Example Web Store A

The first web site we analyzed was a boutique retail portion of a financial services web site. Since the client's core competency was not in retailing, they had little experience in developing a consumer strategy for their on-line retail presence. After several months of fairly successful on-line sales, we examined the site's performance. What we found were some significant stumbling blocks that were preventing customers from completing transactions.

We looked at a two month period of sales and web statistics for the period of October through November, 2000. For those two months, 3950 people quit checkout before completing it. These are customers who had an item in their shopping cart and left the site after becoming involved in the checkout process.

There we 935 orders for this same period. This means that there were over four times as many failures as there were successes (i.e., completed sales). We used these two numbers (e.g., total failures and total completed sales) to come up with a metric we called "total attempts." By adding these two numbers, we see that the number of total attempts for the two month period was 4885 (935 + 3950 = 4885).

By dividing the number of failed attempts by the number of total attempts, we have a failure rate of 80.85%. This means that over three-quarters of the users who tried to purchase something on the site failed and gave up.

Our analysis also revealed that 86.80% of the failures occurred at a log-in screen which was the first page of the checkout process.

These statistics were a surprise to many people in upper management. Many arguments were made about the validity of the numbers. For example, it is believed that at least some of the people counted as failures may have actually come back and completed a purchase later on. Furthermore, the reasons for not completing the transaction may not have had anything to do with the usability of the system. For example, a user at work may have had to terminate a transaction because the boss walked in. Or a modem user may have lost a connection through their Internet service provider.

Many of the arguments were valid, but could not explain the entire phenomenon. So I made concessions and reassessed the data. We adjusted the failure rate to just 25% of the originally recorded number. We then have 987.5 failures (still more failures than orders). I believe that this 25% estimate is absurdly low. The actual number is almost assuredly quite a bit higher. But even with these extremely conservative estimates, we still see a failure rate of over 50%.

The average sale amount for these two months was $56.69 per transaction. Using our conservative estimate of 987.5 failures, the potential lost revenue was $55981.38 (987.5 X 56.69 = 55981.38) for the two month period. If the real number of failures is 50% of the original estimate, that number shoots up to $111,962.75.

Of course, the actual lost revenue could be significantly higher if the true number of failures is closer to the number originally observed. But even with these conservative numbers, the argument is clear. Something had to be done to address these issues.

Our second store example showed a similar pattern of results.