(banan) #1

We have problem with piwik. We use piwik to track visits for many sites, but few of them have big differences in visits (Piwik vs GA). I’ve read that 10% is normal, but 30% with a site that has 13.000 uu (monthly) is a little to much style_emoticons/<#EMO_DIR#>/wink.gif.

Could you tell me what is wrong?

(vipsoft) #2

Google does not disclose its algorithms, so I can’t offer any explanation for the large difference.

Keep in mind:

Piwik has filters for smart bots (e.g. Bing, and now Googlebot) which behave like browsers and execute the javascript tracking code.

Piwik also uses cookies and a heuristic algorithm to recognize return visitors.

Depending on the page load times and order that your pages load the tracking code, some visits may not be recorded if the user clicks away from your page. (GA sometimes has huge latency issues.)

p.s. GA implements “sampling” for large sites, so aspects of your stats may not be reliable to begin with.

(diri) #3

I did some comparisons between multiple kind of tracking programs and found same result:

Piwik is NOT counting correct. It does not matter if it is a high traffic site or a real low traffic one.

Most trivial test to ensure me being correct is to include an image direct before in pages watched and count requests in webserver’s log (similar to IVW).

It might be load time of tracking code so you will loose impressions for this but, it is a real large discrepance when Piwik counts 0 (zero) or almost zero while IVW and similar count 100, 1000 or more.

(vipsoft) #4

Comparing GET requests for static images vs javascript is biased because some browsers prioritize image loading over scripts. Secondly, Piwik does not currently count “clicks” with the (JavaScript disabled). Your tests also can’t factor in browsers that block tracking cookies/scripts (eg Adblock Plus).

A more valid test methodology would be split A/B testing between two or more javascript tracking methods that exclude the web bug, over an extended period of time. And then look at the percentages. It still wouldn’t account for the new smart bots (Bing and Googlebot) which can load and execute JavaScript (but which Piwik now filters out), but it’s a vast improvement if you want to compare apples to apples.