The lost art of efficiency in web design

A few years ago I was interviewing a woman about a travel website that she loved. She explained how it had made her life better.

‘When I’m planning to fly, I block out my evenings for the week and I use this site to find the best deals. I normally spend a couple of hours a night looking for flights. At the end of it I get the best deal,’ she told me happily. I asked her how much  she saved by going to all that trouble. ‘Well,… easily £10 [$15],’ she said, looking a little less happy.

The fact is, people are poor judges of time – especially when they’re kept busy. That means that web design is often unconcerned with actual efficiency. Perceived efficiency is what matters most.

If our attention is engaged, we don’t tend to notice that our time is being wasted. That’s why ‘loading’ animations are so useful – they stop people from getting bored when the computer is taking a long time to respond. They keep up ‘perceived efficiency’ even when actual efficiency is low.

A decade and a half of desktop web design has relied on this kind of misdirection. We’ve been able to focus on effectiveness and satisfaction (both are important and hard to get right) but we’ve not had to think much about efficiency.

That approach isn’t good enough anymore.

The rise of the smartphone means people are trying to do the same things online (buy furniture, edit photos, manage their accounts) but on smaller screens, over slower connections in distracting environments. We need the simplicity that comes from highly efficient interactions. We can’t ignore ‘efficiency’ any longer.

The bad news is, designing for efficiency is demanding, difficult and counter-intuitive.

A few weeks back, we were looking at improving an address input for a responsive website. We needed the address to be validated, so we had the choice of two systems. One was the fairly standard ‘input your post code [zip code] and pick from a list of choices’. The other was a new autocomplete system where you start typing your address and it autocompletes.

It wasn’t clear which was the most efficient. We couldn’t simply run a user test: the interaction took just a few seconds and we’d need to have tested hundreds of people to get a statistically valid answer. Nor could we AB test: it would have required days worth of development effort that no-one was prepared to commit to possibly shaving a few seconds off the interaction.

The normal reaction would be to get everyone in a room and argue about it for an hour. In which case we’d have spent a day or more of consultancy time and the decision would have gone to the loudest person in the room.

Instead we picked up an old tool that fell out of favour in a world were perceived efficiency was king: GOMS KLM. This is a painstaking way of modelling user behaviour one click at a time. Modelling those two simple interactions took about an hour. But it yielded a clear winner based on evidence not opinion. The autocomplete approach was about four seconds faster and less prone to user error. Not the answer that many people were expecting.

It’s that kind of painstaking analysis that’s required if you want to edge an interface to genuine efficiency. But added up, those seconds become minutes and those minutes become an advantage.

A couple of years ago, one of our clients spent £250,000 [$375,000] to shave a few seconds off their shopping basket process. They saw payback in less than a year.

To achieve simplicity, you can’t just focus on efficiency. You need to score high on effectiveness and satisfaction, too. There’s no point in being efficient, yet annoying.

That’s what makes simplicity so hard to achieve – you’re shooting for high scores in all three areas. But there’s a satisfaction to the craft of perfecting an interface in this way. And, of course, it makes a real difference to your users and your bottom line.


  1. I’d be really interested to know how you went about doing this.

    I see a space opening-up for simple apps that support these old HCI techniques.

    • gilescolborne

      Something that handles the KLM rules and maybe allows drag and drop tasks would be very interesting.

  2. Did you just use a spreadsheet at the moment?

  3. gilescolborne

    Yup. Spreadsheet, pencil and paper… depends on the size of the interaction. The ones described above were around 20 steps (it’s always more than you think). The hard part is making sure you’ve not missed a step. Adding up the numbers is easy.

Leave a Reply