February 20, 2008

Multichannel Forensics: PPC, SEO and Online Marketing Example

Please click on the images to enlarge them (you'll need to do this!).

From time to time, I'm asked to describe how Multichannel Forensics can be used by online marketers.

Honestly, the methodology is ideally suited for online marketers ... as long as the online marketer stores key information in a centralized customer database.

Take a look at this online ecosystem. The online marketer tracks source of order in the customer database, categorized as follows:
  • Google, Paid Search
  • Google, Natural Search
  • Yahoo!, Paid Search
  • Yahoo!, Natural Search
  • MSN, Paid Search
  • MSN, Natural Search
  • Portal Advertising, All Sources
  • Shopping Comparison Sites, All Sources
  • Affiliate Marketing, All Sources
  • E-Mail Marketing, All Campaigns
  • Catalog Marketing via Catalog Key Code
  • All Other Online Marketing Sources
  • Organic Online Orders, No Marketing Attribution
Your first instinct might be to nitpick the categorizations, so feel free to use whatever categorizations make the most sense for your business when you apply the methodology. My purpose is to simply illustrate how the methodology might be used.

The second image shows the Migration Probability Table. This illustrates how customers who purchased by various sources last year migrated to different purchase sources this year.

The first image maps the ecosystem.

What can we learn from this analysis? Plenty!
  • Google matters. It is a primary source for new customers. Customers who evolve to loyal status migrate from a newly acquired customer via Google to a Natural Search customer to finally placing Organic Online Orders. In fact, Google has a disproportionate influence over the direction of the brand, as other forms of advertising eventually feed back into Google.
  • Customers are slowly migrating from Yahoo! and MSN to Google. Some customers who used to use Yahoo! and MSN to place orders last year now use Google this year. The Multichannel Forensics analyst should monitor this trend over time, leveraging it in the development of paid search budgets.
  • Yahoo! and MSN purchasers migrate to the E-Mail Marketing channel. The E-Mail services offered by Yahoo! and Microsoft allow a unique multichannel element to occur. In essence, Yahoo! and Microsoft benefit because the customer shifts from purchasing via search to purchasing via E-Mail ... allowing Yahoo! and Microsoft to continue to stay active in the customer relationship with this brand.
  • Customers who respond to Portal Advertising migrate to E-Mail and Catalog Marketing. Obviously, there is a subset of this customer base that takes matters into their own hands (search), and there is a subset responsive to advertising (Portals, E-Mail, Catalog).
  • Shopping Comparison and Affiliate Marketing customers are least loyal in this example. These customers are apparently getting their needs met on a one-time basis, unlikely to shop again.
  • Once a customer purchases without the aid of advertising or search, the customer is the most loyal of all customer types. These customers appear to order because the like the brand, not because they need search or marketing to drive them to the site. This is what a brand ultimately wants. However, the metrics indicate that these customers still use Natural Search via Google results to place orders in the future, clearly illustrating the importance of SEO, even among best customers.
  • The most loyal customers are those who order without advertising, as well as those who order via E-Mail or Catalog Marketing.
  • There is a hurdle for this brand to get over. Recall the disproportionate influence Google has over this brand. Notice that Google responders are not as loyal as are Organic, E-Mail or Catalog Marketing responders. In other words, an over-dependence upon Google results in a less loyal customer base. There is an opportunity to try to convert Google shoppers to other forms of advertising.
If the proper information can be stored in the customer database, really interesting Multichannel Forensics projects can be constructed, limited only by your imagination.

An analysis of this nature also illustrates the limitations of web analytics, where metrics are configured to allow analysts to focus on what a customer does within a single visit. An entire generation of professionals are being trained to believe that the customer relationship is best measured within a single visit. We need well-rounded web analytics professionals who understand how customer relationships evolve over time, folks who can measure the complex relationship between customers, advertising, products, brands and channels. This is the promise of Multichannel Forensics.

3 comments:

  1. Anonymous12:05 PM

    Thank you, Kevin!

    ReplyDelete
  2. Very innovative! You are right that this is the kind of analysis that is way too rarely done in web analytics. As web analysts we are guilty of assigning too much credit to the most recent clickthrough to the site. The history of earlier touch points is too often ignored.

    One could maybe point out that web analysts are not just focused on a single visit anymore. By now, they are capable of assigning the business from delayed conversions during subsequent visits back to the original ad/search clickthrough. Most web analysts can also sum the total business that follows over time.

    But the Multichannel Forensics method of showing the migration paths of customers from channel to channel is absolutely exciting and unique.

    The only other work in this direction that I know of have been studies that investigate the impact of online ad impressions on search. I put some shortcuts to a few of these studies up here: multichannelmetrics.com/DisplayVsSearch.

    Thanks for another great example of Multichannel Forensics!
    Akin

    ReplyDelete
  3. It is true that web analytics tools can focus on more than a single visit. When there's more cross-pollination between web analytics and the SAS folks who've been analyzing customer behavior for the past few decades, then we'll hopefully see better time-based insights.

    ReplyDelete