[TriLUG] Tracking File Downloads

Ken MacKenzie ken at mack-z.com
Sun Jan 25 13:19:08 EST 2015


One thing I will mention is the reason I bring up questioning my stats is
this:

3816.10%614.08 MiBSo there are 38 hits on the episode based on the filter I
first posted.  Total bandwidth is 614mb but the file size of the episode in
question is about 97mb.

So I am thinking that only 6 people downloaded the full episode and someone
else only part of it.

Or...

Perhaps this shows inline plays.  So yeah there is a podcast stream you can
subscribe to that podcatchers use.  But you can also just click the mp3 on
the site.  If your OS/Browser understand opening then perhaps someone
partially listened or clicked and changed their mind and only part of the
data got there.

One of things I am working on is getting us in stitcher.  Just waiting for
the confirmation we are in.  Then I want to integrate the Stitcher player
into the site for people who want to listen on the site.

What this all comes down to is I am trying to stay ahead of the issue of
real metrics tracking to eventually talk to people to sponsor the show
monetarily.  Based on the numbers I am seeing our fan to listener
conversion rate is low.  We have near 700 FB fans but the best I can tell
maybe 1/3 have listened to any portion of the show.  Podcasts being the
netflix of radio of course people could be subscribing but not downloading
or listening yet.  I don't know.

But I want a high degree of confidence in how much the audio file is
getting out there, listeners is harder to track so we have put some easter
eggs in the audio to try to encourage listener response for tracking.  That
information will help us really decide proper marketing strategies.

Ken


On Sun, Jan 25, 2015 at 1:08 PM, Ken M <ken at mack-z.com> wrote:

> No worries, I don't consider it a hijack.  You gave a really good
> suggestion for me to add to my arsenal.
>
> Cloudflare or not I still want to figure out my own on server solution to
> this as well though.  But I think the combination well help me improve the
> confidence I have in report accuracy.
>
> Right now what I am doing for reporting is this.
>
> I have 2 scripts, one meant to be run daily and one hourly.  These scripts
> filter for the site (multisite install) as they build these working logs.
>
> The daily one build temp logs from zcat reads of the archive into
> temporary month day and week logs.  Then those logs are run through
> goaccess to build html reports.
>
> Those daily logs have three modes, Past month, past week, past day.  And
> there are 2 categories for each mode.  Episode download and full site
> statistics.  The latter being an unfiltered run through goaccess.
>
> The hourly job just passes the current nginx access log through goaccess
> filtering only for site and builds a report.
>
> So all of that is working and in drupal I have a content type that only
> the admin and editors can see for site reports and each one is a quick php
> script that checks the presence of the report they are to display.  If it
> is there it puts the last modified date out then includes the html of the
> report.  If not it displays an error msg.
>
> So far an elegant solution.  I need to set up drupal to alter the template
> though for the reports as the overall site template kind of gets in the
> way.  The reports are out of any web root and the pages to include them in
> drupal are secured so no user without the proper credentials can get to
> them.
>
> The last thing I need to do in all this is setup a monthly cron job to
> snapshot each months report to an archive location.
>
> The need for this solution is drupals own site statistics although
> capturing the event are not the most friendly to view and their collection
> hits the DB which is a performance concern.  Analytics with drupal can only
> capture so much and direct file access like podcatchers will use is not one
> of them.
>
> Ken
>
> On Sun, Jan 25, 2015 at 12:57:50PM -0500, Igor Partola wrote:
> > Oh I absolutely understand that. The free CloudFlare plan gives you a
> TTL of one hour which if you are doing active development can be way too
> long. Though this won't help you with analytics, they do have a lower level
> option where they don't cache things at all, but then they also just
> marshal TCP connections around.
> >
> > On the whole, I think a CDN is a very good thing to do for any size
> site. Let them worry about network capacities, uptime, etc. I am certainly
> converting all my sites to use it, though slowly.
> >
> > I'll stop hijacking this thread now as your actual question still needs
> to be answered.
> >
> > Igor
>


More information about the TriLUG mailing list