Landing pages are a special kind of web page with unique design and content requirements. Critically, landing pages live in a parallel universe and should be designed to maximize the behavior you want. Site navigation should not be included. Either/or choices should be avoided. Instead, focus on the one behavior you want. A best practice is to develop one landing page for every search term or phrase that makes up your SEM campaign. In our own testing, we’ve found that developing unique landing pages for each search engine phrase will significantly conversions.

Dos and Don’ts When Building Landing Pages

Relevant Links

Marketing Sherpa
See especially this report on landing-page optimization from February 2005 available for purchase¹

Marketing Experiments
See especially this research note that summarizes much of the work done at this site on optimization of landing pages. Includes a list of vendors such as Offermatica who specialize in developing and hosting of A/B and multivariate testing of landing pages²

Notes

1. Available for purchase for $249↩
2. Membership at this site is free but detailed results on research may require research sponsorship↩

It always annoys me no end, the online world claims analytics and ROI as if the concepts were invented for online world alone. In point of fact, analytics and ROI for marketing purposes were invented “back in the day” by direct marketers like myself and can be applied online and offline. The popularity of online analytics has led many C-level executives to expect the same level of analytics to be available for campaigns fielded offline. Here in lies the rub. It’s almost impossible to deliver real-time analytics to campaigns fielded offline. Unless you think through the analytics around the backend of your campaign in depth from the get go and are prepared to innovate here and make a significant investment in technology, much as the early pioneers in web marketing innovated and invested.

If you are embarking upon this journey, here are some URLs that can help (in no particular order):

• Web Analytics Demystified Maintained by an analyst at Jupiter Research who follows the category for a living. If you have time for only one blog on the subject, this is probably the one. See also his books available for purchase and his sideblog listing of additional blogs that follow the web-analytics space to check out.
• Web Analytics Association While I could wish that this was the Marketing Analytics Association, much of what is here will give you a running start at both offline and online analytics.
• Fireclick Tracks conversion rates and abandon rates by category. Despite the similarity in names, this firm has no relationship to Firewhite.
• Inside Analytics Forward coverage of new products and industry trends, particularly Web 2.0 and how it will impact on web analytics.
• Xavier Casanova’s Blog AKA as the founder of Fireclick. Xaviar may have a new gig but on his blog he still raps poetic about what is happening in his old arena.
• The Dashboard Spy Ever wondered what was on an enterprise’s marketing dashboard? The dashboard spy may be able to show you.

Common tools used by professionals for different purposes:

General Statistical Packages

• SPSS
• SAS
• R Open Source version
• S-Plus … Advanced Statistics for optimization Open Source version | Commercial version

Alternatives to the “Usual Suspects” listed above

• Data Desk Easy to Use
• Stata Time Series
• Minitab

Special Purpose

• Comprehensive Meta-Analysis to summarize and reach conclusions from multiple studies on the same subject.
• Design Expert used for design of experiments from a company called Stat Ease. 
• Power and Precision sample size
• StatXact nonparametric analysis, exact analysis for categorical data, and exact sample size and power.
• XLMiner surprisingly powerful Excel add-in

Multi-variate models
Identifies the relationship between a group of explanatory variables (eg, billing, call center operations, on-site service or employee performance) and the variable of interest (eg, profitability by customer segment)

Market segmentation
Divides the market into groups that are like-minded enough that all customers in a given segment can be marketed to as if they were a single entity. The methodology of choice for segmentation is cluster analysis or discriminate analysis.

Conjoint and discrete-choice models
Models customer choice behavior or the trade-offs customers will make between different product features, including price

Structural-equation models
A powerful, visually represented, multivariate analysis technique that combines factor analysis and regression to study direct and indirect relationships between variables of interest (eg, customer characteristics and experiences and perceptions of the value of their relationship with an energy supplier)

Time-series forecasting
Identifies trends or patterns in data over time and predicts or forecasts future values (e.g., forecasting the need for waste water treatment plant capacity based on predicted water use)

Log-linear models
A sophisticated way to analyze cross-classification tables and test interactions between variables for statistical significance (eg, the test for the existence of a disparate impact of airline baggage screening procedures on selected minority groups in the passenger population)

Survival analysis
Focuses on time as the variable of interest, such as the study of the ‘survival’ i.e. retention rate of particular customers.

Case Note
This is a redacted results analysis for a two-step direct mailing program for Votive, a maker of high- end audio and video equipment. The client–Votive–is fictional.

Step 1 asks Votive’s VARs (value added resellers) to identify themselves as being interested in sponsoring end user seminars in the local market. The end user seminars—Step 2—were designed to create demand for specific Votive products. Results are measured based on the number of VAR requests for kits, number of seminars executed by VARs, and the incremental lift in end user demand.

• Background
• Summary
• Regional Performance
• Prospects vs. Customers
• 2002 vs. 2003 Performance
• List Performance
• Next Steps

Background

In February, we mailed a direct mail solicitation to 20,000 VARs inviting them to participate in our end user seminar program. The call-to-action instructed VARS to go to a special URL and order the kit for fulfillment by return mail. Inside the mailing were the following elements:

• Seminar kit The kit was designed as a “just add water” seminar kit and included a video, ideas on how the VAR could customize the seminar, signage, a high-quality video for use at the seminar itself, plus copies of the direct mail invitation.
• Direct mail to end users Each kit included a sample direct mail invitation as well as information on how the VAR could order invitations and a mailing list.
• Incentive to participate Each VAR was given a sell through goal that was based on their historical run rate. To qualify for the trip, they needed to grow our business by 20% and show proof of performance i.e. that they had fielded the seminars with at least 30 end users.
• Closed loop measurement system End users checked into the seminar online, using a specially developed web application that carried the VARs branding but was hosted by us. In this way, we collected the information we needed to keep in contact with these prospects and also for results measurement.

Summary of Results

This program delivered sales of $2.9 million and an ROI of 2.9 on a contribution margin basis. Based on sell through data after 6 months, each dollar invested in the program generated $6 in revenue and almost $3.85 in contribution margin. This program reached breakeven with two months of launch. Based on performance to date, we recommend that this program be continued in 2004, enabling us to further improve performance based on findings to date. Summary of results obtained Kits requested: Of the 20,000 VARs solicited, 8.4% requested seminar kits, for a total of 1,676 kits requested.

Seminars fielded
Of the 1,676 VARs who requested kits, 704 went on to field seminars with end users, representing a participation rate of 42%. Of this number, 570 of the VARs participated in the same program last year. The remaining were either re-activations or new participants.

Number of end users
The seminars attracted 26,640 end users. Participation varied depending on whether the end user was on our files as an existing customer or was completely new to Votive. In terms of attendees at the seminars, 62% percent were customers, the remainder prospects. The response rate to seminar communications was 2.0% for existing customers and 2.1% for prospects. Overall prospects participation levels are unexpectedly high.

Regional Performance

As expected, the Midwest and South performed better on an ROI basis than the other regions given the traditional higher propensity to buy Votive products in those regions (NW, NE, and Pacific has a higher purchase incidence than South). Pacific lagged in performance, attributable to a misfit between the Votive styling and positioning compared to its major regional competitor. This misfit is highlighted by the relatively small number of units sold.

Prospects vs. Customer Performance

Overall, prospects spend more than customers as they have to buy and end-to-end solution, rather than individual modules. Purchase incidence varies across the two groups and across regions, though prospects buy at respectable rates in all regions except Pacific. Only 7% of prospects who attended seminars purchased Votive’s products, less than half the rate of the next poorest performer. The poor showing generated the only negative ROI in the program. However, existing Pacific customers participated in the seminars at an above average rate. Areas with higher prospect ROI are likely areas to invest greater levels of marketing spend going forward, while markets where prospect ROI are danger zones that require further research. Votive is particularly strong with prospects in the Midwest and posts strong performance with them in the South and Northeast.

The Pacific is not as strong as other regions. We were not able to achieve breakeven on our investment when it came to new prospects (ROI=0.7 with ROI=1.0 required for breakeven). Next year, we need to give the Pacific region more guidance on who they should mail to in order to get to breakeven.

2002 vs. 2003 Performance

Votive fielded a similar program in 2002, but with data collection as results tracking limited to aggregate performance.

Overall program performance improved for the following reasons:

• Excluded from the kit mailing 6,000 of the smaller and least profitable VARs, who as a group did not execute credibly on the program in 2002
• Lowered the cost of the package mailed to the VARs based on their feedback that many of the in-store promotional materials did not drive attendance
• Targeted prospects using externally purchased lists, not only existing customers as we did in 2002
• Selected a higher priced, higher margin item to market

Together, the above actions increased the program ROI by 58%.

List Performance

To support the VARs in their seminar efforts, we continued the following sixteen (16) lists which reached new prospects as opposed to customers and invited them to a local seminar in their area. All the lists had a ROI of 1.0x – meaning all the lists gave a positive return on our investment. Similar to previous catalogs, the audio enthusiast magazines and new homeowner lists were the best performers. The best performing list, List A, had a purchase rate 0.6% and a ROI of 6.2x. The 2nd best performing list, List B, had a purchase incident rate of 0.57% and a ROI of 4.2x. Given the Votive product set, the type of home theater buyers reached through List Y are likely targets. The worst performing list, List P, still managed a respectable ROI of 1.5x. The more general readership of this magazine accounts for its weaker performance than the other lists.

Next Steps

Overall the program drove significant demand and generated a strong ROI. We recommend continuing the program in 2004. For the next wave, we recommend designing in the following test cells to better inform our marketing efforts:

• Gender
• Years at present address
• Age

Based on other data available from your team and third parties, we believe that buyers of your products skew male. We also believe that people wither moving to a new home or who have been in their homes for 6+ years are much more likely to buy. The replacement cycle for your products is 6-9 years, and often people buy in your category when they move to a new residence. In addition, we suspect that the younger average in Pacific may be a key driver in the relatively poor performance of the region. By creating test cells for these items (assuming we can get cells large enough to be statistically significant), we will be able to learn more about the key drivers of success of this two-step program.

Though we believe that continuing to target new customers in Pacific will not generate a positive ROI, we recommend that the effort be continued and that you conduct quantitative research to understand the drivers of this performance.

A system that enables you to understand how a particular customer responds to your marketing initiative. The hallmark of these systems is the ability for the customer to respond via whatever mechanism they prefer: by clicking through to a special landing page online, by calling an 800# or the equivalent, or by filling out a response card and mailing it to your company. Once the customer responds, systems at your company should give you the ability to track where that customer is in the sales pipeline.

Here is a top-down view of a typical system.

The testing hierarchy is a visual diagram that represents the campaign by target segment and details what you plan to test within each target segment. This type of visual can be very helpful, especially when planning a complicated campaign that reaches multiple segments and/or touches the target multiple times.

Key
DR = direct response
DM = direct “snail” mail
EM = email

Click here to see an example of a test matrix.

Sample Size Calculator

Confidence Interval