Visual Saliency Vs Eye tracking Heatmaps

Share

Visual saliency programs exist that are intended to be a cheap alternative, or preliminary step to eyetracking. The basic premise is that our visual attention is drawn to areas of high visual salience, so without testing actual people, we can figure out what people are going to look at in a visual display.

A free version of this software is Fen Gui. On their website they have a very clear explanation of the different visual elements such as contrast, colour, intensity etc. that contribute to visual saliency. Also presented is a comparison with actual eye tracking data. The similarity shown is fairly compelling, but we decided to conduct our own tests… after all if this gives the same results as eye tracking, why would anyone pay the time and expense for eye tracking?

We compared saliency heatmaps with eyetracking heatmaps across a broad range of mediums, webpages, emails, print advertising, packaging, and outdoor scenes. Not surprisingly, the mediums with text were extremely different as the saliency heatmaps do not take into account people’s motivations, but focuses just on potential visual engagement. This isn’t particularly interesting, what is more interesting is how the saliency heatmaps compare to the eyetracking heatmaps for more purely visual mediums such as packaging and outdoor scenes.

Cuing customers to look at your key messages

Share

This is an example we have used for several years and came out of some very early research we did with JCDecaux, but we have never put it into a blog, so I thought I’d revisit it here.

In the research, one of the questions we were exploring was how visual attention can be directed through elements of an advertisement. One such cuing element we investigated was the models visual attention. Most beauty and hair care advertisements follow a basic template of the model looking out engaging with the viewer and a pack shot sharing a significant amount of the ad real estate also. An example of this is the Sunsilk advertisement below, along with a heatmap of 200 people viewing it.

Note that there is very little attention on the pack shot and brand logo. In fact just 6% of the people looked at the pack!

What happens if the model is engaging with the product rather than the viewer? We photoshopped the ad and turned the models eye to the pack shot. See the results below:

This tiny change had a profound effect. 84% of the viewers now looked at the pack! This is 14 times more people looking at the pack. You can also see that there is overall greater engagement with the ad, including more attention to the brand logo at the bottom.

I‘m not suggesting that every ad should have the model looking at the product or key message, but using the model’s gaze can be a very effective way of guiding the viewers gaze and communicating key messages.

There are no hard and fast rules in eyetracking

Share

Skip Fidura from Dot Agency recently posted a great article about eyetracking, “There are no hard and fast rules in eyetracking”.  Please read his article because it’s got some great messages and he really knows what he is talking about form a client’s perspective; he’s seen the ROI from eyetracking pre-tests.

A key point that Skip makes, as the title suggests is that eyetracking will not give you a set of rules to apply to every creative you make.  Eyetracking helps to surface behaviour which can contribute to some guidelines to best take advantage of people’s habits, but generally eyetracking is a diagnostic tool.  It allows creative people to act freely and follow their ideas no matter how off the wall, but then pre-test and make sure the designs are doing their job: standing out and delivering a message.  The ideas can then be taken to the client with some intelligence; some information about which designs work, or which elements of the different design work.

Thanks for your post Skip.

Was Blended Too Good for Google?

Share

Search Engine Strategies is back in London next week and it reminded me of the paper that I gave there last year, I thought it would make a good a blog article, so here it is!

In 2008 Paul McCartney and Heather Mills announced their separation precipitating a lot of press coverage. This coincided with Google’s experimentation with blended search results:

As you can see the blended results get organic positions one, two and nine. We ran a sample of 30 people thru our omnibus study, giving them the task to “Find out more about Paul McCartney”. We left the task deliberately vague because we wanted the users to explore something of interest to themselves. The heatmaps of all 30 users shows that the blended results received a lot of attention:

The problem for Google is that the blended results receive almost all the attention, sucking user gazes away from the sponsored links like little black holes. If users don’t look at a link they can’t even consider clicking it, this is a problem if your revenue model is based on PPC!

A few days before I was due to talk at SES 2008 I thought I’d better check to make sure the data was still current, it wasn’t:

As you can see Google had reduced the number of blended results in the first nine organic places from three to one, de-prioritising the importance of blended.

I did the same search today (10th February 2009) and they have changed the algorithm again:

further de-prioritising the blended results down to organic position five. As we have found in previous research organic listings below rank three are much less valuable that ones in the top three.

We will never know how much revenue Google lost by experimenting with blended results, but I expect it’s more that the $4,000 it would have cost to pre-test the concept with Think. Just goes to show even Google gets it wrong on occasion!

Hollywood & Eyetracking Reveals How To Sell More Magazines

Share

The magazine market is big business!

Size of the UK market:

· 2,600 magazines are sold every minute of every day in the UK

· People in the UK spend around £2bn on magazines each year

· 80% of women regularly read a consumer magazine

· Consumer magazine sales have increased by 4.3% over the past 5 years and the average UK adult now purchases almost 30 magazines every year

(Periodical Publishers Association 2008)

If we asked you what attracts you to a particular magazine on the shelf, you might say ‘the front cover’. If you did, you’d be in agreement with around 75 percent of magazine consumers. The cover of a magazine has enormous influence on most purchasers’ buying decision – all but the most loyal of readers. It can be seen as an ‘advertisement’ for the issue’s content and most women’s mags share a generic layout – a model surrounded by the article headlines.

It is estimated that 60 percent of newsstand purchases are unplanned, and reports indicate that time spent selecting a magazine ranges from forty seconds grazing the newsstand, down to five seconds considering an individual title.

With an impulse buy like this, marketers only have a few critical seconds to appeal to the consumer’s subconscious, to evoke an interest in the magazine and successfully make a sale. Therefore, it is important for marketers to understand what is going on in the consumer’s head when they look at a cover and make their purchasing decision.

How eye tracking can help

Eye tracking offers an objective approach to understanding consumer behaviour. Eye movements can reveal hidden cognitive processes which can help understand and predict purchasing decisions.

To investigate front cover preferences, we conducted an eye tracking experiment using 30 female participants. They were asked to select a magazine they would purchase from eight available. We were particularly interested in researching the effect of two models on the front cover, instead of one and if it would influence the purchaser’s decision. The front covers of ‘Eve’ and ‘Good Housekeeping’ were Photoshopped to remove Patsy and Susannah from the original covers.

Which do you think got the most attention? Which got least?

magazine-covers-not eyetracked

Original (two model) and Photoshopped (one model) covers of Good Housekeeping and Eve magazines.

Heat maps display an aggregation of many individual’s viewing experience. Above, you can see that fixations are generally concentrated around the faces of the models/celebrities. Engaged people often make long fixations, and this is also in line with previous research on non-verbal communication e.g. eye contact indicates interest, attention and involvement. Interestingly, the actual text on the covers is not looked at very much.

N.B - 96% of participants looked at the image of Davina alone compared to just 8% who looked at Patsy and Davina.

96% of participants looked at the image of Davina alone compared to just 8% who looked at Patsy and Davina.

We can see that the image of Davina alone received more attention than the image of both Patsy and Davina (Eve), as demonstrated by the dense area of red. There were also longer fixations on the image of Davina, suggesting higher engagement and cognitive activity taking place here. Perhaps because Patsy and Davina are not normally regarded as a pair, less interest is taken in them.

Trinny and Susannah received more attention that Trinny by herself, because they are normally regarded as a pair. Trinny and Susannah received the most attention of all of the ten magazine covers we tested.

Celebrity parings work better when people believe there is a relationship between the pair – this is a technique Hollywood has been using to great effect for years!

Thanks for reading: http://twitter.com/modestrobert

Why does Think Eyetracking sometimes use a sample of 30?

Share

Everyone knows that the more people you use in any kind of test then the better your sample is likely to represent the population as a whole. This applies to questionnaires, polls, observations, and also where people look when they look at adverts, packaging, homepages, landing pages, emails etc However, in the real world of budgetary and time constraints, you want to test with as few people as possible whilst still being confident that your results are reliable so that you can make decisions with confidence. There are ways that you can determine the appropriate numbers to achieve this – we won’t get into the statistical theory here – but it is important to note that the minimum number of people to use is dependent on what you are measuring. Traditional Market research often uses samples of 100 or more people. At Think Eyetracking we often test 30 people – why is this? Broadly it is because what we are measuring is entirely different; traditional market research usually asks the consumers conscious mind it’s opinion or attitude, at Think we interrogate what the subconscious mind ‘does’ (rather than what the conscious mind it thinks it does!) by recording fixations and saccades. Fixations and saccades are ‘things’ rather than abstract concepts such as opinions and attitudes; because we are counting ‘things’ we can often use smaller samples than traditional market research and achieve statistically reliable results. As a demonstration, here is part of a more formal Monte Carlo Simulation we did:

We asked 150 people to look through the pages of a magazine as they normally would do; there was a mix of advertisements and editorials. The Dove advertisement below was included.

We then created 4 heatmaps from 4 different (randomly assigned) groups of 30 from the total sample of 150.

You can see in these heatmaps above that each sample of 30 produced the similar pattern of visual behavior. And when you run the numbers these are essentially the same too. So that’s a fairly simple explanation, but let’s look a little closer at what’s involved. There are three key elements that can effect the necessary sample size:

  • A) the complexity of the stimuli being shown
  • B) the task being set and
  • C) the variability in the sample

A stimuli with limited complexity and a directive task both limit the need for a larger sample. For example look at figures 1 and 2 below.

It is easy to see that with a less complex stimuli as in figure 1, there will be less variability in people’s visual behavior, and so a smaller sample can be used than would be necessary for figure 2 to find common patterns. Similarly, if the task looking at figure 2 was to find the two black squares, there would be more common behavior than if it was to simply look at the image, and so a smaller sample can be used to find common visual behavior. We have been running commercial eyetracking studies for over six years now (2002 to 2008 at time of writing) and during this time we have got real good at managing elements A through C – stimuli complexity, task, and sample variability – to ensure that a sample of 30 can be enough to provide statistically significant and robust results.

That said, sometimes it is necessary to test with samples greater than 30 and you will be unsurprised to find out that this is when the differences in the things we are testing are very subtle or the image is very visually complex. In these cases we can advise on what will be an appropriate number to sample.

Posted By Lizzie Maughan and Robert Stevens.

Credit Crunch, Theft and Checks

Share

As Christmas is almost here I thought is a good time to revist a blog from last year looking at credit rating agencies Equifax and Experian. Last year we found that 97% of people could not find their Statutory Credit Report using Equifax. A report that you are legally entitled to for just £2.00.

In the intervening time Equifax has made it even harder to find your report: There were two links on Equifax’s home page to your Statutory Credit Report, now there is just one and it is effectively hidden at the bottom of the page with grey text on a grey background!

Equifax Statutory Credit Report

The rest of this blog was originally posted on 9th September 2008.

As the credit crunch worsens and identity theft increases, consumers are well advised to check their credit rating on a regular basis.

Credit Ratings are now more important than ever before, so at Think Eye Tracking we conducted independent research on two major credit reference agency websites; Equifax and Experian.

In the UK, people have a legal right to get a copy of their Credit Rating for £2 ($4US). Equifax and Experian allow people to apply for their Statutory Credit Ratings online. Both websites also offer a number of premium services related to credit ratings which range in price.

The research included eyetracking thirty people as they tried to get their Statutory Credit Rating reports.

With Equifax, of the thirty people tested, only one person saw and clicked on the statutory report. The premium services however were very engaged with. This suggests Equifax maybe pushing the premium services at the expense of the statutory one.

Some of the people tested verbalized their confusion and unhappiness:

“They are trying to force me to buy my credit rating for £12 when I have a legal right to for £2, it’s disgusting.”

“I just can’t find it. Am I being stupid?…. Is it really here?”

“I must have looked for it for four or five minutes without finding it before I gave up.”

Finding the statutory report proved difficult for 29 of the 30 consumers tested as the heatmap below shows, there is only one mouse click on the Statutory Credit Report link:

The testing returned far more positive results from the Experian site. 18 of the 30 people tested clicked the statutory report, but 27 saw it. Consumers who did not choose the statutory report made an informed decision to go for Experian’s one month free trial of their premium service.

The eye tracking heatmap above details where people tested focused on the Experian home page. Eighteen of the thirty testedclicked to access their statutory report and 27 saw it. Some of the people tested stated:

“I looked at the statutory offer but went with the free one because it seemed better.”

“Why isn’t the Exquifax one easy like that?”

Companies do need to balance their commercial interests with their moral and legal obligations. This is necessary to gain the trust and loyalty of consumers. At Think we believe Experian has got the balance right by prioritizing their premium service while also giving consumers statutory rights a significant amount of prominent screen estate alongside a compelling commercial offering.

Is Equifax profiteering from the recession?

Has Google gotten better?

Share

How many times do you use Google each day? Each week? Each month? Countless times! And would you say you use it in the same way you did a couple of years ago?

It’s likely that you don’t.

There has been a significant shift in search pattern behaviour; one that reflects rational and effective time management; one that is a result of habitualisation. Here at Think Eyetracking, we conducted a study into search behaviour to explore this.

In a 2005 study we found that people looked through many of the search results before clicking on one. The distribution of attention using Google can be seen in the heatmap below.

In our most recent study, we had 30 participants search for the broad term, Oasis. PEEP methodology was employed; participants were eyetracked while completing the task and later asked about their usual behaviour.

As seen in the heatmap above, fixations are studded around the top 5 results and the majority of clicks are upon the top 3 results (discounting the sponsored link). The sponsored link was actually not well attended to due to the fact that searchers are now familiar with advertiser placement within Google. The 2008 heatmap supports the recent trend observed by Cornell University (Their study found that the top 3 Google results get 79% of all clicks) and by AOL (Findings were that 63% of clicks were concentrated upon the top three search results).

Furthermore when asked afterwards what they would normally do when they couldn’t find their desired search result on the first page of Google, 87% respondents replied that they would modify the search terms or refine the search by category. 97% of people tested answered that Google was the search engine they most commonly used and out of those people, 87% stated they wouldn’t bother using anything else.

It’s not clear that Google has gotten any better, but certainly our use of it has become habitualised. Google’s popularity and dedicated following mean that a large majority of users have grown extremely familiar with the search giant and refine searches to display exactly what they need within the top 5 results. We now expect to find our required answer in the top 5 results.

The use of Google has become habit and users have optimized their behaviour accordingly; they now act in order to eliminate the need to scroll below the fold or sift through additional pages of results. Nowadays to compete competently, you must know your customer’s search words, and land in the top 5, if not top 3 results.

Why Think Eyetracking?

Share

Think helps deliver on the promise of marketing;

Create and maintain profitable relationships with consumers.

To date marketing has not delivered on its promise; approximately 80% of all new products launches fail to meet pre-launch expectations.

How can a multibillion dollar industry that employs lots of smart people and ideas get it so wrong?

The answer is simple, if not obvious: market research has largely ignored the importance of consumers’ subconscious minds. Consumers are rarely aware of what they are doing and why they are doing it – cognitive scientists have a rule of thumb “unconscious thought is 95 percent of all thought”, some believe this may be a serious underestimate. I know this may be hard to believe, so think for a moment about your daily commute today, how much of it do you remember?

Probably very little, if you take the same route to work each day. This is because your subconscious mind takes you there on autopilot; there is no need to engage your executive, conscious mind unless something novel or unexpected happens. If you need more convincing go and shop your usual weekly or monthly grocery shop at a new store. Pay attention to how it feels to need to engage your executive mind to find items when you are used to using your habitual mind. It is frustrating!

To date, market research has addressed the conscious or executive mind with questionnaires, focus groups, satisfaction surveys and the like. The problem is the conscious mind is at best only partially responsible for the many decisions people make on a daily basis. One unfortunate but real consequence of this is that 43% of smokers relapsed again within 12 months after undergoing surgery to remove cancerous parts of their lung (according to study by Walker, 2006). This is not the action of a rational conscious executive mind. It is the action of an unconscious habitual mind, stuck on a bad autopilot setting.

There is a disconnect between what people say and what they actually do, we believe this disconnect is symptomatic of the two different states of mind. It’s not that consumer lie on purpose, they usually answer truthfully, but using their executive mind and that is only part of the way we all make decisions. At Think we believe we have found a way to bridge the chasm between the executive and habitual minds to help market research deliver on the promise of marketing.

We rarely consciously control our eyes, this fact gives us the ability to see into some of the processes of the powerful habitual mind with eye tracking. Combining this with our experience, dedicated software, and methodologies, allows us to peer into the consumers’ minds to understand which (if any) mind is being engaged.

Our proven approach combines eyetracking with traditional market research to gain deeper understanding of how consumers engage with visual and written creative work on habitual and executive levels.

It’s the beginning of something special.