If you were to measure a website’s momentum, how would you do it?

In a post from last week, I made reference to the need for a momentum metric to buffer the standard strength metric that we get in PageRank. If a PR 8 site hasn’t been updated in 9 months, and is completely blank, that should be reflected in some publicly accessible data.

My question for this post is how would a momentum metric be best accomplished, taking into account the variety of sites that exist. In the comments section of that post, Phillip gave us a start:

Maybe it could be a rating that combines the separate metrics of today, i.e. this number of percentage points is given to Alexa traffic rank, this for PR, this for the number of Technorati mentions, and so on. The question of who gets the biggest weight in the rating would surely be cause of debate.

So why don’t we hack this out in the comment section of this post. Here are the questions to consider:

1. Momentum is a measurement of relative strength over time, and as such requires that we take “snapshots” – what is an ideal frequency of snapshots? One day? One week? One month?

2. To answer question 1, it would help to know the data that we’ll be considering in measuring a site’s momentum. Afterall, why take a daily snapshot if our data is only updated once a week or once every 3 months for that matter. So what data is relevant? Backlinks? Alexa? Technorati? PageRank?

3. How much weight should each of these data points be given?

If you have some ideas on how this topic, feel free to comment, but for answering these 3 specific questions, I propose the following format:

Once a week

Alexa – 25%
Technorati – 25%
PageRank – 15%
Backlinks – 30%
Delicious Votes – 5%

8 thoughts on “If you were to measure a website’s momentum, how would you do it?

  1. I think you are describing a metric that is essentially a weighted derivative value like SEOmoz.org’s pagestrength measurement.

    I think the main thing to consider is that a momentum calculation is not a static picture of metrics like PageStrength, but instead should be a measurement relative to time. It has to be a vector to measure momentum (acceleration/deacceleration).

    This means that you need to capture information over a time period and run that through either a very good machine or possibly through an OLAP or something to get at the information you are looking for. Tie that into an OLAP and you could not only provide metrics of historical terms, but you could then do what if analysis for the future.

    Ergo you don’t just want to measure “What happened in the last 7 days? or last 30 days? or since Date X?” You want to be able to take that information as a piece of data and project out, what will happen in the next 7 days? and what will happen in the next 7 days if I do X, Y, and Z on day 2, 3, or 4 and all the permutations in between.

    Not sure if I mentioned it but OLAP’s are the cat’s meow!

  2. Great formula there, Ryan. As Artem put it nicely, there is little point in tuning the weights this early, as Performancing would be the one to set the standard. The tuning happens a few weeks (and a few hundred blogs) into inception

    IMO a good frequency of snapshots would be once a week — not too burdensome on the server. Of all the metrics we listed (Alexa, PR, Technorati, etc), Technorati would be the one which needs daily snapshots the most. The others can do with weekly updates. Maybe we can have a system of varying frequency of snapshots for each metric?

    It’s good to see PageRank not given that much of a weight 😉 Of all those metrics, I think it’s the easiest to be gamed.

  3. Since there is no definition for the “momentum metric”

    relative strength over time (i.e. strength relative to self)

    What would be interesting is to see this metric used to buffer a raw strength metric. Here’s an equation:


    Where momentum is on a scale -10 to 10

    Let’s assume that PR is the measure of strength for the time being. And let’s assume a PR 8 site with quickly declining Alexa stats, backlinks, etc. Momentum= -5

    8*(1+(-5/10))= 4

    Compare this to a site has a PR8 with less drastic dip in momentum of -1

    8*(1+(-5/10))= 7.2

    So here you could have 2 sites, both with PR8, but the momentum metric would clearly bring out the fact that one site is more healthy than the other.

  4. > I think your formula is weighted correctly.

    Since there is no definition for the “momentum metric” there is a little point in tuning weights

    How about being a bit more scientific?
    1. You could gather, e.g. 100 blogs known by you, range them by your “feeling of momentum”
    2. Then take random 50 of this set and tune your metric so that it correlated well with the “real” feeling of momentum
    3. Verify the formula by seeing if the formula arranges the remaining 50 blogs in the similar order that you did manually

    Then periodically (say twice a year) update the formula by repeating the above procedure.

    It would take quite some time for tuning and a decent expert(s) to provide a feeling of momentum, but the resulting metric might be a respectful one.

  5. Ryan, I had proposed an idea to Mark and Micah that we have a list called the Performancing 20, based on sites registered as stats-public in PMetrics. Maybe your momentum metric could be combined with my list idea?

  6. This is a really interesting idea! I’m wondering if more than just Delicious should be included- maybe also StumbleUpon at least? Maybe even a few of the other social bookmarking sites (just a few). I think your formula is weighted correctly.

Comments are closed.