Search Engine Marketing: How to Successfully Measure Your Web Site Analytics, Part 2

Facebook
Twitter
LinkedIn
Telegram
WhatsApp
Reddit
Email

With the luxury of data in your hands, your ability to quantify your recommendations will set your projects apart, and move them to the top of the queue. This can be fairly simple. Whether focusing on leads, purchases, call avoidance, click-throughs, downloads or any other type of goal, give the achievement of the goal-or conversion-an assumed value. For example, assume that a lead generated on your site is worth $100 to your business and that there is a lead form that generates 50,000 visits per month. Over the last few months, conversion rates have been relatively steady at about 18%. Given the assumed lead value, the form is generating $900,000 per month. Through targeted campaigns, navigation adjustments, form design changes and testing, you estimate that conversion can be increased to as high as 23%. The 5% increase in conversions results in an additional $250,000 per month or $3,000,000 per year. Now this project can be assigned a priority. These recommendations should take precedence over any project not anticipated to generate $250,000 per month. Put another way, every month that the organization waits to implement the change costs $250,000. Use a spreadsheet model so that numbers can be quickly changed if assumptions are challenged.

Use the Best Strategy:

Assume that you have run some stellar analysis and believe in ten recommendations. You want to implement them all, but know there is no way they will all be acted upon. Lead with the one that has the best chance for success even if the quantifiable impact isn’t as large that of another recommendation. Once successful results have been realized, this success will justify more initiatives. With each win, it will become increasingly difficult for the IT organization to prioritize other projects ahead of the ones driven by comprehensive web site analysis.

Lead with your best chance for success even if you can’t quantify the potential as being quite as large as one of your other recommendations.

Hypothesize and Test Results

It is imperative that recommendations are positioned as hypotheses, and that all parties understand that implementing a change hasn’t proven anything. Assume your company is running an email campaign designed to lure visitors back to the site to buy laptops. The target population is 10,000 recipients. The test may be as simple as sending 1500 recipients an email with subject line “Free Trial” and another 1500 recipients an email with subject line “Limited Time Offer.” If the “Limited Time Offer” subject line results in a 10% higher conversion rate, consider sending the remaining 7,000 recipients this version.

Leave a Reply

Your email address will not be published. Required fields are marked *

What Our Clients Say
482 reviews