THE RKGBLOG

Dayparting by Device: The Best Times to Reach Those New iPad Users

With today’s release of the new iPad, marketers can expect to see an appreciable bump in tablet traffic as consumers who pre-ordered early enough will get them in-hand, while countless other shoppers will try their luck at retail stores around the country. We saw such a traffic spike shortly after this past Christmas, with tablets surging to nearly 8% of paid search traffic, nearly double their share of just a couple of months earlier.   Apple has seen its tablet dominance slip a bit from its peak, but they still dwarf the competition in traffic, and they are likely to add a little cushion back to their lead this week.

While tablets may or may not be ushering in the “post-PC” era — if it’s not here already — the reality is that user behavior suggests tablets and desktops are not neatly interchangeable, neither in the minds of users nor in how marketers should address them strategically.  Previously, we showed how traffic patterns throughout the day and week vary significantly by device type, with tablet usage peaking between 9 and 11pm and over the weekend.  We’ve also shown how there are huge differences in ad conversion rates both between device classes and even within those classes.  So, where do we go from here?

First, if you’re not already segmenting your ad campaigns by device on account of the above, you probably should be.  We also know that the value of the ad traffic we generate can vary depending on the time of the click, so hopefully, you’ve also assessed your ad conversion potential by week and/or daypart and are implementing that information into your ad bidding, as well.  But, can we just assume our dayparts are appropriate across all devices?

Hourly Click Trends by Device

As a refresher, let’s go back to how traffic breaks down by device throughout the day.  Since desktop still greatly outpaces mobile traffic, we present traffic from each device class that is normalized against the average hourly level for that class.  This helps show when traffic is relatively high or low for each device:

Again we see clear differences in when each device type is most likely to generate traffic.  Not surprisingly, the iPad and Other Tablet segments show very similar patterns throughout the day and both peak when many on the East Coast are going to bed and those on the West Coast are home from work or school.  Desktop hits a peak at about the same point, but it is far less pronounced.  Desktop traffic is strongest, compared to the other devices, during the workday.

How Revenue Per Click Differs Across Devices

Now let’s take a look at how the revenue we generate for each ad click varies during the course of the day.  Note that this is a different perspective, where each device class is benchmarked not against itself, as above, but to the average for the program as a whole.  This highlights the wide gulf between the iPad/Desktop segments and the Phone/Other Tablet segments:

So, for example, desktop clicks occurring during the 6am Eastern Time hour generate a revenue per click (RPC) that is about average for the program as a whole over the course of the day, while the value of phone clicks is about a fourth that.

We have a couple of concerns here:

1)  Given that traffic trends differently by device, are the overall differences in RPC we see throughout the day — and, possibly, the dayparting data we are using for bidding –  skewed by the shifting share between better and worse performing devices?

2)  How should we be determining the correct dayparts and can or should we apply them equally across all devices?

On the first question, we’ll see below from isolating our largest segment, desktop, against overall RPC, that the net effect of using overall data to inform dayparting for desktop is to suggest bids be brought down more than they should during poorer performing periods and not pushed enough during better times.  For example, 10am desktop clicks generate a 20% higher than the daily overall average RPC, but our overall RPC at that specific time is only 17% higher.

At least for desktop though, this gap is pretty consistent throughout the day:

Even though the relative traffic levels of the other segments can shift significantly from hour to hour, those levels are not large enough to introduce appreciable spikes one way or the other in our overall RPC, which is dominated by desktop data.

Still, the short of it is that if we are relying on overall RPC trends to determine dayparting, it will likely not deliver the appropriate bid pushes and pullbacks.

Isolating Devices for Dayparting Calculations

Instead, if we are already segmenting devices out in our ad campaigns, we should be isolating them in how we determine our daypart bidding adjustments as well.  The correct comparison is to each device’s own average RPC,  as we’ve done below:

This view accentuates some of the noise in the data, particularly for Phones and Other Tablets, but it does give us a better view of when ad clicks from each device class are more valuable and to what extent.

Looking again at the 10am hour, we find that desktop revenue per click is 22% higher than the average desktop RPC for the rest of the day.  At the same time, iPad RPC is only about 7% higher than it normally is.  Again, our overall RPC was 20% higher at 10am, which, had we not isolated the iPad, would imply a far greater dayparting bid push for it then we’d want.

Also of note, the value of phone traffic takes an appreciable dive during the peak evening drive time on the East Coast.  We don’t see the same effect with the iPad or even the Other Tablet segment, which adds some additional weight to the argument that they are closer in affiliation to desktops than phones.

More generally, while the 10,000 foot view of revenue per click by device may seem to show fairly similar patterns throughout the day, there are some gains to be made from digging into the details.  For a well-established program that is otherwise firing on all cylinders, getting desktop bids 5% more accurate — or adjusting iPad bids even more — can make a real difference.

Technorati Tags:

  • Mark Ballard
    Mark Ballard is Director of Research at RKG.
  • Comments
    6 Responses to “Dayparting by Device: The Best Times to Reach Those New iPad Users”
    1. Davd Smith says:

      wow, good stuff! analytics at it’s best.

    2. Terry Whalen says:

      Hi Mark,

      I think you might get even better optimization if you also take into account average CPC rather than just RPC. RPC may go up and down, but if average CPCs go up and down in line with RPCs, then you may not have a benefit from dayparting. Maybe average ROAS-per-click would be a good metric, since it takes the revenue and the cost into account.

      Thoughts?

    3. Mark Ballard Mark Ballard says:

      A good point, Terry. If the competitive landscape shifts dramatically throughout the day, you could find the CPC you pay changes even if your bid does not. And, generally, an ROAS that exhibits very spiky behavior over time, whether it’s over the course of the day or over several months, is a bad sign. RPC itself isn’t an entirely independent variable – bidding higher will draw more lower quality search network traffic, for one – but, if we’re doing a good job, I think it’s a better reflection of the potential for dayparting than ROAS would be. We are already pushing and pulling back on bids throughout the day, so a view of our ROAS change is going to be more muted, if not flat.

    Trackbacks
    Check out what others are saying...
    1. [...] Shared Dayparting by Device: The Best Times to Reach Those New iPad Users | RKG Blog. [...]

    2. [...] Mark Ballard over at the Rimm-Kaufman Group has a great article which shows charts breaking down PPC stats by device. It is interesting to see how vastly different the data is even for the iPhone, iPad and other [...]

    3. [...] or time-of-day. The problem is that the same schedule is applied across all devices, even if RPC (revenue per click) trends are different. Whereas before, you could make separate campaigns for desktop and mobile, setting different [...]