The Conversion Chronicles, resources for improving your online conversion rates

All Conversions Are Not Created Equal


Forward to a friend      
We respect your friends privacy

Remember when online success was measured in ´eyeballs?´ Turned out revenue and profitability per customer mattered more. Some folks are risking the same mistake with testing.

So if you have an ecommerce site running a test online, and your main measurement is conversions, you may actually be placing more value on the metric than it deserves. That´s because the value of a conversion -- that is, a specific action taken by a customer -- often depends on the dollars it generates.

Testing for conversion rate remains useful if you simply want to see if you can persuade visitors to do something, or if you´re just tr
ying to increase the number of sales. But when you´re selling merchandise or services that have a dollar amount attached, then all customers are not created equal.

Go beyond conversions to look at revenue, because a test that stimulates people to spend more money may, for example, be more effective than one that makes them spend more frequently.

Imagine you´re running an A/B test with two different offers. You´re driving traffic from a single ad and diverting visitors to two different pages: one is a control, with a headline and call to action you´ve been using for years. The other has a new headline and call to action. Beyond conversions, consider tracking the following metrics for each branch of the test:

Metric #1. Revenue, total sales

This is the aggregate amount of sales for each branch. Think of this metric in terms of selling Girl Scout cookies: it doesn´t matter to the troop leader if one little girl sold 200 boxes to one person or 20 boxes each to 10 people. The total amount of sales remains the same.

By comparing the total sales revenue of each branch of a test, you can see in aggregate which branch is performing best.

Metric #2. Average order value

When you divide total revenue by number of orders, you get average order value. Whereas conversion rate measures the number of customers your landing page has an impact on, average order value tells you how much of an impact your landing page is having on them.

For example, if landing page A, with the old headline, has an average order value of $100, and landing page B (new headline) has an average order value of $104, you know that your new headline is convincing visitors to spend more. They may be buying less often, but that´s not what you´re measuring here. This metric is simply to see which branch pushes an already converted customer to spend more money.

Metric #3. Revenue per visit

Because it´s impossible to have an equal number of people visit each branch, this metric simply normalizes the sales number for each branch. Derive this number by dividing the total sales by number of visits.

Metric #4. Revenue per visitor

This metric is separate from revenue per visit, because a visitor may return to your site three or four times before making a purchase. To determine revenue per visitor, take the total sales divided by total number of visitors.

Revenue per visitor is perhaps the most interesting metric, b ecause it takes into account both average order value and conversion rate. If you´re increasing the number of sales through landing page A, but increasing average order value on landing page B, this metric will let you know which of those landing pages is actually the most effective, based on revenue.

We have found that it´s the most valuable measurement tool for determining the effectiveness of different test recipes.

--Two other things to remember:

Before you use the above metrics, you´ll want to eliminate the ´noise´ or skewing effect of very large orders, which happen on almost every ecommerce site. If your average order is $75 or $100, but one branch of your test brings in a $6,000 order, filter out that order so it doesn´t skew results.

Finally, as with any other test metric, be sure you obtain an appropriate level of statistical confidence, so you´re not basing your conclusion on false or inaccurate data. (For more information on confidence levels, see our article in last month´s issue of Selling by Design.)

More from this months issue | Archived chronicles | More from this author
Matthew RocheAuthor: Matthew Roche, CEO and Co-President - Offermatica

Matthew Roche is Co-President and CEO of Offermatica. Offermatica provide a scientific testing platform for applying A/B, multivariate and Taguchi testing to increase online sales.