“Data Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.
Today’s column is written by Rob Leathern, CEO of Optimal, Inc..
Having had the fortune (or misfortune) of being involved in online advertising product development over the past nine years, I’ve been able to watch how, on average, each successive advertising server and system gets built faster and faster, with fewer of the old (product, engineering or business) mistakes. One thing that is still universally done very poorly by vendors and publisher ad systems alike, however, is the handling of time.
This article is not about real-time bidding – which mostly means “impression-level bidding,” and which is only available for a relatively small percentage of online inventory and even less mobile inventory. No, I’m talking about good old dayparting (turning your ad campaigns – bids – on or off at different times of the day or week) and event-based triggering (changing campaigns when something new happens). Time-of-day and event-based targeting, for the most part, are absent or really suck today. And that poor implementation is a growing problem, considering that this type of targeting will become ever more crucial in a multi-screen, increasingly mobile world.
Let’s start with the desktop. Have you tried to daypart an online search or display campaign recently? The user interfaces for setting and changing when ads run is confusing (Google AdWords), dysfunctional (Bing) or entirely absent (Facebook). And they typically don’t account for differences in users’ time zones! Yahoo’s Right Media platform is a notable exception (although it still has its 2008-era user interface).
In other words, you basically have to create different campaigns for each time zone. Meanwhile, I’d bet $5 that fewer than one in five ad-tech executives could tell you what percentage of people live in each of the US time zones (For the record, it’s 47% Eastern, 14% Pacific, 33% Central, 5% Mountain and less than 1% Hawaii-Aleutian). It annoys me that some major platforms report everything in Pacific Time (which, again, represents only 14% of users) with no ability to adjust, but – even worse – some insist on reporting in Coordinated Universal Time (UTC). At least the engineers seem to prefer UTC!
Is it just that we don’t “get” time, time zones and time changes…or is it a publisher-driven conspiracy? My theory is that it’s the fault of a relatively efficient market. I think that, until now, dayparting and event-triggering opportunities simply haven’t been compelling enough to justify large-scale investments by publishers or ad systems.
Most ad buyers aren’t going to take the time to manually analyze small-scale differences, but algorithms can find these things and make adjustments at an advertiser level or across multiple systems. Here’s some free, real, actionable data: Mobile ad-conversion rates drop 10%-15% on Fridays, and Twitter and Facebook mobile ads on Android get better engagement rates than on iPhones.
Here are some of challenges to dayparting and event-triggered advertising:
Ad-system changes seldom happen immediately. Most systems have a delay before ads turn on or off and before bid changes are enacted. AdWords is way ahead of Bing in minimizing this lag; Bing only lets you make changes based on seemingly arbitrary buckets of time and, according to a support article in February, it may “take anywhere between 1-4 hours for the changes [to your dayparting] to go live.”
Publishers want to spend all of your budget, every day. This may seem like an obvious point, but dayparting messes up big publishers’ budget and pacing algorithms and introduces way more complexity than you might think. If certain times of day are generally less desirable to multiple advertisers, what are they to do? They can’t exactly throw that traffic away, and thus many would prefer to avoid dealing with this issue altogether.
Collecting hourly or time zone-dependent data is not just a storage issue. Storage is cheap, but the data volume isn’t the biggest issue; the biggest difficulty is the ability to rerun user-level analysis across arbitrary time periods, which is non-trivial and adds complexity. Counting impressions is far easier than counting unique users. Therefore, most systems don’t let you reaggregate data based on different time zones or time periods.
Mobile, which is still very new, is more fragmented than desktop – and it makes time far more important. If we haven’t yet sorted out time-based ad buying and bidding on the desktop, optimists would say that we might be able to leapfrog the issue in the further-fragmented mobile ecosystem – but realists are probably rightfully skeptical. We’ll see.
Precise location is still “El Dorado,” but we need to solve time first. Nobody is getting a Starbucks coupon when they’re walking by a Starbucks. This is the oft-supposed ideal of mobile advertising. I’d argue it would probably be unprofitable for Starbucks, as it would likely just subsidize the company’s existing customers, but let’s not open that can of worms. I think time-based advertising on mobile will happen in ways where location can be actually useful. For example, a restaurant might bump up its bids 500% the hour before lunch time and – later, when available – in a 10-block radius around its location, but probably not while someone is walking by, as it’s unlikely that person would be looking at his or her phone right then.
Mobile promises to make solving these challenges even more fruitful than they might have been in the past:
Google and Facebook authentication make logged-in mobile usage significant. This is a big one. The promise of consistently being able to address users across both mobile and desktop means that sequencing interactions – finding a way to gather combined data on a desktop or tablet – and getting lighter interactions on a phone may become a viable model.
Smartphones are both standalone AND companion devices. Having an unrestricted Internet-access device in workplaces that might restrict desktop usage is one opportunity where retaining cross-device context could be valuable. But smartphones are also companions to other media, which enables the following point:
Television plus mobile will be the catalyst for time- and event-based triggers. We know ahead of time when certain television shows will be playing. Either we can get the TV ad schedules ahead of time or – more likely, though not easily – we can recognize television ads playing in real-time and trigger simultaneous mobile and/or social ads (or show-specific tweets) based on those television ads. (This is already starting to happen; see this story about Twitter Amplify, announced in May.) This ability not only ups the ante for dayparting, putting it into the context for users’ local time (again, time zone issues come into play here, except for live events broadcast on TV), but also creates a unique new category of event-based advertising.
My guess? You’re going to hear a lot more about dayparting and time-based advertising in the latter half of 2013.