One of the major changes we’ve made to the bidding page (although it appears relatively minor on the page itself) is how bids end. Before when choosing an expiry date for your bid, you’d just click on the appropriate link, and a calendar showing the current month would pop up. You’d flip through to the month you wanted (assuming, of course, if you want the bid to expire on a day in the current month), choose a date and time, and you’d be good to go. It looked like this:

and when you’d click on that yellow “ending” text, you’d get something link this:

Pretty simple, right? We thought so too, until we discovered this one interface element caused a small market drop 12 times a year.

The issue is that most bidders don’t really care when their bid ends. Some do, of course (which is why it’s great to have this feature), but we discovered that when most bidders would choose an expiry date, they’d use approximates, saying things like “I want this bid to last about 2 weeks” and select a date appropriately. So far so good. But rather than always choosing a day about 14 days away, the date bidders would choose was being influenced by the interface.

Let’s say it’s the 15th of March. Our bidder, wanting his or her bid to last about two weeks, would most likely click on the last day of the month, the 31st, since that’s pretty close to his or her goal of two weeks. One thing they wouldn’t do was click on the calendar once to move it ahead to April, and click again to select April 1st. Selecting March 31st instead of April 1st saves our bidder two clicks, and the two dates are both close enough to being two weeks away from the 15th that it’s not really worth the hassle.

You may see where we’re going with this!

Now let’s say it’s the 16th of the month, and we’re in the same scenario. Our bidder is again most likely to select March 31st, since it saves time and clicks, and while it’s not as close to two weeks away as it was on the 15th, it’s still a pretty good fit. And as days go by, this bidder keeps selecting the 31st of March as their “two weeks away” point until we’re so close to it that it’s not a reasonable approximation of “two weeks away” anymore. At this point the bidder either thinks “Ah well, one week is enough” and again selects March 31st, or they bite the bullet and click twice more to select a later date in April.

The result of this was that we would see a spike in expiring bids at the end of every month. This was less than ideal for publishers, because at the end of every month all these bids would disappear en masse, and bid prices would drop across the board. Of course, that meant prices were cheaper for advertisers, but as we’ve seen recently in the world economy, a stable market is often preferable to one with drastic changes in it, even if this volatility is somewhat predictable. It would usually take a day or two to push prices up to where they’d been before the month ended.

Our new interface is similar, with an important difference: if you want to enter in an expiry date (and you don’t have to), you now have two ways of expressing this. The first (and default) way is to select the number of days (or weeks, or months, or years) that you want your bid to last. The second option is the one we discussed, with the pop-up calendar that allows you to choose a particular day. It looks like this:

What this new interface does that the old one didn’t is make the desire for efficiency we all share (some might call this “laziness”, but I disagree!) into a positive thing. If you’re there thinking, “I want this bid to last two weeks, and I don’t particularly feel like looking at a calendar to figure out what day that actually is”, this allows you to enter in the most important information (the length of your bid) without having to transform it to a date. The result is that bid expiries are no longer clumped around the end of the month, but happen throughout it instead. This means is a more stable and consistant market, which is good for everyone!

It goes to show you the importance of interfaces: even small things like this can have larger effects on the system as a whole. Normally improvements to our pages like this one are made without this sort of post to go with it, but it was really interesting for us to discover what was going on, why, and to redesign (and test, and redesign, and test, and redesign) the page appropriately. I hope you found it as interesting as we did!