10911410_541306596011052_6244709954541121276_o

Key takeaways from Superweek 2015

Events

In business decision making we shouldn't be relying on hunches because it would be like playing a lottery. It is completely fine to have an opinion, but only when it is backed up with data will it become a valid choice.

A week ago, I’ve been on a web analytics conference named Superweek. It is one of the top conferences in Europe and for sure among the top 5 most important conferences in the world of analytics and data-driven decision making in the online businesses. Names of the speakers like Doug Hall, Simo Ahava, Yehoshua Coren, Peter O’Neil and Craig Sullivan (among many other industry thought-leaders) and about 100 other participants from various agencies, SaaS companies and freelancers should well describe the importance of the event.

Data beats opinion

The moto of the conference since its beginnings a couple of years ago is “Data beats opinion”. Obviously, the point is not to make business decisions based on assumptions, a “gut feeling” or experience only, but to rather gather and analyse the relevant data and then decide on the next steps. Since there was almost 30 different sessions and various speakers, I’ll choose a couple of main points to present.

1. State of the analytics

In today’s world there are many analytics challenges: tracking a web session across multiple devices, tagging and tracking off-the-web traffic sources; where is the line of one’s privacy and should the companies share their users’ behaviour data to make one big knowledge base? We are living in the time of many devices surrounding us, we are using at least a smart phone and a computer (not to mention multiple devices), with the “Internet of Everything” concept we might soon be living in a world where our fridge will order our food instead of us to fill itself and our washing machine rinsing our clothes to keep it fresh as it detected we are late from work.

Amount of information gathered about a website visitor or a service user is getting much bigger, so we cannot even rely any more on one tool that will be able to track and gather it all. We need to use data visualisation systems like Tableau to be able to connect all the resources and make reports and informed decisions based on them.

10953925_540475709427474_2658684572308952154_o

There was an interesting discussion about whether analytics tools like Google Analytics (GA) should track web sessions (visits) like they are currently, or should it be showing behaviour of each separate visitor (like Kissmetrics does). I can say that I think it is still a good solution to use Kissmetrics for SaaS or any other business that needs to improve their service on a individual needs level, while GA should stay with aggregated data for all of the visits.

 

2. Conversion rate improvements and testing

Conversion rate optimisation (or CRO) exists as a term for a couple of years already and as a dispersed set of optimisation techniques for much longer. It should be a field of marketing and analytics that enables us to improve user experience and fulfil expectations through continuously testing and tweaking web pages’ copy,  layout, navigation, calls to action, etc. It is not limited to the web sites though: you can be testing a mobile app as well.

What we raised as one of the most important concerns (because it is rarely spoken about) is a level of education among the practitioners about the statistics and the relevance of the test results. Split testing tools are presenting the whole process as an easy one that anyone can do and that if you get to the magic number of 95% (talking about the percentage of probability test result is not random), you are a winner. Point that is being missed is that sample size of people being part of the experiment is the most important, length of the test, seasonality of visits and sales and some other factors. Result is that people are testing on their own (because it is that easy), reaching non-valid results while being assured they are the winners and that they have just vastly improved their conversion rates, while failing unfortunate as they are not seeing the numbers rise in their bank account. The result is a loss of trust in the whole concept and reverting back to not testing and just doing some random site redesign every couple of years without backing it up with data.

On top of that comes the common misconception that there is a set of “guaranteed” and universal tests that everybody should do and that will and can improve every website: button colours, headline texts, images, etc. Blog posts on the subject of CRO and even whole websites are propagating the idea of the various test cases that are successful and that all of us should implement and test because they will work. First fallacy there is that there are no universal good case practices that work all the time. Second fallacy is that each and every test should be a winner and that the tests that haven’t improved the conversion rates are the losers, which is so far from the good practice because those “losing” tests tell us so much about what we shouldn’t do and are as valid learning experiences as the “winning” tests.

10855075_540475459427499_8055254465530713171_o

 

3. Technical integrations and development

Tag management systems (like Google Tag Manager or Adobe Dynamic Tag Manager) are becoming more popular and needed on a daily basis. Instead of integrating a series of scripts and various event-triggered codes on the website pages, one would put only one script like the previously mentioned services and all the other scripts inside of it. Benefits are the reduced number of inquiries sent to the third party services (you would have only one – Tag manager’s), faster deployment and changes to the code, as well as possibility to do all of this without involving guys from your IT/development department.

Simo Ahava, Phil Pearce and Yehoshua Coren were truly invaluable in demonstrating their solutions and ideas for working with advanced setups for both Google Analytics and Google Tag Manager. Simo Ahava’s GTM Tools, Yehoshua Coren’s Analytics Ninja Tools (in beta currently) and Phil Pearce’s idea for JSON-LD used for SEO (search engine optimisation) and GA integrations and improvements through Google Tag manager are all fantastic solutions for many problems in analytics community.

DSC_3931

 Bonus: Google Search in 2015

We had two presentations about development to follow in 2015 in Google’s Search department, which is obviously one of the most important influencers in the internet sphere.

2015 will be the year when mobile experience will be important factor for ranking. Building websites which are responsive or adapted to mobile will be required for high rankings in search results. HTTPS as the secure communications protocol will be increasingly important as well for search rankings. There are a couple more inside information to share, but I’ll respect our speaker from Google who told us to not spread the word about some other details. :)

But, I can share one of my opinions and strong impressions after the conference about Google’s changes.

Google Maps began covering interiors of the objects like shops, restaurants, hotels, and if we allow them tracking on our Android devices, they are able to know where we were, even which product we were looking at in a store. On one side, this will improve Google users’ experience because we will be getting more focused and relevant ads and offers, we analytics guys will be able to build better products based on more information about behaviours of the users, but on the other side, why should anyone know all about our browsing and shopping habits? Especially when that someone is bragging about protecting each of our privacies.

Instead of a summary

Being held on a snowy mountain top hotel east from Budapest, Hungary, Superweek is truly a high quality conference where one can hear various opinions on relevant topics from the analytics field and for sure a good networking event. My job now is to implement as many of tips and knowledge gained into our own ManageWP which I surely look forward to executing.