the next two months so that means that a community moderator can literally not look two months into the future they can literally see what will happen with.

every single participants in terms of frequency of participation and quality of participation in terms of how many times they will react to a certain task or a certain question and how insightful their kind of answer will be we use data that is both quantitative and qualitative for that we use how have they participated in the last two months whats the kind of parts a patient pattern and we also look at.

what they have sat in the last two months and we look at all their kind of verb attempts we apply text analytics to that and our text analytics system is looking at whats the cognitive effort that they have put in whats the cognitive effort they needed to give that answer so well look at their participation patterns both in terms of frequency and in terms of cognitive.

efforts put in to give an answer and based on the patterns of the last two months we predict for every single man remember the patterns for the next two months and based on that we can score every participants we can tell for every participant what kind of participant they will be based on quality and quantity of their answers in two months from now will they be a community star so somebody who participates all the.

time and gives very insightful answers or will they be in two months for now a high potential so somebody who is participating from time to time but when they participate their answers are really great their contributions are fantastic or will they be louder on the low ends pacifists or lawyers and its especially when we see that we will have a lot of pacifists and a lot of annoys.

that we need to take action that we want to take action today to make sure that the future will not happen that they will not become the participants that we have predicted for the next two months so how is our system doing well today we are able to predict with 78% of accuracy what their quantity of participation will be and 71% of accuracy what their quality of.

participation will be in the world of prediction systems this is really good how is the scoring compared to human moderators well a human moderator will would probably do slightly better if he or she has the time for that but the thing is a human moderator will never look at two months of data for every single individual participants to predict what that person will do in the next two months and they will for sure not do that for 100 or 150 participants so this system is maybe not as good in predicting the future as a human but it.

can do this in no time almost in real time for a large scale of participants so that’s the beauty of the system its the difference between the system doing it for you at slightly lower accuracy or you not doing it at all as a moderator so that’s really cool so the moderator really has a view on what will happen with my community in two months from now with every single participant and then there.

is also the prevention capability of course the system learns how it can prevent the future from happening so the system really gives for every single participant based on the quadrant theyre in a suggestion to the moderator what action they should take so normally a community moderator will manage the community as a whole the system here allows to take a very specific action.

for a group of participants given the quadrant theyre in so a lawyers get a different kind of prevention approach than pacifists for example and the system based on who those participants our will give a suggestion what kind of action could be taken for that group of participants so for some participants the system will suggest send them a reminder email send those guys a motivation email pick up the phone and call those five participants or send a.

text – message to those ten so that means that not only the community moderator can become a proactive manager of his or her community they look at what will happen from now but they also take action today to make sure that it will not happen they can for example have three groups of participants then who they need to call 20 that they need to send a text message and five others that need to get a motivational email so.

its saving the moderator time because the moderator doesnt lose time in the future when the health of the community goes down though he or she takes action today to prevent the future from happening and he or she is using a very personalized approach to be sure that every single participant will gets back on the right track so this back end system this kind of AI driven dashboard is really a time saver for community managers and it allows them to keep.

their community healthy all the time and it allows them to not only have some kind of general kind of engagement approach towards the community but to also have a personalized kind of approach for certain member groups what have we learned about AI while building the system well first of all that for AI systems you need a lot of data so its really good that in the last 10 years we have done almost a thousand communities because that means that we have a lot of data about what kind of.

behavioral patterns are of community participants and its all of that data that allows us to see patterns to teach our AI system what community members do and how they behave and based on those learnings our AI system is making predictions for the future so capture data that’s the first thing you need to do because its only when you have a lot of data that AI systems can learn from their data to make predictions secondly we have also learned quite a couple of things about adoption of this systems by our moderators we have seen that for the.

people that have been part of the creation of this system and that have been part of the tests that we have done with this system that you need to tell them what the AI system is doing what the logic is behind what the machine is doing because its when they understand what data the Machine uses what kind of patterns the machine recognizes and how the Machine makes its predictions and its suggestions for a kind of.

personalized approach for each of the members its only then that our moderators will trust a machine so dont make AI systems black box make them Y box machines will only be trusted by humans if people though what the machine is actually doing and last but not least we have created here yes a AI dashboard for our moderators to look into the future and to manage their community in a proactive way but based on all those predictions based on all.

those patterns we also some best practices about moderating communities so lets do the dashboards the system is giving us also all kind of other kind of best practices for example we have learned that’s a funny thing that if a community moderator talks a lot about him or herself that that’s detrimental for the help of the community that’s one of the kind of more general learnings di system has given.

us next to the prediction dashboard so that was the first case study there was a back-end system that was a system that moderators can use to help them to manage their community in a better way in a more proactive way and to be more productive the second system is not in the field of inside generation but in the field of inside activation and its not a back-end system but its a front aunt system and here I want to introduce.

this case by referring to the movie her in the movie heard the actor Urich in Phoenix is actually falling in love with his AI personal assistant by the way the AI as a system has the voice of Scarlett Johansson and you see in the movie that little by little he is trusting the system that little by little the system becomes part of his hate day to day life.

and that little by little he is literally falling in love with her so if we take this science fiction movie if we take this vision and we try to bring it back to the day to day job of a corporate market researcher we try to take it back to the world of bringing insights to as many people as possible within an organization what are the problems we need to solve well first of all there can be a lot of insights within an organization but probably they are locked up somewhere in a PowerPoint.

report of 100 slides that sits on a server so if a marketeer wants to find back that one little insight that he or she could use to make a better decision for the next innovation or the next marketing campaign its almost impossible for that person to find it back or it would take a lot of time even for the corporate researcher they didnt call or mail its still a huge effort to go and look in different PowerPoint.

reports go through lots and lots of slides to find that one little insight or key factor key figure that could help that marketing person to take a better decision so if we could solve that that would be great if consumer insights and the right consumer insights would be at the fingertips of a marketing person and a corporate market researcher that would be fantastic secondly if consumer insights could be used by as many people as possible within an organization that would increase the return on investment in.

consumer insights quite drastically so we want more effectivity by bringing it to more people we want more efficiency by having literally consumer insights and our fingertips so how can we use AI for that well we have created an insight activation chat BOTS we have created a smart assistant for market researchers and we actually have created a robot colleague if you want for every CMI.

manager in the world so as a front and system we use here chat bots chat bots are did you apps if you want because apps you need to open them you need to fill out lets say a forum before the app can give you what you need but you dont need to open an app you know you just chat with them in messenger whatsapp or Skype for business.

and you dont need to fill out a forum though you literally have a conversation with them so that’s why today but chat bots are called the new apps so we use that as a front-end system and we made the bots smart by using again just like in the previous example narrow AI meaning that its on the one hand rule-based the system knows if you ask.

the question what the answer is and the system is also to some extent self learning it gets smarter if it knows more about you about your company and about the type of questions you ask so what can this smart assistant this inside activation chat bot do well galvan as we call the bots can do three things for you the first thing he can do is it can impersonate a consumer for example if you have a Sigmund tation study and for.

each of the segments you have created a persona galvan can become that persona so suppose that you one of your segments are young ladies and you have created a persona called Joanna well galvan can become Joanna so you can have a chat conversation with Joanna Delfin will use all the key facts figures and insights that he knows about that particular segments to.

give answers so this is a way to meet up with customers to meet up with an aggregation of all your customers and to bring personas to life in a slightly different way the second use case is where Galvan becomes your coach so Galvan is connected to a database of insights a database of key facts and key figures its still a database that is curated by humans but Galvan knows how to find things back in a database and.

give it to you in a smart way by answering your questions or even if its not clear to him what you need asking a couple of follow-up questions but that database is also in the cases where our clients have our social network to distribute insights throughout the organization or studio its also linked to what other people within the organization are doing with those insights so in the second use case Galvan is telling you what other people are doing with certain insights other.

people like you and in the morning when you first come on whatsapp or slack or Skype for business or facebook Messenger government say given that other people are using this particular insight other people like you probably you also want to notice inside because maybe you also want to apply that insight to your work because a lot of your colleagues that do a similar job are working already with that inside we envision a future where Galvan can look into your calendar and.

can tell you well based on the meetings that are on your plate today given the fact that you’re for example meeting up with the advertising agency about that new campaign for that particular target group maybe it can be handy to read these three insights because they can serve you during the meeting the third use case is where Galvan becomes lets say your assistant where you sit into a meeting.

you for example are deciding together with a design agency on the next pack for your product and you’re doubting between option a and option B you can then take your phone start a chat with Galvin and ask Kelvin Kelvin what do we know about their kind of distinctive assets a packaging for that particular product should have to be successful on shelf to catch immediately in the store the attention of potential customers and Galvan will look up what he knows about.

that and will tell you in just a matter of seconds you can use that to make the decision and you can continue your meeting so here Galvin literally becomes your kind of personal assistant your kind of insight activation helper what have we learned well the people had infused the system find adoption really easy because its like chatting to a person in your messenger system in the system you already use so there is a very low barrier to start to use Galvin is Galvan perfect no is Galvan super smart.

not yet but Galvan can do the trick Galvan can bring a persona to life Galvan can recommend you what you should read Galvan can help you during the meeting in a better way than a search engine would do so people are quite satisfied with Galvan today its helping market research managers to save time its bringing insights at their fingertips anywhere anytime its helping so also the corporate research to save time because they say that 20%.

of their time goes to answering those type of basic questions that Galvin now answers for them it makes consumer insights more used and popular within the organization and eventually it makes that consumer insights are used to make decisions so that more decisions are consumer centric decisions so what have we learned by building this front-end AI system well first of all that its all.

about relevance its not about what can we do for AI systems its about what can a our systems do for us so that means that when we think about AI systems we need to look at whats the market research process whats the process of activating an insight within an organization and how can a I help us to take away repetitive tasks to do things quicker to do things different to do things better here we literally looked.

at how can insights come to life within an organization what could be three use cases in this case where AI could help us and where a chatbot could help us so we define the three cases that I just explained to you based on an investigation looking at human behavior at the behavior of marketing people CMI managers when it comes to activating insights secondly we also learned that.

next to the fact that the system needs to be smart the system needs to have also chit-chat capabilities its not only that the system needs to follow certain logic it also needs to be able to behave a little bit as a human to have a certain personality to be able to ask you how your day went to be able to.

ask you if he was of help last time that you used him why is that important well because of the fact that it feels a little bit more human that such as chad bob has personality is crucial for adoption people find it fun and interesting to talk to Galvin they want to see how he will not only help them but also how he will react to certain questions to the chitchat people even ask Galvin Galvin can you tell us a joke and how Galvin knows a couple of jokes are not the best jokes but it.

makes that Galvin feels a little bit more like a colleague do we pretend that Galvin is as good as a human no its about managing expectations on the one hand and we dont give Galvin a human face we give him a robot face because for now that is what those AI systems are they are pieces of technology that have specific intelligence to do.

specific things very well so for now all those AI systems are little robots are little smart machines that are literally our little helpers to do parts of the research process or the insight activation process quicker better faster cheaper so were coming to the end of this this webinar what can I say to conclude well first of all I think the time is now to stop the talking and start the experimentation we have started first experiments you have seen.

to where we build it create it and build a back-end system for our moderators and where we created a front-end system for some of our clients already a system to help you to generate insights in a different matter way a system to activate insights in a batter in a different way but we can think of many more possibilities throughout the research process and the inside activation process where AI systems can help us we can think about systems where they suggest what type of questions we should ask given the briefing we can think.

about buts that can help us to do moderation we can think about smart systems that I can do surveying in a different way there are already many systems that help you with the analysis of quantitative and qualitative data there are systems that can help you to bring the ride inside to the right person at the right moment and that’s the exercise we should do that’s where we should think about what can narrow AI do for us already today and I would say start building those systems start experimenting learn and share your learnings so that as an industry we can.

move forward because if you look at what has happened in the last ten years in the market research industry in the last 20 years in the market research industry we have seen online coming we have seen social technologies coming we have seen mobile coming and they all had a huge impact on where we are today in market research digital social and mobile if we look at AI I think the.

impact of AI on market research will be bigger than those three waves of change together it will shake up the market research industry and when you look at the last 10 years we have missed a lot of opportunities as the researchers we have given away a lot of the social media analytics to players outside of the market research industry to functions outside of market research within companies the same for the explosion of data and big data and Linux we have given away a lot of that to other players and to other and functions within companies lets not.

miss the boat for with AI lets jump on that AI train and yes some of the experiments will be disappointing some of the experiments will fail but its about learning its about start doing things so please start experimenting and please allow us as an agency also to experiment together with you and if we think about the future of our industry I would also like to encourage you to take the great survey we will share the link.

with you in the chat we will also share the link with you when we share the presentation with you later today over email yeah our industry is one that meets innovation and I think taking a survey like that is also helping us to understand where the future of our industry lies so take the survey and start to think about what a I can do for you start to think about your first AI experiment so thank you very much for listening to this smart Ts webinar Im.

happy to answer some questions right now that you posed on on the on the chat Im also happy to answer questions over email or on Twitter and LinkedIn later today and if you enjoyed this Smarties please take a look at our Smarties website because we have two interesting ones coming up as well so I open the floor now for four.

questions first question is one from mark what are prerequisites to benefits optimally from artificial intelligence you mentioned having much data but how structured tact should this be are there other important pre requisites very good question so data is important if you look at the two systems that we have created the first one is making use of millions of data points millions of data points of what community members have done in different communities and different communities over the last ten years is all of that data are very structured in the beginning no its.

literally different data sets sometimes coded in a different way but our data scientist has done a lot of work of bringing those different data sets together so if you can make sure that its somewhat structured that’s already a good beginning but those data scientists are really very smart in bringing different data sets together and and in structuring all of that so capturing data is the first thing to do secondly I think its about thinking very carefully throughout the research process in the inside activation process.

where can a I actually help me where do I have very repetitive tasks where do I have needs to be smarter so to say so its mapping out where can a I help me or my colleagues where can it be relevant for us so that’s the second exercise you should you should do and then the third kind of recommendation I would give a stick back but start small because its all about using the data and certain pieces of software to create meaning to create value so there is a.

lot of going back and four words between what data you have whats technically possible and where you can create value for the users of those systems so I think make but start small and do a lot of iterations so I hope that that was an answer to your to your question we will in the next couple of months by the way write a more extensive paper around this topic so we will keep you posted around that as well so if there are no further questions.

here on the chat Oh mark has a follow-up question really great so your first case mentioned as suggestions presented to moderation how strict is the follow up on these suggestions are they fed back in the system do you have already results on feedback good question so we have built a system and tested it with a couple of our moderators it our suggestions so that means that this is.

the system makes suggestions its still up to the moderator to decide what to do or not so that’s an important remark to make and yes in the future we will include the feedback from our moderators as a variable in in the prediction model but for now we havent done that yet but its still important to to know that the system is not making a decision for the.

moderator the system is suggestion is making a suggestion to the moderator so after the questions I will close up here but feel free to email me Tom at insights dot EU or to drop me a question on LinkedIn or Twitter so thank you very much have a great day and we hope to welcome you soon at one of our future events or webinars thank you very much.