Hints for Contextually-intelligent Design

Remember the old carbon-paper credit card receipts? They don't really exist anymore because credit card companies stay in business by designing all their processes towards fraud reduction. And by implementing constant improvements, they have becomes better at it than any other type of service.

A huge shift began happening a couple decades ago. Real time approval, first with dialup modems, allowed them to begin to intelligently process transaction records to determine what their customers are up to. And of course deny or accept transactions based on the patterns they found. Their success rate, and the seamless ease of use to consumers has been an object lesson in both fraud prevention and general customer management analytics for at least a decade.

But there is a further lesson in this, as when it falls down they are very bad. Just yesterday Alison went to buy a ticket for her annual trip to Canada. Denied. I had to call the card company's fraud hotline to clear the transaction. Did you see the part where I said "annual trip." There is a pattern, but their system has been set up to recognize only short-term patterns. In this it is missing long-term patterns, inconveniencing us and probably missing out on an important understanding of our actual purchasing behavior. (Oh, it's not just us. Barbara hardly even tries to use her cards when speaking in the UK, and every other person I know had had declines or has to call ahead when travelling).

A lot of us are old enough, or nerdy enough, that we're used to hacking the system, installing updates and working around bugs. But as mobiles move to be appliance-like – as they become more invisible, seamless, and delightful – the more that cracks in the UI become glaring.

Like the way that credit card anti-fraud processing is so seamless any failures in its intelligence are glaring, contextually-intelligent mobile devices, apps and sites will need to be designed not just to be a little smart, but so that any failures are not jarring to the user.

This has been soaking in my brain overnight, and I've started thinking more and more about how the mobile, internet-connected world should be a lot more intelligent and contextual today.

• Make your information intelligent – Far too much customer information is locked away in specialized data stores. While certain visions of an ideal world would say it should all (barring security and privacy concerns) be open, you might not want to open your customer records to everyone. But I'll bet you still need more openness. Make records more easy to get to, process, index and search than they are now. Make the information semantic, so it can be understood in bits and pieces, and searches for meaning more easily. Your information is a key asset, so you need to be able to use it to stay in business, as well as to serve more intelligent, useful information to your customers.
• Sense your customers – Normally when I say this, I am referring to device-based sensors, and am reminding everyone to use all of them; gesture fails because most use accelerometers, but adding cameras, gps, bluetooth, etc. will provide additional contextual information. But you can also be clever without access to device sensors. Try to pick up context from every touchpoint customers have with you. What time are they accessing the website, from what device. Make sure customer care emails and calls are annotated to include key concerns. And don't force-sort, eliminating secondary issues of note. What can be divined from these patterns of use? Even if nothing today, maybe something eventually.
• Interpret their needs – How are you going to find out what all that data means? You'll have to talk to actual people. You need to hire research for this, and it's likely to need to turn into an ongoing research program, that tracks changes in user needs and satisfaction and behavior over time. Don't just guess, don't assume your mom or your poker buddies are typical of your user base, and don't take user comments at face value. Hire research folks to do this for you.
• Remember your customers – This is the direct lesson from our yearly travel story above. People have behaviors that are at human speed. Just because they only perform an activity every year, or decade, doesn't make it less relevant. Don't delete accounts after a month of disuse, and don't automatically disregard rare occurrences as spurious. Try to understand the users, and make your systems work for them for a change. Research teams will love to parse this information as well.
• Design your product like a toaster – Like you, I'm a big nerd. I have a toaster oven, with several inter-related controls. I do all sorts of stuff with it, practically making dinner sometimes. But to get mass-market adoption you need a toaster. Bread in, lever down, wait... toast up. You can choose the degree of toastyness, but it's a persistent control, so you don't have to select the setting each time. Simple products that meet specific user needs, easily and seamlessly, are appliance-like devices that succeed. Consider features carefully, and refer to the user behaviors and needs I just told you to gather up.

• Start serving contextually – Whatever service you offer, especially if it's already mobile, start making it contextual. The simplest changes to your static site can add a lot of value. Can you serve only information relevant to the current time? A lot of stuff (weather, say) can be useful without pinpoint location, and region can be guessed at more easily than you might think.

And as you proceed, and upgrade over the years, keep all of this in mind. Don't forget your users, add features for the hell of it or build more locked-away data stores. Think contextually and you'll be ready for the future of mobile.