how to define "popular"

2 Aug 2010 - 5:01pm
6 years ago
3 replies
708 reads
Diana Wynne

The B2B site I'm working on will have several crowdsourced modules such as popular articles, most saved documents, and recent popular searches. These can be filtered by role or region, in order to show "most popular for people like me."
I'm wondering if anyone has found or documented standards on how popular links are calculated, for example whether they are total number of clicks or unique visitors. This isn't a public site, so we're less concerned with advertiser conventions than what will product meaningful results for user.
I did a little research on timeframe for popular, which roughly corresponds to publication cycle. So a daily newspaper popular is last 24 hours, but a quarterly publication might be 90 days.
The requirements for reporting are somewhat different than for presenting to users. For example an admin might want to know what the most clicked links for January were, whereas users just want the current period.
Thanks for any thoughts,
Diana WynneSan Francisco


3 Aug 2010 - 4:13pm
Greg Johnson

My $0.02:

Popularity would best be quantified as a percentage of people who experienced the site in relation to the number of people who performed the desired action. For example: if I were and I wanted to show the most popular articles, my algorithm would produce the top 10 articles with the highest percentage of people who dugg the article of those who saw the article for all articles over a certain threshold.

You need to be able to quantify: goal vs. the number of users who had the opportunity to complete that goal. That could be an email to friend link or someone who spent more than a predetermined amount of time on a page.

I don't think that popularity is as simple as the number of views or clicks though.

3 Aug 2010 - 10:13pm
Diana Wynne

Thanks Greg. Meaningful action is an important point. And agreed, there's nothing simple about Popular.

Unfortunately in this case, they're just reading information. We started out having separate lists for viewed versus downloaded, but with PDF's displayed in some users' browsers, the line is less useful.

We will measure how many times a page is favorited, but it's a private network pushing content rather than a public site with social networking features. There are calls to action, mostly offline, which are tracked separately.


3 Aug 2010 - 11:53pm
Gino Rodrigues

IsnĀ“t popularity based on goals going to bias the result? There are other good opportunities in other metrics that can really convey popularity, when properly combined and tested:

Hits per timeframe - as you found on your research, it may be very difficult to predict the result based on varying timeframes. Maybe you can use the cycles to weight the score, or simply group the presentation of content with similar cycles.

Absolute number of hits - can be used to measure long term popularity, that goes beyond framed measures.

Interactions - content that has been e-mailed, tagged, bookmarked, rated, commented, dugg, ... shows itself as an object of some kind of valuable interest.

Conversion - a contact, a request, an answer, .... desired actions are as useful as any interaction, but count specially for ROI measures.

Hits from popular searches - as you said those are already identifiable, search popularity not only crosses through the information architecture, but also creates means to enhance it.

Sources - referrer analysis can be made and then automated. For example: internal hits can be considered more valuable; hits from Google can mean less, because of its potencial volume; hits from twitter or mail can reveal an active interest to spread the content;. hits from a certain campaign cam be treated only temporarily.

Well, these are my "any thoughts" =)

Syndicate content Get the feed