Loading Profile...
Feed Engagement API Averages
http://api.postrank.com/v2/feed/engag...
{
"5ce49a804180f62171ff48275bd08b59": {
"sum": 470.0,
"count": 11,
"average": 22.3809523809524,
"min": 7.0,
"max": 114.0
}
}
I can't understand where the 'average' comes from, I would have thought the average is sum divided by count, which is about 42, not 22.
{
"5ce49a804180f62171ff48275bd08b59": {
"sum": 470.0,
"count": 11,
"average": 22.3809523809524,
"min": 7.0,
"max": 114.0
}
}
I can't understand where the 'average' comes from, I would have thought the average is sum divided by count, which is about 42, not 22.
I’m unsure
1
person has this question
I have this question, too!
Tell me when someone answers.
The more people who ask this question, the more it gets noticed.
The more people who ask this question, the more it gets noticed.
The best answer from the company
-
Hi David, this may be a poor choice of keys on our part...
Count = Number of time periods where engagement was found over the interval (aka, non zero)
Average = average engagement over entire time period (including days with 0 engagement)
Your query is for ~3 weeks in December, hence 470 / ~21 days, whereas count is the actual number of days with non zero engagement.
I’m confident
The company says
this answers the question
Create a customer community for your own organization
Plans starting at $19/month
-
Inappropriate?Hi David, this may be a poor choice of keys on our part...
Count = Number of time periods where engagement was found over the interval (aka, non zero)
Average = average engagement over entire time period (including days with 0 engagement)
Your query is for ~3 weeks in December, hence 470 / ~21 days, whereas count is the actual number of days with non zero engagement.
I’m confident
The company says
this answers the question
-
Inappropriate?Thanks so much for that Ilya. If I can pick your brains a little...
What does it mean if a blog has no engagement score on a particular day? Is it that there was genuinely NO engagement at all, or some other reason?
I'm sort of leaning to thinking of sum divided by count as a better measure of engagement, but it depends on your answer to the above.
I’m thankful
1 Comment Add a comment
This answers the question
-
David, no engagement means that PostRank did not receive any activity from any of the sources we track (bottom of: http://www.postrank.com/postrank). The page may well have received pageviews, so the engagement number is for off-site activity.
In other words, we don't store 0 values in our database, so a missing day is implicitly a 0. In all of our calculations at PostRank we use the full period between 30 days - this keeps it fair for everybody. Otherwise you're competing on biggest peaks of engagement. -
Inappropriate?Hi Ilya, thanks again.
This is my site incidentally, why does it have no comment data except if you go back to early November?
http://www.postrank.com/feed/2788b841...
I’m undecided
1 Comment Add a comment
This answers the question
-
David, what do you mean? If I hover over the postrank scores on some of the recent articles I am seeing comments come up.
-
Inappropriate?Sorry, yes, I can see them now, I swear 5 minutes ago it was only twitter showing, sorry....
I’m sad
1 Comment Add a comment
This answers the question
-
Hmm, that's entirely possible. Comments check are not real-time (unfortunately), so our system could have scheduled your URL's for update just a few minutes ago.
-
Inappropriate?Can I ask about this:
http://api.postrank.com/v2/feed/engag...
no data.
http://www.postrank.com/feed/f4ffc455...
has data
I’m undecided
-
Inappropriate?David, it was a cache error - found the problem and it should be resolved now. Try changing the end parameter by a second, ex:
http://api.postrank.com/v2/feed/engag... -
Inappropriate?Your comments counting can't seem to handle pagination.
Look, when I checked my own site via api for first 21 days of december the average engagement was 6.6. Now you change timestamp by one second and it's 657, it's impossible for such a difference with just one second. So I'm guessing it wasn't fully indexed or included until I asked about it above, or a technical reason.
For the api I'll be checking a lot of pretty obscure sites in indonesia so just wondering if postrank can cover them.
So, if I query for the site id and it doesn't exist, does your system take note of this and then queue the site for inclusion?
Sometimes the site id exists but there is no engagement data, same question, will it likely update at some point once the system knows that that site is being asked about?
I’m undecided
-
Inappropriate?David, all good questions.
> Your comments counting can't seem to handle pagination.
Depends on the site, but we do require custom templates to handle most pagination cases. If it's a big or popular site we're more than happy to add pagination support for it.
> So I'm guessing it wasn't fully indexed or included until I asked about it above, or a technical reason.
The feed you're looking at is new to our system, which means we kicked off comments discovery the moment we saw it. One limitation of comments checks is the fact that we can't reliably extract the publication date of each comment, which means that we attribute all of the engagement to that first day when they were discovered. After this and once we're following the feed, we do periodic and dynamic interval checks for each story, which allows to grab the comments very close to their publication time. That's the reason you were seeing that big spike after changing the time.
> For the api I'll be checking a lot of pretty obscure sites in indonesia so just wondering if postrank can cover them.
As long as there is twitter, delicious, etc. activity, it will definitely work. If there are big and popular sites in Indonesia that we could take into account, please let us know - would love to investigate those.. We're always looking to extend our coverage.
> So, if I query for the site id and it doesn't exist, does your system take note of this and then queue the site for inclusion?
Yes. When you query our FeedInfo API, by default the feed will be added to our index if we haven't seen it before. If you don't want that behavior, you can pass a "noindex" parameter.
> Sometimes the site id exists but there is no engagement data, same question, will it likely update at some point once the system knows that that site is being asked about?
If ID exists and there is no data, it means we haven't seen any off-site activity for those URLs. Having said that, we don't check for comments for stories we don't know. So, if you add a new feed, we'll schedule comments checks, which in turn, might result in found engagement.
I’m confident
-
Inappropriate?"The feed you're looking at is new to our system, which means we kicked off comments discovery the moment we saw it"
It had five subscribers, although none seemed active.
"If there are big and popular sites in Indonesia that we could take into account, please let us know - would love to investigate those.. We're always looking to extend our coverage."
Big and popular sites in Indonesia tend not to be blogs....http://www.jokosusilo.com/ would be the most popular, as it's a very commercial site I don't include it on my list here http://blogs.indonesiamatters.com/, those are the sites I'll be checking, assuming I ever get an api key....
I’m undecided
1 Comment Add a comment
This answers the question
-
David, are there any social sites where users share links, vote on stories, etc? Specifically in Indonesia? We could take a look at adding those into our metrics.
-
Inappropriate?Here's one - http://www.lintasberita.com/
Seems to be an awful lot of blogspot.com sites in the voting so it looks a bit spammed and I've always had the suspicion their voting numbers are faked upwards, but there is some genuine stuff there, some of it good too.
Plurk also, not a voting site, but it's more popular in indonesia than twitter, not for much longer though. -
Inappropriate?Thanks David. Lintasberita doesn't have an easy API, but I've added it to our list to look into - we have some stuff in the works that may enable us to track sites such as lintasberita.
Out of curiosity, why do you say "not for much longer" for Plurk? I've heard they got blocked in China, is the same thing happening in Indonesia? -
Inappropriate?I did this graph in May 2009
But now:
so I was wrong, twitter more popular now1 Comment Add a comment
This answers the question
-
Ah, gotcha. Good for Twitter. Not so good for Plurk.
-
Inappropriate?Yes, so, for my list, do you think it possible to track those sites in a useful way via the api? I'm just looking at the average engagement numbers.1 Comment Add a comment
This answers the question
-
David, given the popularity of twitter - that's already one good signal. One strategy I would advise is: run the list of sites you want to track against our API, that'll make us import the feeds if we haven't seen them before and also run comment checks for them. From there, you can take a closer look at the engagement numbers, etc.
-
Inappropriate?Thanks Ilya, I've already added in most of the sites into my own postrank account and that seems to have had an effect. How do I get an api key, I did ask Melanie for one but no response.1 Comment Add a comment
This answers the question
-
This comment was removed on 12/29/09.
see the change log -
Inappropriate?Thanks Ilya, um can you delete that key now as somebody else might use it!1 Comment Add a comment
This answers the question
-
Good idea! Done.
-
Inappropriate?I'm having trouble with this one - http://www.postrank.com/feed/858f3924...
I couldn't get its id from the api so added it manually, but I can't get any engagement score and it won't show up in my subscriptions.1 Comment Add a comment
This answers the question
-
David, are the calls to the API's returning an error? Can you please give me the URL you're trying to query? (minus your private key)
-
Inappropriate?To get the id I use this - http://api.postrank.com/v2/feed/info?...=
then append the url http://www.benablog.com/
'feed is not in our system'
So I couldn't get it, then manually added 858f3924fdda2d89b3f25c123d561c97. If I query that one manually I get all 0's. No, just tried, works now. -
Inappropriate?Another thing is I'd like to display a 'best of the best' top ten posts from all my subs, but when I try the user - topic api or use my rss feed for the topic I get sort of random rubbish results. The top result with 9.8 postrank is this - htt*://madrotter.blogspot.com/2010/01/around-world-musical-schizophrenia-beat.html
It would be embarrassing to have a top ten list with that on top! -
Inappropriate?Looks like you fixed this, seems to have improved.
I think, to complain again, that postrank scores are too dependent on frequency of posting, blogs that have only one post a week, but with lots of activity on them, just can't compete against blogs that post day after day and have moderate to high activity on the posts.
EMPLOYEE
