Going beyond tracking and logging

Taking a step or three beyond the analytics programs

[iamlost]

OK, this is WebmasterWorld, and this is the Website Analytics - Tracking and Logging forum.

Are there any members who do NOT use Google Analytics?
Are there any members who think that analytics is not only great fun but a critical business requirement?
Are there any members who go beyond simply Tracking and Logging?
Are there any members who take the input from their raw data and/or analytics program(s), i.e. Piwik, GoAn, and then go that extra step or three, i.e. visitor fingerprinting, probabilistically assigning referrer fingerprinting to users, contextually targeting visitors, calculating referrer value, bot defences quality control, link rot checking, etc. et al?

If so let's have a conversation.

I'll start.
Most webdevs know the conversion value of Google traffic, many know that of other SEs, direct (bookmark, type-in), etc., some even know the difference between mobile and desk/laptop and tablet. Of course the rich SE query data has mostly gone MIA except for the vague WMT bucket sort sop to complaints. And if SEs or just-plain-Google is your sole traffic source the following is moot.

What about the individual value of non-search referrers? What is the connection that brought the visitor from that particular site to yours? Very few have a clue. But one can gather a whole lot of contextual data that all together can attach a pseudo-persona and behaviour pattern and yes, a value, to visitors from each referring site.

Dump logfiles into DB and sort by domain. Check each referrer page individually (manually or automagically depending on volume) and pull title, meta description, header, subheaders, referring link anchor text, surrounding paragraph text. Add in your page's same information. Add in visitors clicktrack, conversions, etc. Over time, given that that link continues to send traffic a pattern usually emerges.

A pattern of behaviour attached to a type of person interested in certain specific commonalities between that site/page and yours. And while that link on that page exists subsequent traffic is identified within certain bounds (from rather amorphous to quite specific). With known context ad and aff targeting is enabled far beyond mere default scattergun.
Note: if you don't know the value of a referrer how do you know whether to be upset when it stops sending traffic? Or that perhaps reaching out might be in order as worth the effort?

Plus with actual contextual content delivery even greater benefits may accrue to both visitor and site. In testing now.
Thanks to the wonders of analytics.

What do you do with your analytics analysis?
How far do you stretch beyond the analytics business as usual?
?

[DixonJones]

There was a very interesting technology, called visitorville (still seems to be there) that did a great job of turning raw logs into personas. Also turned your logs into a game of SimCity. But ultimately (when I tried it a LONG time ago) it didn't take too many lines of log to bring it to its knees. Might be better now though!

Anyway - to jump into your conversation...

We have always used Piwik as our prime source, hosted on our own machines. GA keeps on telling me I have too much traffic for their system... which is very nice of them, but not inspiring. But in any event, I think the influence of the content is often more important than the traffic.

> if you don't know the value of a referrer how do you know whether to be upset when it stops sending traffic? <
Very true, but the mix is not that straightforward. the "value" of a referrer is The amount of Life Time Value they generate less the Resource involved in maintaining the relationship. Now if you have 10 referrers... I guess you can work out the time and energy and possibly money that you invest in dancing the Google tune, but beyond that it gets chaotic. Looking at my stats for yesterday, 35% of new user signups were marked as Direct Entry by Piwik! We can track Emails, ads (not that we have many) and of course we love links, but still 35% came in as "direct entry". In fact that is GIGO (garbage In Garbage Out) because the truth is that they either chose not to be tracked, or the referrer string got lost in translation... or they switch device from mobile to PC when they decided to sign up. Whatever the options, I highly doubt 35% of my conversions yesterday suddenly decided to come direct out of the blue.

I saw an article in the Drum the other days saying Bot fraud will cost advertisers $7bn in 2016. Now Please don't get me started on whether the Drum is the fountain of all wisdom or just a gust blogger's link farm, but I doubt the numbers are a million miles off. Presumably these bots don't "buy", but that buggers up monitoring on a freemium model, because if the bot signs up for a fake free account, then the costs start there... not when they pay (which they never will).

So my resolution for 2016 has been to stop concentrating on the top of the funnel for a bit, and start concentrating on that point between "interested" and "custoer". As a SAAS, our bread gets buttered only when the credit card comes out. Until recently we gave out 30 day trials to people, as long as they put skin in the game and listened to a demo. But how long does it take to make a buying decision? NOT 30 days! You buy a house in 30 days, not a SAAS. So it turns out that the critical KPI for me is not ROI, but the time it takes us to decide whether a free user is a "Buyer, Die-er or spammer".

They didn't teach me THAT at log school.

[iamlost]

We all here (so far as I know :)) have limited capabilities/resources and so must carefully budget areas of greatest return for least relative cost/effort - one after the other. Certainly addressing the differences between 'interested' and 'customer' could well produce significant benefits from both. What is important is to keep working the visitor-site-benefits-conversion connections wherever a bottleneck or opportunity present; change is a given, just sitting back and leaving things on auto will eventually ruin any business because competitors are just that.

All analytics tools are flawed and come with built in irritants/shortcomings but at least Piwik is open source and customisable. As to referrer and user-agent, both are optional includes and even worse easily spoofed. I appreciate but never use them as primary identifiers.

Very true, but the mix is not that straightforward. the "value" of a referrer is The amount of Life Time Value they generate less the Resource involved in maintaining the relationship. Now if you have 10 referrers... I guess you can work out the time and energy and possibly money that you invest in dancing the Google tune, but beyond that it gets chaotic.

Do Google focussed/reliant sites really have that few page traffic referrers? My 'worst connected' page has 90-some different domain external backlinks that have sent traffic in the past year; many have thousands, some hundreds of thousands.

I fully agree with the importance of LTV as a boundary. I began with the goal of the best 1% and have worked up to above 5%. My tests show that somewhere above 10% link instability aka chaotic change begins to be significant so that is my current endpoint target: the top 10% external traffic referrers for each page. Given the number of pages and their links that it is still a serious undertaking. Nor is maintenance/updating trival.

However, as you mentioned, look at the effort that so many put into Google and their marginal bucket sort of search terms. General searchers who are among the worst converting traffic a site (at least mine) can receive. Better to know more about that best identifiable 10% of referrers that already do convert well (5-plus times Google) and provide them a more compelling reason to convert, to return, to recommend; given evidence to date a very nice almost double increase. And that is without factoring in visitor fingerprinting and individual personalisation.
Note: I am working my rear off to get away from general search as a major traffic source; have been for a decade. The more the SEs attempt to answer and hold rather than refer, the more their results go GIGO, the harder I work.

I have always appreciated software/SAAS 30-day trials as an opportunity to get some task(s) done not as a road to purchase; not of course what you like to hear but are no doubt quite familiar with. I would think that if someone is really serious one to three days is sufficient trial for most offerings.

I am confused by:

...it turns out that the critical KPI for me is not ROI, but the time it takes us to decide whether a free user is a "Buyer, Die-er or spammer".

Are the two really that disconnected? I would think that faster differentiation would improve ROI? Or am I missing something?

Contextual delivery is my development route forward (as it appears also to be Google's...) and just as geo-location is a critical mobile consideration (as opposed to important for non-mobile) so too is where the visitor came from. Truly direct is one thing but most came via some other site and knowing about that site-page is a contextually critical input if they are new and still important if they are return.

Of course, every business model, vertical/niche, target market et al somewhat to significantly differs.
Note: which is why cookie cutter sites perform so mediocre to average, they are.

I just find it bemusing that so many put so much into Google and not the far greater elsewhere. This dependance on third parties to do/supply everything is simple, easy, and incredibly limiting.

Breaking free begins with logging and tracking.
Then come analyses, hypotheses, and tests, rinse and repeat.
Then come analyses, plans, and actions, rinse and repeat.

[lucy24]

Are there any members who do NOT use Google Analytics?

Well, yes. But the rest of your post kinda implies that in this line you meant "analytics generically", whether GA or Piwik or something else.

[iamlost]

@lucy24: I guess it does although it was really exasperation of this forum being so GoAn post heavy and basic.
The rest was meant to ask what folks do with their analysis, how far they stretch in which directions. Guess I was clear as mud. It's just that I find the possibilities fascinating and love to muck about.

[lucy24]

I have piwik, but can go months on end (literally) without checking it. Mainly I run my raw logs periodically. Intervals range from a few days to a full month, because my sites range from tiny to microscopic. I've got everything on a fixed schedule. There's information you can get only from logs and not from analytics, and conversely there's information you can get only from analytics and not from logs.

Raw logs tell me that such-and-such IP has opted out of analytics, so from now on I will only see them in logs, not in Piwik.

Raw logs tell me how many previously unknown (and hence not blocked) robots visited-- and, for that matter, how many known robots made a request that was promptly blocked.

Analytics tells me when someone followed a link on some page of mine to go off-site. (As noted elsewhere: when people leave my site, as eventually they must, it's gratifying to think that they're going where I sent them. You could also observe, cold-bloodedly, that if they go from search engine to your site to a third site, it creates a longer time gap between leaving the SERP and coming back to try something else.)

Analytics tells me when someone revisited a page that's still in their browser cache, so logs won't show a fresh request.

There are obvious advantages to having analytics living on your own server and subject to your own access controls. You might still find analytics claiming so-and-so was on your site when they really weren't-- but at least you won't find claims that someone visited when you know darn well you've blocked them outright.

Oh, yes. In addition to logs and analytics, I track request headers for pages. I've even got the header-tracking code on the 403 page, so in the rare case that someone requests an image and is blocked for reasons unknown, I can figure out why. I kinda think it was iBill who originally posted the code. I've only recently figured out that it's a heck of a lot more useful if you also record what URI they requested (not a header per se, so you need one more line of code*).


* In my case two lines, because the first thing I tried didn't work and I was testing on a live site so I didn't want to take the time to clean it up, just went to something I knew would work. I did say I only speak three words of php, but you thought I was exaggerating.