Adrian Scott . com

 Press Coverage
 Past Clients

 Web of Finance
Scientific Internet Marketing

by Adrian Scott . com

The key to effective marketing on the Internet is the ability to test and measure marketing initiatives. This approach has proven its value in direct marketing and database marketing and can be easily extended to the Internet.

Scientific Internet Marketing can be used to test, evaluate and fine-tune initiatives such as Web advertising, marketing Web sites, and e-mailouts. In this article, we look at how you can implement it at various levels in your Internet marketing efforts.

Most Internet organizations are currently at levels 1 and 2, Eyeballs and results. Major corporations, such as financial services firm Charles Schwab & Co. and technology company Hewlett-Packard are currently developing more advanced levels of Internet marketing, though these new developments are only starting to be released on the Internet. (In many cases, they are released for an "extranet", a special private Web area for special customers or prospects.)

Level 1: Eyeballs

The first level of scientific marketing is the equivalent of measuring eyeballs, analyzing how many people click on a Web ad or how many people visit a Web site.

Ad click-through is generally reported by publishers selling the ads. Basic statistics on Web site visitors can be gathered by analyzing Web log files using Unix shareware software, or downloadable PC log analysis software (such as Net.Genesis' Net.Analysis).

Level 2: Results

The second level is results-oriented. You start off with a goal to measure your initiative with, such as "getting new subscribers" for a subscription Web site. When visitors enter your site as a results of your marketing initiative, you tag them with a unique identification number (using cookie technology) and store that ID number with the information in your list of people who came in through that initiative.

Then, when visitors subscribe, you record the ID number. In this way, you can find out what your conversion rate is for the marketing initiative -- what percentage of respondents become subscribers. The nice thing about cookie technology is that the ID tag will still be on the visitor even if it several days between when they respond to the marketing initiative and when they purchase a subscription.

Example: Jerry's Online BubbleGum Store and Web ads

Jerry has a (fictional) Web site where people can purchase bubblegum online. Jerry creates a co-marketing relationship with Lucy's Online Chocolate Store, so that her site has a link to his site. Jerry would like to know how many new orders he is getting from this co-marketing relationship, so he can figure out if he should extend the relationship, create more co-marketing relationships with other sites, or focus on other marketing initiatives.

Jerry has Lucy make the link to a special URL at his site, The URL is a CGI program that assigns a unique ID using a cookie, and stores the ID in a list. Then Jerry modifies the order-taking CGI program on his site to capture the ID number on all orders that are placed. After these two changes, Jerry waits three days and then does a comparison of these two lists to find out how many people who clicked through the link have later actually purchased bubblegum.

He finds (for instance) that 40 people have clicked through the link (been issued ID numbers), and 6 of these people have made orders. Thus he has a 15% conversion rate.

Jerry can even take this to another level, by looking at the dollar amounts of these orders, and evaluating the effectiveness not just by conversion rate but also by dollar value.

With this system in place Jerry can also make a deal with Lucy to give her 5% of any orders that are generated by referrals from her site.

Level 3: Testing

With level 2, Jerry had a way to measure the results of his co-marketing initiative, but no easy way to fine-tune how the co-marketing link worked.

Level 3 adds on to levels 1 and 2 by providing marketers with the ability to test and fine-tune their messages. For instance, a co-marketing Web link might be a graphic image (e.g. a GIF). The graphic image might include a headline. It's quite possible that slight changes in the headline wording or appearance could have significant impact onto its effectiveness.

Alternatively, a marketer might want to experiment with the jump page in an online ad campaign, the page that people go to when they click-through on an ad. There are so many things that can be tested, such as the number and kind of links on it, the headline, the amount and kind of graphics, etc.

These kinds of details can be tested by presenting two creative executions on a random basis and tracking results. For instance, for jump page testing, when a person clicks through the ad, they are randomly presented with one of the two possible pages. In addition, they are given a unique ID (as discussed in level 2) and the Web server also records the ID number along with with of the

2 pages the visitor was presented. Then, at the result page, such as an order form or other sign-up, the ID is recorded again.

Then the marketer can analyze which of the two creative executions was more effective looking at which generated more orders or sign-ups.

Example: Pro Piano Online Magazine and an E-Mailout

These methods can be applied to both Web and e-mail initiatives. For instance, a fictional online magazine we'll call Pro Piano has a list of e-mail addresses of people that might be interested in it. Pro Piano wants to send an electronic mail-out to people on this list, but would like to test and fine-tune the wording and presentation of the e-mail before sending it out to everyone on the list.

Pro Piano creates two versions of the e-mail, one which places an emphasis on their free online library of MIDI files and another which focuses on their articles. Then a small sample of the original e-list is selected and randomly divided in two. Each of the two e-mail creatives contains a different URL

for readers to go to.

The URLs mentioned in the e-mail go to CGI programs which do ID tagging as discussed above, plus record which of the two creatives the person had received.

After the sample mailouts are sent out and the results are collected from the CGI programs' log files, they are analyzed. Pro Piano can then find out, for instance, that the e-mail emphasizing free MIDI files generates more site visits, though fewer new subscriptions than the e-mail emphasizing articles.

Note that though we mentioned using cookies in the above example, it wasn't totally necessary if the subscription form had a field for e-mail addresses. We could keep a list of which creative was sent to each e-mail address,

and then cross-reference that with a list of e-mail addresses entered in the subscription form.

Level 4: The Database

You may have noticed that in the previous example we didn't really need to assign a unique ID to each visitor. We could have only tagged them with the information that they were using creative #1 or #2. Similarly, in Level 2 we only needed to tag the visitor with the information that they enetered through

the co-marketing link, rather than a unique ID.

However, if you do go ahead and assign unique ID's, you can gradually build up a database of information about each site visitor. This database can record what parts of the site the visitor visits, where they enter the site from, what they enter in forms such as surveys and sign ups, and other information. You can use this information to market on a one-to-one basis, presenting personalized offers aimed at the visitor's interests and needs.

Example: Mega-Big Computer Corporation home page

While looking for laserjet printers, John visits Mega-Big Computer Corporation's home page. It takes 5 clicks for John to find the laserjet printer area, a bit of a hassle. Then he looks for their new 3D game software, and finally finds it. Two days later, John returns to check the printer specs again.

This time the home page has two links at the top-right corner, one to "Laserjet Printers" and one to "3D Games". In the middle of the page is scrolling text which mentions a special offer on laser printers for customers who order within 24 hours, and invites John to click for an order form.

Database Technology

In level 4, I introduced the concept of using a database. For major organizations with serious operations, a standard business database such as Oracle, Sybase or Informix is needed. These can be interfaced to the Internet using Perl/SQL or Java Database Connectivity (JDBC). At the same time, a small site or small organization can get started without a full-blown database system using, for instance Unix's DBM files and Perl for CGI programming. DBM files can be used for randomly accessing and modifying data, so they are very useful in creating a level 4 marketing iniative.

The Next Level

What level is your organization at and where do you want to be next? What's beyond the Database level?

The Web gives us the potential to have a whole sales dialogue with visitors, qualifying prospects and selling them products and services. It also gives us the capability to carry on a relationship with customers using a memory, so that return visitors are welcomed as such and given special service.

Companies with experience in direct marketing will quickly move towards these techniques of scientific Internet marketing. These techniques are applicable to all companies marketing over the Internet.

Adrian Scott, Ph.D. ( is CEO and founder of Aereal Inc. ( Aereal delivers Internet business solutions combining marketing and technology for finance and high-tech corporations including Bank of America, Charles Schwab & Co., Symantec, Hewlett-Packard, Technology Funding, (Electric Classifieds) and GMO Advertising.


URL is uniform resource locator, which is essentially the address for a Web page.

CGI is common gateway interface, which lets programs run on a Web server to generate a Web page and do other functions like processing forms.

GIF, graphic image format, is a commonly-used format for graphics on the Internet.

Cookies are a Web technology that lets a Web site store a piece of information on a user's hard disk that can be accessed by the Web site on subsequent visits of the user to that Web site.

Adrian Scott . com is CEO and founder of Aereal Inc., which delivers Internet business solutions leveraging technology and marketing. Aereal's clients include Bank of America, Charles Schwab & Co.,, Technology Funding, Symantec and Hewlett-Packard. Adrian Scott . com is a contributing author to 6 Internet books, including Web Publishing Unleashed and World Wide Web Unleashed from, and has written articles for periodicals including NetscapeWorld, Java Report and He has performed with the New York City Ballet, sung with Placido Domingo, and graduated from Rensselaer Polytechnic Institute at age 16.

Copyright 1997 Aereal Inc. All Rights Reserved.