Dayparts, Day of Week, and Other Cherry-Picking Techniques

With the advent of preset dayparting, the average Google advertiser now has just a bit more of the power an API (define)-driven campaign management solution possesses. At the grocery store, we all pick the best cherries from the pile. In SEM (define), we’d love to have a similar option. Dayparting is one of the most powerful methods of cherry-picking the best clicks out of a clickstream.

In this column, I’ll cover Google’s newly released ad scheduling feature for AdWords and how to use MSN adCenter’s powerful “incremental pricing for targeting” feature. Because dayparting and other methods of cherry-picking clicks are based on a sound strategic foundation, we’ll discuss not just the how, but the why.

When it’s time to execute more sophisticated dayparting and cherry-picking strategies, you’ll likely need to look at one of the more robust API-driven campaign management solutions or consider deploying some extra labor to both analyze data and generate the calculations necessary to carry out your strategy. For example, one of the more powerful ways API-connected technology and labor can be used to extend the cherry-picking concept further, across all search engines, is through geographic campaign resegmentation. Geographic overlays, combined with dayparting, can really be powerful, as we’ll see.

Why should you cherry-pick? The kinds of visitors you’re likely to get at your site based on time of day or geography are quite different. Visitors may differ by a variety of critically important factors, including:

• Average age
• Gender
• Household income
• Surfing location: from work (broadband) or home (possibly dial-up)
• Stage in the buying cycle (ready to buy or researching in a leisurely fashion)
• Computer speed, operating system, and screen resolution
• Psychographic or psychological profile (personas, lifestyle segment)
• Geekiness (I like this one because it indicates the level of comfort the visitor has with technology, computers and the Internet. The more geeky person may be better at intuiting information from a less-than-perfect user interface)

All these factors add up to one thing: your predicted conversion based on geography and time of day will be dramatically different. If a campaign is national and the national average conversion rate for a week’s worth of traffic (e.g., a typical amount of traffic measured in a Web analytics package) converts at 3 percent, we can index that average conversion to 100 percent. Then by measuring the clicks a different way, segmented by time of day in our time zone, we may find the following pattern:

Time	Conversion Rate (%)	Index (Based on 100)
Midnight to 5 a.m.	2.1	70
5 a.m. to 10 a.m.	2.7	90
10 a.m. to 3 p.m.	3.5	117
3 p.m. to 8 p.m.	3.2	107
8 p.m. to midnight	2.9	97

It’s important to note the index numbers don’t average out to 100 because the traffic levels at different times of the day differ. It’s also important to note that if this analysis were done on a nationwide U.S.-based campaign, the data would be a bit fuzzier than we’d like because there would be four time zones represented. We don’t know the full impact of the time zones, but even based on mushy data there’s a huge swing.

By using the index data, you can estimate how much more or less than your current bid you’d be willing to pay. This is killer because a 17 percent bid increase may mean several hops up in position for many keywords. This kind of bid tuning is particularly powerful when your average position number is three or less. The bump in position may well put you into the hot zone of high traffic areas, right at a time you can afford to be there.

An even more in-depth analysis of your data by IP address cluster (by time zone) will likely yield an even stronger relationship between conversion and time of day. While you’re appending geography data to your click and conversion data (assuming you use a technology that facilitates such appending), why not segment clicks and conversions by geography even more tightly than time zone? If you have a large campaign, you should be able to group clicks by designated marketing area (DMA).

You’ll be able to identify the geography of most clicks, but clearly some clicks will have an ambiguous geographic location, such as those from AOL’s proxy servers in Virginia, which cover the entire nation of subscribers. But for those clicks you can identify by geography, the same indexing analysis can be performed. Some cities may index significantly higher than others. This information allows you to further cherry-pick.

Use geographic index data to restructure your campaign into common index groups that share the same time zone. For example, if Chicago and Dallas both index at 115, then a Google campaign could be set up to specifically target those cities. The dayparting information could be added to the city index, resulting in a cumulative index of well above 130. That’s a 30 percent higher bid possibility.

Use some important high-volume keywords that are running at middle positions as a test bed for your cherry-picking strategy. Keywords aren’t the only way to target in Google and MSN. Dayparting and geography are powerful filters, too.

If this seems like a lot of work, it is. Like it or not, SEM is getting more mathematical. If you feel as if the competition’s winning but can’t put your finger on how, chances are they’re tapping the power of cherry-picking, leaving the sour fruit for you to pick over. Wouldn’t you rather have the plump, firm, juicy, and delicious clicks for yourself?

Meet Kevin at Search Engine Strategies in San Jose, August 7-10, 2006, at the San Jose McEnery Convention Center.

Want more search information? ClickZ SEM Archives contain all our search columns, organized by topic.