Focus on the consumer, not devices or channels
We all know that it can be problematic to think and plan in device or channel-based silos. Doing cross-device right requires that we build plans around the customer and their particular cross-device usage patterns. Consumer media time is highly fragmented, and they use different devices for different tasks. True cross-device marketing refers to targeting the same individuals across the devices that they use to access the web.
Reaching some people via PC and other people via a separate mobile campaign isn't really cross-device marketing, and a great deal of industry research has shown that reaching people across their devices drives significantly better results.
This isn't a semantic nuance -- it's fundamental to marketing in a multi-device world. According to research from Sophos, the average American spreads their digital usage across about three devices. And comScore says that fully one-third of digital time is now spent on mobile devices. To reach and persuade the user in the most powerful way we need to move our messages with them as they flip between devices throughout their days.
Ensure sound user-to-device matching
It's darn hard to link all of an individual's devices and behaviors to a single user ID. The Sophos study I mentioned earlier said that on average our digital behavior is stored across 12 IDs. An ID for your PC, another two or more for your phone and tablet browsing, and a number of IDs for the various apps you use. Uniting all of these data spaces to one common ID is very difficult -- very, very difficult.
I'm going to outline two different approaches companies use to accomplish user-to-device matching: inferred device matching and login-based matching.
Inferred device matching is a multistep process that, like it says on the tin, infers a match between a user and a set of devices. Broadly, doing it involves a couple of steps. First, you need to associate everything that happens on a single device to one device ID. All the data sets for each of the IDs on a device must be aggregated to a single profile.
Then you have to find ways to associate devices together to a household level. One of the key methods used is by examining the IP addresses we use. Multiple devices on a home IP, for example, are much more likely to be used by the same user than devices that connect via different households.
Household level matching is an important advance for the industry. But it is not -- emphatically not -- user-level matching. Consider a household with a family of four -- mom, dad, and two teens. If these users are typical Americans, there would be 12 devices associated with a household ID. A brand might get better results by targeting the dad across all these devices than by advertising on 12 random devices. But a great deal of that media would be wasted because pops really only uses three of them.
Device-to-user matching is inexact. Broadly, matching is enhanced with more data points. So, for example, seeing two devices connecting on the same IP address provides some evidence of them being used by the same person. But if that IP address is for the Starbucks at 40th and Lex, the likelihood for user-to-device accuracy based upon one recorded connection per device is quite low. We need more data to make a more accurate determination.
Now, if two devices both connect at the same 16 distinct IP addresses over a five-day period, there's significantly more certainty that they share the same user. It's still possible that those two devices aren't used by the same person, but the more data points you add to your equation, the greater that likelihood becomes. IP addresses aren't the only thing that can be used to infer a match. Similar browsing habits, for example, could also provide some additional assurance of a match.
Shared devices have the potential for clouding the soup even more, though new methodologies for associating sets of behavior on a device with different users have been deployed by some providers.
Inferred device matching is the most common method in use today. The other method is based upon recording first party logins on different devices and then matching these PCs, phones, and tablets using that login as a data point. If, for example, I login to the same financial portfolio account on PC and smartphone, the odds are pretty good that both devices are mine, or at least often used by me.
This user-to-device matching method works best for a company that has a lot of logged-in users. Otherwise, scale would be a real challenge. Further, what's absolutely critical here is that any PII must be disassociated from that profile. Note also that login-based matching has a high degree of accuracy, but this method also has the potential for mismatch if the logins are shared. Nothing is 100 percent -- at least not yet.
Shared devices can create problems here as well, and it's likely that login-based solutions have or will deploy inferred device matching in some form in addition to login-based matching. In short, the level of accuracy is a very important consideration because it's reaching the same users across multiple devices that has such a great impact on results.
Leverage rich behavior and interest insight across all devices and channels
The availability and depth of user interest data to target individuals often varies significantly by channel and device. You can, for example, get really granular in audience development in PC-based display, but interest-based targeting in mobile is often more rudimentary. One key reason is that pure-play mobile vendors see less of a user's total digital activity.
But they do have some really valuable information that pure-play PC vendors don't have: rich, reliable understanding of the mobile behavior of a person. And you need both.
In my view, doing cross-device right requires use of a profile that has both rich interest and behavior data and sound device-usage insights. Interest-based targeting should be just as rich for tablet and phone users as PC users. That requires great user-to-device matching combined with deep audience behavior and interest data.
Look for three kinds of cross-device scale
Whether your brand is large or small, make sure that you field cross-device programs with the scale necessary to really move the needle.
In cross-device, you need to pay attention to three kinds of scale:
- Number of cross-device profiles: How many cross-device profiles can be reached? Potential cross-device reach varies quite widely.
- Number of data points per profile: You can ignore this if you are truly a mass marketer. But if your targeting is at all granular, make sure that your partner can provide enough profiles at the necessary level of specificity. Note also that unique data points are more valuable than common ones.
- Number of potential cross-device ad opportunities: Bigger inventory footprints increase the odds that appropriate inventory can be associated with a cross-device profile so that more impressions can be delivered.
Note also that scale is in many cases inversely related to profile accuracy. To some degree, higher accuracy standards may reduce the number of profiles available. It's like any QA standard -- the more precise the spec, the fewer things that will pass it. The implication here is to inquire about the accuracy standard and ensure it lives up to your needs.
One other thing to be aware of is that scale is constantly increasing for many vendors in the industry. Today a partner might have 30 million profiles, next month 34 million. Just make sure that your expectations match the scale they can offer when you sign that IO.
Engage using the interactive strengths of each device
This is where I'm going to lose some of you. We have to make a real effort to capitalize on the characteristics of the medium if we are to maximize cross-device campaign results. Recognizing that standard units are easier to execute and sometimes key to delivering an acceptable overall CPM, we also need to identify and include more engaging and interactive units in our programs.
Many people choose mobile programs, for example, made up almost entirely of 300x50s, those little slivers at the bottom of your iPhone screen. Reported interaction and click rates for such units are rather high -- but much of that performance probably relates more to fat-finger syndrome than actual noticing value.
Make units with real stopping power a centerpiece of your programs. Add multiple ways to interact. Give people a choice of CTAs. These are all good best creative practices -- but they are arguably even more important to maximizing cross-device performance metrics because they increase the likelihood that they will actually see and experience the multiplicative effects of brand exposure on multiple screens.
Make learning and insights a priority
Cross-device is not something you can set and forget. All industry stakeholders are learning and improving their approaches -- and results -- every day. Set learning objectives with every campaign. What additional cross-device target insight can I learn from this effort?
Genuine multi-device behavioral understanding is in its infancy. Help your baby grow. Your programs can progressively reveal unique aspects of your audience's unique cross-device behaviors. They can provide perspective that can also help guide strategies in other areas of marketing. Ensure that you the partner or internal analytical resources necessary to determine more than your campaign CTR and CPA.
For years many marketers have approached mobile and tablet advertising with a combination of facts and hunches. Let's make a concerted effort to minimize the hunches in 2013. As you formulate your cross-device programs, ask the questions necessary for you to identify the strategies that can really meet your needs. Lots of people are getting great results from cross-device marketing, but the best results are going to go to those that focus on the essentials as they formulate their action plans.
On Twitter? Follow iMedia Connection at @iMediaTweet.
"Vector illustration of several multimedia devices," "Perfect smiles," "Smart phone concept with beautiful blonde woman," "Man's head falling to bits," "A locked smartphone,"and "Computer keyboard e-learning concept" images via Shutterstock.