These days there is a overabundance of information available about every single person using the internet. Ten years ago people said “I know you are gathering data about me, but I hope you use it to benefit me”. Now people have stopped asking politely and added incognito filters, ad blockers and other means of privacy to their browsers to block the marketing bots. This is all a futile act when in the same time people share their life to the fullest on multiple social media accounts. People constantly leak sensitive material about themselves and give AI and marketing algorithms a way into their mind.
I give out a lot of my data to the different cloud services on the web. I use different fitness trackers and sleep analysis tools. These are all mixed and matched with my IFTTT integrations and therefore are leaked throughout the world. I don’t care. I would actually hope that someone somewhere would pick up on my patterns of body temperature and alertness by offering me something that would help if there is for example a flu lurking in the future for me. What I do try to keep secret are details that could help uncovering my passwords, like maiden name of my mom, my childhood pet names, etc. Usually people still give all these out to funny memes and miniature apps on Facebook and Instagram.
Remember this one?
Someone certainly went for that too, but most went for the previous questionnaire revealing all your password protection question answers.
But marketers are not trying to steal my credit card, right? Well, they just want to ship you goods and get you to buy them. How accurate the data should then be to be effective? Normally the gathered data is less than 20% accurate, which is enough for good results. Actually for every percentage point more accurate than 20%, the revenue of using this data decreases since the target audience gets smaller. And as Simon said in his talk, the data needs to be timely to be interesting. In this context timely means, that even though you hit your target group, only some of them are in the market for that type of goods right there and then. To get conversion out of your campaign you need to hit wide enough to find all parties that see your message as relevant and timely for them.
I went through a couple of examples of when we use the data we have about a person and don’t try to understand what it means. When we don’t ask WHY or HOW COME.
I commute to Helsinki from Tampere and I love coffee and coffee shops. I am a social individual and this has made different algorithms target me as a potential coffee shop goer in Tampere. But actually when back at home, I tend to invite friends over and offer them coffee from my french press. Where I do go to coffee shops is in Helsinki, where the luxury of my home equipment doesn’t exist. So targeting me to go to Tampere coffee shops doesn’t work at all. This is inconvenient for me, since I don’t get coupons and offers about the new coffee shops in Helsinki that I could try.
A few years back I noticed a peculiar commonality in my Facebook feed advertisements. They all were boasting some sort of artificial fertilisation services. Apparently algorithms had decided that a person in just over their 30’s with a spouse had to want children. And since there were none, there was a need to “correct” the situation – so the bombarding of ads from different clinics etc. When I didn’t click on those adds, the algorithms decided, that the problem must be something else. Maybe my spouse is the issue here? So they added dating app ads to the mix. Interesting combination indeed when these two co-existed in my feed for some time. When I didn’t respond to these either, the algorithm decided that maybe I have managed to hide my children from the internet and I have them. So started the diaper, children toys and clothes ads. Even today I have “matching pyjamas for all family” type advertisements at my FB. The algorithm didn’t have the option anywhere that I might not want children and that that was “the issue”. It decided based on roughly 15% accuracy that “all women want children”.
IF this was something involuntary or there was a tragic reason for us not having kids, how do you think I would feel about those ads and brands behind them? So should the advertisers care about personalisation gone wrong, as long as they hit their conversion targets?
They have to. In today’s connected world my bad experience and hurt feelings can reach a multitude of people in the brand’s desired target group. By crossing someone’s core values you will get thrown under the bus in the internet. Even though usually we say 4-5% is a good conversion rate, as a marketer one should care about the remaining 95%. Try to understand why it didn’t resonate? Was the timing off? Was is irrelevant and why if so?
Even though a multitude of websites tell you that they are tracking you to provide better user experience, it is not that simple. User experience cannot be build only on data available. No brain, no gain. To improve customer experience you need to take into account all channels and context of the customer. You need to figure out what is important and what is meaningful. In addition to algorithms, you need to be human.
You can use analytics to spot the trends on data and then ask only few people why does that happen. Always ask “why” five times, to get to the root cause. Like my coffee shop example. Why don’t I respond to offers at my area? —> I go to coffee shops at Helsinki. —> Why is that? —> I have cool gadgets at my home to serve coffee. —> Why is that? —> Since I love good coffee and friends gathering around to my home. Etc.
But could you NOT do personalisation and targeting? No, everyone else does it, so you have to too. There is expectation level to this, and people will be offended if you are not interested in them like every other brand they encounter. But be good at it. People will feel offended also if you do personalisation and you do it only based on data and you happen to be wrong. Also don’t over do it and be spooky.