Over-specific Targeting and Segmentation

 In my previous posts, I've talked about how you can analyze website data, or A/B test data, and use it to identify winning segments of users.

However, your wonderful new test design may be an unfortunate loser.  It may have shown dreadful results - all the KPIs you can think of are in the red.  There's no way to salvage this one, every single metric shows a drop for Recipe B.  And I suppose you have two options:  learn from the data and try again, or segment the data and see why it lost.

And so begins a trip down a very dangerous rabbit-hole.  

"If we look at new versus return visitors, we find that return visitors didn't perform as badly as new visitors."

"And if we look at return visitors who were visiting on a mobile device instead of a laptop or desktop, then we see that performance is actually slightly better."

"And if we look at return visitors who visited on a mobile device and were looking for our higher-price products, then we actually see an improvement."

Great.  But after three rounds of filtering, targeting or segmenting (your choice of terminology), you've gone from 50% of traffic (the test population) down to 4.3% of traffic.  Is it really worth spending time, effort and energy to provide 4.3% of your traffic with a unique experience?  If you're a luxury brand like Rolls Royce, or Beaverbrooks, then possibly, yes.  If you're selling discounted pet supplies, perhaps not.

But we can get into this level of detail with our targeting and personalisation campaigns too.  I've previously talked about the challenges of setting up a personalisation campaign - the first is obtaining and analysing the data, the second is having the content to present to user segments.  But assuming we can make a decent effort at both, we don't want to get into too much detail in our segments.  For example:

What do you show in the homepage hero banner?  Or what do you show in your "We think you'll like this..." module?  Do you stand at the front of your virtual store, identifying customers based on data such as previous visits or previous purchases, and say, "I think you'd like to buy this Lego set.  It's not discounted, there's no extra incentive to buy, but we watched you on your previous visit, and we think this Lego set is for you."

Is your targeting that good?

In my experience and conversations with other professionals, Netflix and Amazon are often cited as the leaders in targeting.  "Because you watched Star Trek: Voyager" is a reasonable and transparent explanation of the recommendations that Netflix shows me.  And sure enough, around half of the recommendations are actually interesting to me - some of them I've seen before, some of them aren't my cup of tea.  And when you have the opportunity to present me with 42 options (the screen scrolls horizontally seven times) then you can show me specific examples.  If I don't know what I'm looking for, this is a good place to start.

So you could stand at the front of your virtual toy store and say, "We can recommend these Lego models...." and show 42 from your catalogue.  And why not?

If that's not feasible (perhaps due to challenges with obtaining stock values - there's nothing worse than actively recommending an item that's out of stock) then you can be less specific.  Standing on the virtual front door of your virtual store, you could offer, "Would you like to see our Lego models? Please walk this way" instead of "We think you want this model."  You're more likely to get a positive reaction, for a start; in a world where engagement metrics are the king of the KPIs, you're at least more likely to see better results from being a little less specific.  You can certainly expect to be more successful with a broad recommendation than with no targeting at all.  Compare, for example, "Welcome to our toy shop!  These are our favourite toys!" with "We think you're interested in construction toys."  The first is symptomatic of the "We want to sell you this" which pervades many home-page banners, instead of the notion that we find out what our customers want to buy, and show them that.  I'll leave that one there for now, but at least some level of targeting is better than none (and probably better than over-targeting too).

If we can say with 75% chance that a visitor is looking for Lego models, but only a 23% chance that it's Lego Technic (the advanced, engineering-level Lego), and only a 5% likelihood that it's a Lego Technic Race Car, then perhaps leading with one specific model is too much.  It would be better to suggest the Lego Technic range, and direct users to a category page and let them find their own way from there.
     

Your virtual store could be selling electronics; home appliances; books; streaming TV shows; or whichever, but Lego has the advantage of being widely known globally, and very visual and tangible.  Insert your relevant product subcategory in here (I suppose if I had been paying attention to your browsing habits, I could have personalised the content of the blog to make it more relevant to you.  Maybe next time!).

Other Web Analytics and Testing Articles I've Written

How not to segment test data (even if your stakeholders ask you to)

Web Analytics - Gathering Requirements from Stakeholders

Analysis is Easy, Interpretation Less So
Telling a Story with Web Analytics Data
Reporting, Analysing, Testing and Forecasting
Pages with Zero Traffic

Personalization, Segmentation or Targeting

Following all my recent posts on targeting (or personalization), I was discussing website content changes with a colleague. I was explaining how we could test some form of interactive, real-time changes on our site.  His comments were that this wasn't real 1-to-1 personalization and what I was actually doing was just segmentation and content retargeting.  This started me thinking, and so I'd like to share my thoughts on 1-to-1 targeting is possible, easy and worth the effort.  Or should we be satisfied with segmentation and retargeting?

1-to-1 targeting requires the ability to show any content to any user. It probably needs a hige repository of content that can be accessed to show content that isn't shown to other users, but which is deemed optimal for a particular user.

1. How do you decide which type of user this particular user should be classed as?  

2.  How do you determine which content to show this particular user (or type of user)?

3.  When the targeting doesn't give great results, how can you tell if the problem is with 1. or 2.?

And, as a follow-up question, why is "targeted" content drawn from a library held in higher esteem than retargeting existing content? Is it better because it's so difficult to set up?

Content retargeting - moving existing content on the page - does not require new content, but "isn't real 1-to-1 targeting."  This is true, but I would argue that the difference - mathematically at least - is negligible.  The huge library of targeted content isn't going to be able to match the potential combinations of content that can be achieved just by flipping page content around to promote a particular group of products.

In previous examples on targeting, I've looked at having four product categories that can be targeted.

How many combinations are there for the four products A, B, C, D?

4 * 3 * 2 * 1 = 24

There are four options for the first placement, leaving three for the second placement, two options for the third and only one left for the final place.

This is a relatively simple example - most websites have more than just four products or product categories in their catalogue (even Apple, with its limited product range, has more than four).

Let's jump up to six products:

6 * 5 * 4 * 3 * 2 * 1 = 720.

At this point, retargeting is going to start scaling far more easily than 1-to-1 personalization. 

Admittedly, it's highly unlikely that all 720 combinations are going to be used and shown with equal probability - we will probably see maybe 6-10 combinations that are shown most often, as users visit just one or two product categories and identify themselves as menswear, casual clothes, or womenswear customers.  The remaining three or four categories aren't relevant to these customers, and so we don't retarget hat content.  I mean: if a user is visiting menswear and men's shoes, then they aren't going to be interested in womenswear and casual clothing, so the sequence of those categories is going to be irrelevant and unchanged.

So, we can group users into one of 720 "segments", not based on how we segment them, but how they segment themselves.  This leads to a pseudo-bespoke browsing experience (it isn't 1-to-1, but the numbers are high enough for it to be indistinguishable) that doesn't require the overhead of a huge library of product content waiting to be accessed.

When does the difference between true personalization and segmented retargeting become indistinguishable?  Are we chasing true 1-to-1 personalization when it isn't even beneficial to the customers' experience?

I would say that it's when the number of combinations of retargeted content becomes so large that users are seeing a targeted experience each time they come to the page.  Or, when the number of combinations is greater than the number of users who visit the page.  Personalization is usually perceived - and presented - as the holy grail of Web experience, but in my view it's unnecessary, unattainable and frequently unlikely to actually get off the drawing board. Why not try something that could give actual results, provide improved customer experience and could be set up this side of Christmas?

Other Web Analytics and Testing Articles I've Written

How not to segment test data (even if your stakeholders ask you to)

Web Analytics - Gathering Requirements from Stakeholders

Analysis is Easy, Interpretation Less So
Telling a Story with Web Analytics Data
Reporting, Analysing, Testing and Forecasting
Pages with Zero Traffic

Why Personalisation Programs Struggle

So why aren’t we living in a world of perfect personalisation? We've been hearing for a while that it'll be the next big thing, so why isn't it happening?

Because it’s hard.  There's just too much to consider, especially if you're after the ultimate goal of 1-to-1 personalisation.

In my experience, there are three areas where personalisation strategies come completely unstuck.  The first is in the data capture, the second is the classification and design of ‘personas’, and the third is in the visual design.

1. Data capture:  what data can you access?

Search keywords?
PPC campaign information?
Marketing campaign engagement?
Browsing history?
Purchase history?
Can you get geographic or demographic information?
Surely you can’t form a 1x1 relationship between each individual user and their experience? 
Previous purchaser?  And are you going to try and sell them another one of what they just bought?
Traffic source:  search/display/social?
What products are they looking at?
What have they added to basket?

2. Classification:  how are you going to decide how to aggregate and categorise all this data?  

Is it a new user?  Return user?

And the biggest crunch:  how are you going to then transfer these classifications to your Content Management System, or to your Targeting engine, so that it knows which category to place User #12345 into.  And that’s just where the fun begins.

And how do you choose the right data?  I'm personally becoming bored of seeing recommendations based on items I've bought:  "You bought this printer... how about this printer?" and "You recently purchased a new pair of shoes... would you like to buy a pair of shoes?" As an industry we seem to lack the sophistication that says, "You bought this printer - would you like to buy some ink for it?" or "You bought these shoes, would you like to buy this polish, or these laces?"

3. Visual Design

For each category or persona that you identify, you will need to have a corresponding version of your site.  For example, you’ll need to have a banner that promotes a particular product category (a holiday in France, the Caribbean, the Mediterranean, the USA); or you may need to have links to content about men’s shoes; women’s shoes; slippers or sports shoes. 

And your site merchandising team now needs to multiply its efforts for its campaigns. 

Previously, they needed one banner for the pre-Christmas campaign; now, they need to produce four, five or more instead.  This comes as they are approaching their busiest period (because that’s when you’ll get more traffic in and want to maximise its performance) and haven’t got time to generate duplicated content just for one banner.

Fortunately, there are ways of minimizing the headaches that you can encounter when you’re trying to get personalization up and running (or keeping it going).

Why not take the existing content, and show it to users in a different order?  Years ago, there was a mantra (with a meme, probably) going around that told us to 'Remember: There is no fold' but I've never subscribed to that view.  Analytics regularly shows us that most users don't scroll down to see our wonderful content lying just below the edge of their monitor (or their phone screen).  So, if you can identify a customer as someone looking for men's shoes, or women's sports shoes, or a 4x4, or a hatchback, or a plasma TV, then why not show that particular product category first (i.e. above the fold, or at least the first thing below it)?

4. Solutions

The flavour du jour in our house is Airfix modelling - building 1/72 or 1/48 scale vehicles and aircraft, so let's use that as an example, and visit  one of the largest online modelling stores in the UK, Wonderland Models.

Their homepage has a very large leading banner, which rotates like a carousel around five different images: a branding image; radio controlled cars; toy animals and figures; toys and playsets; and plastic model kits.  The opportunity here is to target users (either return visitors, which is easier, or new users, which is trickier) and show them the banner which is most relevant to them. 

The Wonderland Models homepage.  The black line is the fold on my desktop.

How do you select which banner?  By using the data that users are sharing with you - their previous visits, items they've browsed (or added to cart), or what they're looking for in your site search... and so on.  Here, the question of targeted content is simpler - show them the existing banner which closest matches their needs - but the data is trickier.  However, the banners and categories will help you determine the data categorization that you need to - you'll probably find this in your site architecture.

However, here's the bonus:  when you've classified (or segmented) your user, you can use this content again... lower down on the same page.  Most sites duplicate their links, or have multiple links around similar themes, and Wonderland Models is no exception. Here, the secondary categories are Radio Control; Models and Kits; Toys and Collectables; Paints, Tools and Materials; Model Railways and Sale.  These overlap with the banner categories, and with a bit of tweaking, the same data source could be used to drive targeting in both segments. 

As I covered in a previous blog about targeting the sequence of online banners, the win here is that with six categories (and a large part of the web page being targeted), there are thirty different combinations for just the first two slots, with six options for the first position, and five for the second.  This will be useful as the content is long and requires considerable scrolling.

The second and third folds on Wonderland Models.  The black lines show the folds.

Most analytics packages have an integration with CMS’s or targeting platforms.  Adobe Analytics has Target, which is its testing and targeting tool.  It's possible to connect the data from Analytics into Target (and I suspect your Adobe support team would be happy to help) and then use this to make an educated guess on which content to show to your visitors.  At the very least, you could run an A/B test.

5. The Challenge

The main reason personalization programs struggle to get going is (and I hate to use this expression, but here goes) that they aren't agile enough.  At a time when ecommerce is starting to use the product model and forming agile teams, it seems like personalization is often stuck in a waterfall approach.  There's no plan to form a minimum viable product, and try small steps - instead, it's wholesale all-in build-the-monolith, which takes months, then suffers a "funding reprioritization" since the program has nothing to show for its money so far...  this makes it even harder to gain traction (and funding) next time around.

6. The Start

So, don't be afraid to start small.  If you're resequencing the existing content on your home page, and you have three pieces of content, then there are six different ways that the content can be shown.  Without getting into the maths, there's ABC, ACB, BAC, BCA, CAB and CBA.  And you've already created six segments for six personas.  Or at least you've started, and that's what matters.  I've mentioned in a previous article about personalization and sequencing that if you can add in more content into your 'content bank' then the number of variations you can show increases exponentially.  So if you can show the value of resequencing what you already have, then you are in a stronger position to ask for additional content.  Engaging with an already-overloaded merchandising team is going to slow you down and frustrate them, so only work with them when you have something up-and-running to demonstrate.

Remember - start small, build up your MVP and only bring in stakeholders when you need to.  If you want to travel far, travel together, but if you want to travel quickly, travel light!

Combinations and Permutations

PERMUTATIONS AND COMBINATIONS

After mentioning permutations and combinations in my previous blog post on targeting, I thought it was time to provide a more mathematical treatment of them.  Everybody talks about them as a pair (in the same way as people tend to say 'look and feel', or 'design and technology').  

Let's start with an example:  three banners are to be shown on a website homepage. If we simplify and call the different pictures A, B and C, then one order in which they can be hung is A, B, C and another is A, C, B.

Each of these arrangements is called a permutation of the three pictures (and there are further possible permutations), i.e, a permutation is an ordered arrangement of a number of items.

Suppose, however, that seven banners are available for presenting on the website, and only three of them can be displayed. This time a choice has first to be made. If we call the seven banners A, B, C, D, E, F and G, one possible choice of the three pictures for display is A, B, and C - ignoring the sequence of the banners. Regardless of the order in which they are then hung this group of three is just one choice and is called a combination.

A, B, C
A, C, B
B, A, C
B, C, A
C, A, B
C, B, A

are six different permutations; but only one combination - thus:  a combination is an unordered selection of a number of items from a given set.

In this post,  I will discuss methods for finding the total number of ways of arranging items (permutations) or choosing groups of items (combinations) from a given set. But before we do so it is critical that we're able to distinguish between permutations and combinations.  They are not the same, and the terms shouldn't be used interchangeably.

For example:  a news website has ten news articles on its site, but the home page layout means that only five can be shown, in a vertical column. While they cannot display all ten of the articles, they must choose a group of five. The order in which the site selects the five articles is irrelevant (in this case); the set of five is only one combination. Once they have made the choice, they are then able to place the five articles in various different orders on the display stand. Now the site team are arranging them and each arrangement is a permutation, i.e a particular set of five articles is one combination, but that one combination can be arranged to give several different permutations.

1.  The King's Health is Failing
2.  Peace Treaty Signed!
3.  Life found on Mars!
4. Bungled Theft on the Railway
5. Jack the Ripper
6. Reports of My Death Greatly Exaggerated
7. Lottery Winner Buys Football Team
8.  New 007 is a Woman
9. Crop Circles - The Answer
10. Price of Eggs falls 10%

In each of these examples, decide if the question is asking for a number of permutations, or a number of combinations.

How many arrangements of the letters A, B, C are there?
Arrangements means the sequence is important, so this means permutations.

A team of six members is chosen from a group of eight. How many different

teams can be selected?
The sequence is not important, so this means combinations.

A person can take eight records to a desert island, chosen from his own

selection of one hundred records. How many different sets of records could he choose?
Different sets, again the sequence is not critical, so these are combinations.

The first, second and third prizes for a raffle are awarded by drawing tickets

from a box of five hundred. In how many ways can the prizes be won?
Here, there's a difference between the order (or sequence, or arrangement) of the three prizes, so we're looking at permutations.

Combinations:  the sequence is not important.
Permutations:  the sequence is important.

Other reading you may find interesting:

If you're interested in how to use this to improve your website, I can recommend this article on personalisation and targeting and this one on why personalisation programs struggle (hint: they don't make good use of maths).

I've also written a more practical article on how to use combinations and permutations, looking at Targeting Website Banners.

Alternatively, if you like the maths of combinations and permutations, I can suggest Multiplications Puzzles

Targeted Banners: A study in permutations

"How are our banners performing?"

It's a question I'm being asked increasingly frequently, as we step up our on-site marketing. And banners are banners: they've been around for years; customers are accustomed to them (and possibly tired of them) and the challenge is to make them fresh, useful, relevant and just plain interesting. However, banners are easy, straightforward and simple to execute, measure and analyse.  You didn't think I was going to recommend adding banners without commenting on KPIs for them?

"Which banners are doing the best?"

So the challenge becomes: how do we make sure we have the right banner for the right customers?  How do we drive clicks and - more importantly - increase revenue?

Some form of targeting helps, and this could be keyword, or geo targeting, or behavioural targeting.  But why not try promoting multiple products instead of just one (or just one family of products)?  If you can split a banner slot into two, you can promote twice as many products in the same space.  And, if you can use some form of targeting, then the options for what to show increase significantly. 

Let's take an online toy retailer as an example.  What would you promote on your home page if you were an online toy shop?

There are several categories you might want to feature:

Construction toys

Dolls

Dinosaurs 

Cars

Robots

Board games 

Outdoor toys 

However, your website design only allows space (or 'real estate') for three.  And besides, you've found that having more than three diluted the effectiveness of them - your visitors get "banner blindness".  So which three do you show?  

You could determine which three to show based on various factors:

What does the customer search for?  If they search for "Lego" or "Jurassic Park", then they'll probably appreciate the banners for construction toys and dinosaurs.  Setting up some form of tracking on search usage across the site, and then matching this to the banner categories will enable you to show content that's more likely to be appealing to your visitors.

Better still, what was the inbound keyword that your user searched for?  I know Google doesn't share natural search terms (for some strange reason, the tracking only applies to Google paid search terms), but if you can access the search term that the visitor used when she came to your site, then you can start targeting her from the moment she arrives on your site, and that's a key advantage.

Alternatively, which marketing campaign did she click when she came to your site?  Was it an email that advertised your range of sports toys?

Or which pages has this visitor been viewing? If they're browsing your pages on Barbie, Sindy and similar, then this gives you a better indication of her purchase intent.

What has this customer purchased before? This will take longer for your targeting to initialise as you'll have to wait for your customer's first purchase, and once it's running it will be less dynamic than the other methods, but will be more specific as you'll know that this user has made a purchase in this category before.  The segment will be smaller, but have a higher likelihood to purchase.  Thinking outside the toy store example to other industries, maybe you could target your banners based on items added to basket (cart); videos viewed on site (what were they promoting); PDF downloads and other success events on your site.

So, you can see, there are various options for how you target, and you can then determine how effective each method is, through testing.  The model doesn't have to be perfect, it just has to be testable, and you can be confident that your testing model will go some way towards making your site more relevant (and better converting) for your visitors.  After all, there are so many variations, surely there's a good chance that you'll find a better set of banners than the one you show to all your generic customers.

So, how many permutations are there?

In our example, there are three slots available on the site, and we have seven different banners we can show.  Here are some example images, taken from various online sites.  In practice, these would be more uniform in design and messaging.

We have three slots in total:

The first slot could be filled by one of seven images.

The second slot can be filled by one of the remaining six images.

The third slot can be filled by one of the five images that we haven't used yet.

7 * 6 * 5 = 210 permutations (we use permutations here, because we can't use the same banner in two slots - that would give us combinations, and would be an even larger number).

If we decided we wanted to use only six of the banners - for example if we decide that board games aren't relevant any more - then the calculation would be:

6 (for the first slot) * 5 (for the second slot) * 4 (for the third slot) = 120

Conversely, if we introduced soft toys as an extra variation, so that we had a bank of eight banners altogether, then we'd have:

8 * 7 * 6 = 336

This is an example of "permutations without repetition" - the sequence of the banners is important, and we only show each banner once (we don't repeat them).  I recommend this site for more on the calculations of the number of permutations.  I've also written an article explaining the difference between combinations and permutations (they're not the same thing). The short answer is that the more slots you have, and more banners you have, the more permutations there are (significantly increasing with scale), and the greater the likelihood of showing the best banners to your users.

So, target your banners - you'll be able to dynamically target your content to your users, and start to reduce the guesswork from your marketing.  Even the smallest increases in possible locations or banners will rapidly improve your chances of presenting the ideal banner (if not the ideal permutation) to your users.

Web Analytics and Testing: Summary so far

It's hard to believe that it's two years since I posted my first blog post on web analytics.  I'd decided to take the step of sharing a solution I'd found to a question I'd once been asked by a senior manager:  "Show me all the pages on our site which aren't getting any traffic."  It's a good question, but not one that's easy to answer, and as it happened, it was a real puzzler for me at the time, and I couldn't come up with the answer quickly enough.  Before I could devise the answer, we were already moving on to the next project.  But I did find an answer (although we never implemented it), and thought about how to share it.

Nevertheless, I decided to blog about my solution, and my first blog post was received kindly by the online community, and so I started writing more around web analytics - sporadically, to be sure - and covering online testing, which is my real area of interest.

Here's a summary of the web analytics and online testing posts that I've written over the last two years.

Pages with Zero Traffic
Here's where it all started, back in May 2011, with the problem I outlined above.  How can you identify which pages on your site aren't getting traffic, when the only tools you have are tag-based (or server-log-based), and which only fire when they are visited?

Web Analytics - Reporting, Forecasting, Testing and Analysing
What do these different terms mean in web analytics?  What's the difference between them - aren't they just the same thing?

Web Analytics - Experimenting to Test a Hypothesis
My first post dedicated entirely to testing - my main online interest.  It's okay to test - in fact, it's a great idea - but you need to know why you're testing, and what you hope to achieve from the test.  This is an introduction to testing, discussing what the point of testing should be.

Web Analytics - Who determines an actionable insight?
The drive in analytics is for actionable insights:  "The data shows this, this and this, so we should make this change on our site to improve performance."  The insight is what the data shows; the actionable part is the "we should make this change".  If you're the analyst, you may think you decide what's actionable or not, but do you?  This is a discussion around the limitations of actionability, and a reminder to focus your analysis on things that really can be actionable.

Web Analytics - What makes testing iterative?
What does iterative testing mean?  Can't you just test anything, and implement it if it wins?  Isn't all testing iterative?  This article looks at what iteration means, and how to become more successful at testing (or at least learn more) by thinking about testing as a consecutive series, not a large number of disconnected one-off events.

A/B testing - A Beginning
The basic principles of A/B testing - since I've been talking about it for some time, here's an explanation of what it does and how it works.  A convenient place to start from when going on to the next topic...

Intro To Multi Variate Testing
...and the differences between MVT and A/B.
Multi-Variate Testing
Multi Variate Testing - MVT  - is a more complicated but powerful way of optimising the online experience, by changing a multitude of variables in one go.  I use a few examples to explain how it works, and how multiple variables can be changed in one test, and still provide meaningful results.  I also discuss the range of tools available in the market at the moment, and the potential drawbacks of not doing MVT correctly.

Web Analytics:  Who holds the steering wheel?
This post was inspired by a video presentation from the Omniture (Adobe) EMEA Summit in 2011.  It showed how web analytics could power your website into the future, at high speed and with great performance, like a Formula 1 racing car.  My question in response was, "Who holds the steering wheel?" I discuss how it's possible to propose improvements to a site by looking at the data and demonstrating what the uplift could be, but how it all comes down to the driver, who provides the direction and, also importantly, has his foot on the brake.

Web Analytics:  A Medical Emergency
This post starts with a discussion about a medical emergency (based on the UK TV series 'Casualty') and looks at how we, as web analysts, provide stats and KPIs to our stakeholders and managers.  Do we provide a medical readout, where all the metrics are understood by both sides (blood pressure, temperature, pulse rate...) or are we constantly finding new and wonderful metrics which aren't clearly understood and are not actionable?  If you only had 10 seconds to provide the week's KPIs to your web manager, would you be able to do it?  Which would you select, and why?

Web Analytics:  Bounce Rate Issues
Bounce rate (the number of people who exit your site after loading just one page, divided by all the people who landed on that page) is a useful but dangerous measure of page performance.  What's the target bounce rate for a page?  Does it have one?  Does it vary by segment (where is the traffic coming from? Do you have the search term?  Is it paid search or natural?)?  Whose fault is it if the bounce rate gets worse?  Why?  It's a hotly debated topic, with marketing and web content teams pointing the finger at each other.  So, whose fault is it, and how can the situation be improved?

Why are your pages getting no traffic?
Having discussed a few months earlier how to identify which pages aren't getting any traffic, this is the follow-up - why aren't your pages getting traffic?  I look at potential reasons - on-site and off-site, and technical (did somebody forget to tag the new campaign page?).

A beginner's social media strategy
Not strictly web analytics or testing, but a one-off foray into social media strategy.  It's like testing - make sure you know what the plan is before you start, or you're unlikely to be successful!

The Emerging Role of the Analyst
A post I wrote specifically for another site - hosted on my blog, but with reciprocal links to a central site where other bloggers share their thoughts on how Web Analytics, and Web Analysts in particular, are becoming more important in e-commerce.

MVT:  A simplified explanation of complex interactions

Multi Variate Testing involves making changes to a number of parts of a page, and then testing the overall result.  Each part can have two or more different versions, and this makes the maths complicated.  An additional issue occurs when one version of one part of a page interacts (either supports or negates) with another part of the page.  Sometimes there's a positive reinforcement, where the two parts work together well, either by echoing the same sales sentiment or by both showing the same product, or whatever.  Sometimes, there's a disconnect between one part and another (e.g. a headline and a picture may not work well together).  This is called an interaction - where one variable reacts with another - and I explain this in more detail.

Too Big Data
Too big to be useful?  To be informative?  It's one thing to collect a user's name, address, blood type, inside leg measurement and eye colour, but what's the point?  It all comes back to one thing:  actionable insights.

Personalisation
The current online political topic:  how much information are web analysts and marketers allowed to collect and use?  I start with an offline parallel and then discuss whether we're becoming overly paranoid about online data collection.

What is Direct Traffic?
After a year of not blogging about web analytics (it was a busy year), I return with an article about a topic I have thought about for a long time.  Direct traffic is described by some people as some of the best traffic you can get, but my experiences have taught me that it can be very different from the 'success of offline or word-of-mouth marketing'.  In fact, it can totally ruin your analysis - here's my view.

Testing - Iterating or Creating?
Having mentioned iterative testing before, I write here about the difference between planned iterative testing, and planned creative testing.  I explain the potential risks and rewards of creative testing (trying something completely new) versus the smaller risks and rewards of iterative testing (improving on something you tested before).

And finally...

A/B testing - where to test
This will form part of a series - I've looked at why we test, and now this is where.  I'll also be looking at how long to test for, and what to test next!

It's been a very exciting two years... and I'm looking forward to learning and then writing more about testing and analytics in the future!

Web Analytics: Personalisation

Last Friday night, I had to transfer some money from my savings account to my current account, and in the process encountered an interesting case of personalisation.


Withdrawing the cash from the savings at the building society was a typically anonymous matter, even though I had to provide my account passbook and photo ID, but this only became apparent when I paid the money into my bank, just across the road.  I only had to provide the money and the debit card for my bank account, but as soon as my card had been scanned, the bank clerk began addressing me as David, and just by doing that, provided a much more personal service.


Earlier in the evening, I phoned the local take-away restaurant, and on the way back from the bank, I called in to pick up my order. I'd called them from my home landline, but hadn't provided a name or address.  However, I've ordered from the take-away before, and they'd evidently stored my data: at the top of the receipt for my order were my full name and address.  As I mentioned, I hadn't provided any information at all when I phoned the order through.  Was it surprising to see my name and address on the receipt?  Absolutely. Was it un-nerving?  Perhaps, but it's more a reflection of a local business using data and information to their advantage.  I don't know if they're going to use my purchase preferences to offer me particular choices or offers next time I order... I'll let you know.


Online, I'm not surprised when Amazon, or eBay, or any other e-commerce site, uses my login details and my activity on their site to try to provide me with relevant content or advertising.  So I've been searching for a particular author, or a particular album, movie or laptop - should I really be surprised that they've noticed, and now they're using the promotional space on their sites to show me advertising of similar products?  Is this scary new technology?  Or is it something that's been around for many years, and this is just its newest incarnation?


Back when I was at high school, I had a part time job as a sales assistant at the local shoe store.  It was easy enough - serve the customers, keep the shop floor well-stocked, tidy away surplus stock into the storage room.  Part of the sales training (it wasn't extensive) was to try to cross-sell - shoelaces, polish, all that stuff, and to sell to customers when we didn't have what they wanted.  For example - "Do you have this shoe in my size?"  A quick trip to the stock room would reveal that we didn't, but a check around the shelves would show that we had it in blue, or brown instead.  Or perhaps, if it was a shoe that looked like it was for the office, did we have a similar style.  Was it good customer service?  Was it personalisation?  I would certainly hope so, as it led to me selling many pairs of shoes (and frequent declines, but that was part of the job).  Did customers question how I'd manage to come with potential alternatives?  Did they marvel at the apparent depths of the stock room, or think it was freaky or scary that I'd been able to anticipate their needs, based on just one query?


Perhaps, then, we shouldn't be surprised, or alarmed, when a computer algorithm looks at our on-site browsing habits and tries to provide us with what we appear to be searching for.