Measuring a Niche Website

December 4th, 2007 by Lars Johansson

Stardoll is not your average website. Its target audience is 7–17 year old girls. Stardoll doesn’t spend a lot on marketing, have more than twelve million users, and is backed by Sequoia Capital and Index Ventures.

Too be honest I never would have expected it to be a big hit back when it was called Lisa’s Paperdolls. Stardoll has explored a niche that would have been difficult to make profitable before the mass adoption of the Internet.

Early version of Stardoll

I discussed Stardoll during one of my many interviews with multivariate testing providers. It made for an interesting contrast compared to Sony Ericsson and Volvo.

I am now revisiting Stardoll to talk about measurement and analysis. It’s a challenging website to tackle so I asked myself what I would measure on site like that.

A good web analytics program should focus on what matters. Stardoll makes its money by selling Stardollars, a currency used for purchasing clothes and accessories for your doll. Virtual stores where you can buy the latest fashion include DKNY, Stuff by Hilary Duff and mary-kateandashley. So for a site like this you want to convert visitors into members and members into buyers. Hopefully doing so sooner rather than later, keeping in mind that some actions that don’t lead to a purchase today may lead to a purchase tomorrow.

Early version of Stardoll

The following are the KPIs I would focus on. Keep in mind that it’s trends that you should keep an eye on.

Landing Page Bounce Rate per Marketing Activity = # of Visitors who Only Visited One Page, Segmented by Marketing Activity (i.e. came from Google Adwords, a particular banner etc.) / # of Visitors Segmented by Marketing Activity

New Visitor to Member Conversion Rate = Total # of New Members / Total # of New Visitors

This, like many other, KPIs should be segmented to show how each individual campaign is performing.

Average Visits Prior to Becoming Member = # of Pre-Conversion Visits / # of Total Conversions

Also measure how many visits to the website it takes to get one purchase. Monitor the trend.

Average Cost per Visitor = Acquisition Marketing Costs / # of Visitors from Marketing Efforts

Average Cost per Member Acquisition = Acquisition Marketing Costs / # of New Members from Marketing Efforts

Purchase Completion Rate = Total # of Purchases/ Total # of Visits Where a Purchase Was Initiated

Average Items per Completed Purchase = # of Items Purchased / # of Completed Purchases

Average Purchase Value = Revenue Generated / # of Purchases Made

Average Revenue per Visitor = Revenue Generated / # of Visitors

Average Revenue per Member = Revenue Generated / # of Members

Tell-a-Friend by Email Conversion Rate = # of Tell-a-Friend Forms Sent / # of Tell-a-Friend Forms Loaded

Average Visits per Visitor = Total # of Visits / Total # of Visitors

Log-in Rate = # of Logged in Visitors / # of Visitors

Percentage of High, Medium and Low Frequency Visitors
# of Low Frequency Visitors / # of Visitors = % Low Frequency Visitors
# of Medium Frequency Visitors / # of Visitors = % Medium Frequency Visitors
# of High Frequency Visitors / # of Visitors = % High Frequency Visitors

Percentage of High, Medium and Low Recency Visitors
# of Low Recency Visitors / # of Visitors = % Low Recency Visitors
# of Medium Recency Visitors / # of Visitors = % Medium Recency Visitors
# of High Recency Visitors / # of Visitors = % High Recency Visitors

What constitutes low, medium, and high recency and frequency has to be established based on the specific website.

Member Retention Rate = # of Repeat Member Visitors / # of Member Visitors

Member Attrition (Churn) Rate = 1-( # of Repeat Member Visitors / # of Member Visitors)

Another, better, way to look at it is to check how many of those who were members at the beginning of the month are still members at the end of the month. Now, you can be a passive, non-paying, member so it is worthwhile to use segmentation – for instance by looking at Superstar memberships (paid memberships).

I would make the internal search engine more efficient and then make the search box more prominent. I would use search to drive sales and measure the following.

Percent Zero Result Searches = # of Searches that did not return a result / # of Searches

Percent Relevant Search Results = # of Searches That Returned a Result And Lead to a Click on a Search Result / # of Searches

Percent Irrelevant Search Results = # of Site Exits from a Search Results Page / # of Visits to a Search Results Page

Percent Repeat Searchers = # of Visitors Searching at Least Twice / # of Visitors Searching at Least Once

Search to Purchase Conversion Rate = # of Purchases Attributed to Searchers / # of Visits to a Search Results Page

Purchases by Customer Support Replies = # of Members That Got Questions Answered and Then Made a Purchase / # of Members That Got Questions Answered

Current version of Stardoll

Member Engagement Score

Measuring engagement is the cool kid on the block in web analytics, but it’s not an easy thing to measure. For Stardoll I would suggest that scores are based on how likely each action is to drive sales. Therefore some background work has to be done before the scores can be determined. Monetize all the actions! The actions I would score:

• Created a MeDoll
• Created a scenery
• Edited profile
• Dressed a celebrity doll
• Saved a celebrity doll to an album
• Uploaded an image
• Added a friend
• Joined a club
• Started a club
• Used the tell-a-friend function
• Wrote an entry in a guestbook
• Wrote an entry in a Starblog
• Rated a MeDoll
• Got a pet
• Chatted
• Commented an album
• Watched Stardoll Cinema
• Added clothes to closet (shopping cart)
• Bought clothes
• Sold clothes in StarBazaar
• Paid for membership (1 month)
• Paid for membership (3 months)
• Paid for membership (6 months)
• Paid for membership (12 months)
• Entered a gift code

Is “Dress up one of these dolls” or “5 dolls like x” based on the selections made by other members? If not, maybe try that Amazon-like approach.

I would also try to list the dolls that have generated the most money first when you arrive at the site, or highlight them in some other way.

I would then use on-site behavioral targeting/individually adapted content to show the previously viewed items in the main container to returning visitors. Make them feel welcome back and make it easier for them to find what they looked at before.

I haven’t seen any sign of an opt-in email newsletter, but if there is one I would use recency/latency to determine when it’s time to offer a special incentive to decrease churn.

I would segment all KPIs by language/site when relevant.

Phew! How’s that for a start?

Other Posts (Automated Suggestions)

Measuring a Niche Website Part II | The First Five KPIs to Focus on
Top Five Posts of 2007
Top Blog Posts 2009
Measuring Online Content Using Google Analytics
CPM Is Dead

Share the love: