On visits and visitors …

I have a Google News alert on the phrase “web analytics” that had the most interesting summary I’ve seen in a long time:

The Web Numbers Game
Multichannel Merchant – Stamford,CT,USA
… Adds John Squire, vice president of product strategy for San Mateo, CA-based Web analytics firm Coremetrics: “We think [Belkin’s argument] is fundamentally flawed …

The entire article is at Multichannel Merchant magazine online. The basic argument is that some people think it’s better to use “visits” to measure conversion than “visitors”, ostensibly because every visit is an opportunity to convert.

Uh, what?

While I’m inclined to agree with Jason Palmer from WebTrends and John Squire from Coremetrics on this issue for no other reason than I know both guys moderately well and very much respect their opinions, the debate about whether online retailers should use “visits” or “visitors” in their conversion rate calculations is moot.

Use BOTH Visit- AND Visitor-Based Conversion Rate Calculations

Every online retailer should be using two very basic and very much standard calculations:

• Order Conversion Rate (OCR) defined as the number of orders taken divided by the total number of visits to the web site during the same period.
• Buyer Conversion Rate (BCR) defined as the number of customers converted divided by the total number of visitors to the web site during the same period.

Setting aside for now any issues associated with the definition of “visitor”, examining these two conversion rates side-by-side gives you unique perspective into your customer base. Do you sell low-consideration items? Likely your OCR and BCR will be similar. Do you sell high-consideration items? Likely your OCR will be low but your BCR higher, especially if you’re looking across weeks or months.

The example given in the article, one where one visitor visits four times and purchases twice, yielding a OCR of 50% and a BCR of 200%, is strangely presented as if the BCR is “bad information.” The original author states (lifted from the article):

“If you use weekly unique visitors, my conversion rate is 200%. If you use visits, my conversion rate is 50%. Which is a better representation of site effectiveness? Clearly, the 50% [number] is much more valuable in understanding where your site may or may not be performing optimally.”

Really?

I don’t understand why any good retailer doesn’t want to know that some percentage of their audience is making more than one purchase during the period under examination? Is order conversion rate a better indicator of site effectiveness? Probably, but it’s a poor indicator of customer loyalty. Is buyer conversion rate a better indicator of customer loyalty? Perhaps, but it’s a less-good indicator of whether your site suffers from process abandonment issues.

Personally as an online retailer, I want both rates.

I need my buyer conversion rate because, much like Intuit who sells TurboTax, I sell “moderate consideration” items but I don’t expect to sell more than one or two items to any given customer. The “every visit is an opportunity to convert” mindset doesn’t help me understand which of my marketing efforts are effective in the long run.

But I need my order conversion rate because I believe in controlled experimentation and want to maximize the likelihood than when a visitor does decide to cart one of my books that they’ll complete the purchase. Here if I focus exclusively on my buyer conversion rate but look at short periods of time then I’ll be sad since it often takes folks more than one visit to make the purchase.

Applying Order and Buyer Conversion Rates to Referring Sources

All of the above is profoundly more interesting when considered in the context of referring sources (domains, campaigns, feeds, etc.) Here I watch my order and buyer conversion rates closely to better understand which referring sources are sending me highly qualified traffic. Consider two examples:

1. Google Japan (www.google.co.jp) referred visitors to my site have a buyer conversion rate of 4.5% and an order conversion rate of 3.4% (a difference of 24 percent)
2. Hurol Inan (www.hurolinan.com) referred visitors to my site have a buyer conversion rate of 1.0% and an order conversion rate of 0.9% (a difference of under 5 percent)
3. Avinash Kaushik (www.kaushik.net) referred visitors to my site have an order and buyer conversion rate of 0.0%

What does this tell me?

• Visitors from Google Japan visit more often before making their purchase, but when they make up their mind a higher percentage are likely to complete the transaction.
• Visitors from Hurol Inan are less likely to make the purchase, but those that purchase don’t take multiple sessions to complete the transaction.
• Avinash doesn’t link to my site or talk about my books very often

See how that works?

Want a Really Interesting Metric?

One thing I calculate to help me better understand my order and buyer conversion rates is the percentage-wise difference between the two. Basically:

(BCR – OCR)/BCR = Percent Difference between Buyer & Order Conversion Rates

This way I can rank-order by referring sources and campaigns to look for sources that are likely to convert more like Google Japan and more like Hurol Inan in the example above. In my dataset, this calculation ranges from 60% at www.fortune-cookie.com (don’t ask) down to -1.3% for visitors referred from www.comcast.net. The negative number tells me that visitors are making repeat purchases (something I honestly do want to know, call me crazy!)

But Wait, There’s More!

All of this doesn’t tell me one very important thing, whether I’m likely to get purchases directly from my referring sources, or if purchases are mostly latent (happening in subsequent sessions). To track this, I add an additional column to my referring source analysis for “latent conversion” which in Visual Sciences I simply define as:

Visitors Who Buy Having a Session Count > 1 / Visitors Who Buy

Because I’m using visitor-based tracking, I have access to the total session count for all visitors referred from a particular source. Now I can create a report that has referrer, percent of sessions, BCR, OCR, the percentage difference between BCR and OCR, latent conversions and purchase value. This report can then be used to learn things like:

• Jim Sterne’s Emetrics.org web site sends me very qualified visitors who convert quickly.
• Google sends me a huge volume of moderately qualified visitors who convert more slowly.
• Avinash still doesn’t love me.

Why Some People Don’t Like Visitor-Based Conversion Rates

One thing worth mentioning is that some people don’t like visitor-based conversion rates. Now I’m not 100% sure why this is but here are some points of speculation:

• Their web analytics application doesn’t really support visitor-based tracking, instead opting to squeeze visitors into oddly shaped buckets (ask Avinash about “Daily Unique Visitors” if you want the detail here …)
• Their web analytics application only supports visitor-based tracking through the use of expensive and impractical data warehouse requests
• Their visitor tracking is based on third-party cookies which have been shown to degrade as a unique identifier over time

Of course I’ve seen people’s lists of “other reasons” that visitor-based conversion metrics are inaccurate, things like “people use two web browsers” and the such. My feeling is that these arguements are designed to obfuscate a larger problem, most likely third-party cookie deletion.

Suffice to say, the buyer conversion rate degrades in value directly with your cookie deletion rate. Based on the research I’ve done, this can cause some serious problems if you’re selling high consideration items. If you’re still relying exclusively on a third-party cookie for you web analytics you need to take this into account.