Timothy Yuen -
Wednesday, March, 09, 2016
Those of us who spend their working lives up to their elbows in Google Analytics – like I do – take a lot of its great functions and quirky complications for granted.
So I’ve decided that now is a good time to start sharing some of my knowledge to help you get the most out of this incredibly powerful tool. There will be many interesting conversations to come!
To start with, I have recently been looking at the User ID setting for session unification and I thought I’d share a recent discovery with you. It will help you understand some of those mysterious gaps in your User ID view.
While doing some testing, I’ve discovered a User ID issue, related to session unification, that many of you may not even notice. Until now, I had always thought that all hits before the User ID is set would be unified as coming from the same User. It turns out that this is incorrect.
Session unification only happens for all hits before the first hit containing the User ID. Any hits without the User ID, after the initial hit, are not counted.
What are the implications of this you ask?
Well, imagine you are tracking the User ID when a user logs onto your site. That user logs out, or times out, browses a few pages and then they log back in – still within a session. Their User ID will not be logged in the few pages they browsed after logging out. And any subsequent pages they view or events without the User ID will no longer be tracked or unified into the session.
I had a sift through the Google Analytics documentation, looking for some clarity, and all I could find was this cryptic sentence:
Now that we know this, what should we do?
As per the Google Analytics statement, ALL hits should contain the User ID once it is known. But if it is not known or tracked, you may face some sticky questions about why certain pages or data are missing in the User ID view.
You’re probably wondering, “how on earth do I go about explaining that missing (in between) data to senior management?” And it’s all about the user choosing to maintain their privacy.
If the user has chosen to ‘log out’, then their subsequent pages/events aren’t tracked. By logging out they have opted out from being tracked.
I believe that this is the basis of the Google Analytics logic for not automatically tying together all of a user’s sessions, once their user ID is identified. The missing data is adhering to, and protecting, the privacy of the user.
Rod Jacka -
Thursday, December, 12, 2013
One of the podcasts that I regularly subscribe to has a great interview with Eric Siegel a leading figure in the area of predictive analytics on the topic of the benefits of this area and the potential risks to privacy.
I attended one of Eric Siegel’s classes in Washington DC a few years back and have followed his conference Predictive Analytics World since.
The interview is a nice balance on the benefits that predictive analytics can provide and the risks to our privacy. As a practitioner in this area I am very sensitive to how data can be used for both good and not so good.
Eric’s recent book (Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die) features the case of Hewlett Packard and their use of predictive analytics to score their employees based on their likelihood to look for another job. Eric’s article in the recent Analytics Magazine is a must read on the downside of this approach. Another better known case of the potential for privacy infringement is the now famous Target pregnancy prediction case which started with a presentation at the Predictive Analytics World conference.
In my view the privacy of the individual must be respected with any analytics initiative and that privacy implications must be carefully thought through.
Rod Jacka -
Friday, November, 08, 2013
Communicating complex and varied data in a useful and meaningful way is always a challenge. How can you convey multidimensional data in a way that allows viewers to take in the whole picture at a glance?
As a dedicated audiobook listener I regularly scour Audible.com for interesting books to listen to. I’m investing 4 to 10 hours of my life in each one, so I need to know whether a book is going to be worth it. For each book, Audible provides charts, based on audience feedback, that provide very detailed data in a simple and comparable way.
Each chart allows me to see how others responded to the book, providing me with both the average (the stars at the top), and the distribution (the bars below), across three dimensions. It also provides the number of respondents in each group.
The colour scheme of the Audible charts is also helps viewers interpret the data meaningfully — with the overall average rating in red and the detailed information in grey. This is very much in keeping with what Stephen Few and others recommend for the design of charts like this.
How do other online retailers show this kind of data?
Audible’s parent company Amazon provides a similar chart, but only for a single dimension. This limits the value of the chart as an aid to decision making — it’s not clear what aspects of the book were most successful for these reviewers.
Australian online store OO.com.au only provides an average rating which doesn’t allow me to see how many people highly rated the product and how many didn’t. While I like the pros, cons and best uses categories, the layout of this chart causes the viewer to do a lot more work than they should have to.
US clothing retailer LLBean.com uses a popup, restricting the view of the detailed breakdown to those who roll over the average star rating.
Getting your charts and reports right is a key impact factor in your success, whether you are selling audiobooks, or reporting to the C-suite. Spending time on making your data clear, compelling and instantly comprehensible will always be a good investment.
Rod Jacka -
Tuesday, May, 28, 2013
My good friend Daniel Waisberg has one of my articles at Online Behaviour on how I see creativity is a key component in your analytics skill set.
Learn how to be both an artist and an engineer and how both of these skills are necessary for you to excel as an analyst.
Rod Jacka -
Monday, March, 05, 2012
In this article we’re discussing the impact of seasonality in your online marketing strategy. Not all industries enjoy an endless stream of online traffic, so making provision for the spikes and troughs lows applies to online traffic, as well as offline sales.
Whilst an organisation is hard pressed to forecast them absolutely, predicting seasonal variations as distinct from a normal trend is part of strategic planning. A seasonal variation may be anticipated because of social custom (e.g. Christmas) but closing the accuracy gap between the expected variation and actual becomes the challenge.
How can we leverage the positive and minimise the negative aspects of online seasonality?
1. Plan correctly
As always, planning is paramount. Some questions which might feature in your plan include;
- What is the incubation period for a sale during the season in question?
- How far ahead do your clients make their purchase?
- How far in advance of that date are they doing their research to aid the buying decision?
- Is a branding campaign – for the particular season – required if people don’t actively search on keywords?
- Marketing creep is the sometimes vexing practice of advertisers bringing forward sales period of the particular season and examples might include:
- Christmas decorations on sale in August
- Chocolate Easter eggs displayed in January
- TV coverage in mid-summer of the start of Winter football training
The challenge is how far can this be advanced without irritating/ offending your prospects?
Has your creative been varied to reflect the season? Does it need a new call-to-action for the season? (Remember it will need changing again immediately after the season!)
Does a competitor’s offer need to be matched or bettered to keep you in the race?
2. Budget correctly
Which serves your needs better? Higher – or steady – spending if the season brings with it more sales
What seasonal sales spikes which occur regardless? There are examples of predictable highs
- Activity for tax agents at EOFY
- Back-to-school supplies & shoes in January
- Fitness clubs & tanning salons at end-Winter
- Auto service centres, at start of school holidays
And predictable lows
- Chocolate sales immediately after Easter
- Slower retail sales prior to EOFY (stock taking)
- Family holiday locations at the start of the new school year
- Party hire suppliers immediately after Christmas/ New Year
Just like the winter ski resort marketing their ‘off’ season as a cooling Summer escape, could marketing spend be channelled to.
What is the year-round average sale value? (How much is the average sale off-season?)
How does it compare to the seasonal average sales value? (Does the average customer buy more in season?)
Can the target CPA be increased? (If yes, could provision be made for a higher CPA inside the season)
Which categories yield best returns? Apportion more budget to higher yielding products/ categories
3. Track, Test & Compare
Comparing the outcome of this year’s seasonal sales (versus forecasts) is the foundation of future forecasting accuracy. It is through the comparison of actual with predicted that insights to future seasonal marketing are offered. This is where well configured web analytics is invaluable.
And a final word: don’t forget to factor your site’s mobile device compatibility into your marketing plans. There’s no escaping the rapid rise of mobile!
Rod Jacka -
Tuesday, January, 10, 2012
The fundamental things marketing professionals want to learn about customers online are similar to what we want to learn offline. For example, we would like to understand and refine our target segments. We would like to better understand the customer journey, demographics, psychographics, consumer behaviour and decision making process, how best to differentiate, propensity to buy, seasonality factors etc. Similarly to direct mail, the web can provide you with timely data to demonstrate and quantify the effectiveness of campaigns.
We have a feedback mechanism through analytics that can tell us whether or not we are on track with our marketing objectives – provided the data is treated in the correct manner (e.g. ‘comparing apples with apples’). As the ability to measure online activity with analytics increases we can also learn additional things about our customers relating to buying behaviour. For example:
- Understanding the language that customers use and where they are in the buying cycle by examining the keywords that they use to find the site and search its contents;
- Watching what customers do versus what they say;
- Learning what offers visitors respond to best. Small changes to a web page can make big differences in results;
- Using visitor behaviour information to create targeted offers, identify the propensity to purchase and estimate the sales pipeline.
Successful companies who are ahead of the pack are now investing in learning about and understanding the online customer. They realise that converting digital data into insights has currency. As customers become more empowered the emphasis will be on understanding the individual in real time and responding accordingly.
Further reading on the shift to data-driven marketing and the implications of the empowered consumer can be found in the IBM CMO Study