Understanding Image Psychology

The Problem:

Imagery is the most powerful form of content, and yet we aren’t using data to quantify why or why not something resonates with our audience.

  • 50% of CMOs admit they don’t know what it is about their creatives that generates awareness, engagement, and sales.
  • By 2020, ad spending on visual-led platforms will be 75% higher than on search.


The Solution:

Identify and tag all aspects of your images, and then compare engagement with them to identify which aspects are more important.

Interesting Stories:
  • A town in Japan had a crime problem, so they added blue lights in strategic locations to see if this would impact crime. The results of this experiment found that, with no other variables, crime decreased by 9% in the first year.
  • The New York Times did a study to see if a font could affect a reader’s trust. They found that “Baskerville” was the most trustworthy font and “Comic Sans” was the least trustworthy font.
    • Download Baskerville here.
  • See which Netflix icon they found had the highest engagement:
  • An interesting finding by a Picasso Lab’s customer: images without a subject’s head performed better than if their head was included.



How-To Overview:

Course Content

Total learning: 13 lessons Time: 10 weeks
  • Section 1  0/13


1.  Overview

Hi my name is Dillon, today I will be speaking to  you about how the hyper personalization of the marketing funnel is approaching the final frontier  which is imagery in how we are using the artificial intelligence machine learning to kind of tackle that.  The problem with imagery is for too long it has remained entirely subjective in the void of data and using artificial intelligence the machine learning were are trying to make it more scientific process. From watching this you will understand exactly how images affect people’s emotions and how you are able to change the imagery you make for your company to increase conversions and get a better engagement from your audience and the people you want to target.


  1. Imagery Is Not Quantified & Why You Should Start

In the creative development process as a whole is very subjective and largely devoid of data. Although there are a multitude of very talented creative people out there when it comes to making images, videos, and whatever else large amounts of the creative process are from their background, from their history from what they like and from what they don’t like and not from data. There are kind of what we call four part of the marketing funnel. The who, the where, the when, and the what. The who, the where, and the when are fully automated or almost fully automated and measurable. And when I speak about who I am talking about your target audience. Who were you going after? There might be a makeup company that is going after teenagers from the ages of I don’t know 13 to 18 with their shiny lip gloss. Where large beer brand might be going after young men ages 18 to 24. 21 to 24. That’s the who and that’s a pretty automated and measurable part of the marketing funnel that we understand. And there’s the where and when I say where I mean two things. One, there is the platform that you’re using. Some brands obviously have found tremendous success with Instagram while others have found success on Pinterest or LinkedIn or whatever. One of the where is the platform. Every company knows where their ads and their assets perform better than others when it comes to platforms. And it’s pretty easy to tell. The other where I was referring to is location. A lot of larger brands especially have multiple offices. They’re international brands and they have different audiences.


The reality is is one size does not fit all. And people in different areas perceive creatives differently. And so all of that data however is pretty automated. You can see where your customers are coming from. Where people are posting about your company. The IP addresses all of that is pretty automated, measurable and easy to find out. The when is another thing another kind of part of the funnel that is pretty automated and measurable. I personally as the head of growth here at Picasso have used it in email marketing and newsletter campaigns. We know that it’s probably not best to send out a B2B marketing email at 1:00 in the morning on a Saturday or Friday night. But otherwise you kind of have a bunch of data points, hundreds of millions of data points as to when is best to send an email and when is best put out a tweet or an Instagram post. That is all pretty automated, measurable and easy to do as much of other softwares. What we think as is the final frontier is the what. Like I said, the who, the when, the where are automatable but when it comes to the creative development process a lot of senior leaders give complete control to the creative people and although they’re very talented generally like I said doesn’t have much to do with data but rather their personal opinions. A study done by the Content Marketing Institute actually found that 50% of chief marketing officers admitted to knowing that they didn’t really know what Creative’s, a generated awareness, engagement and sales.


And so we think that the image is the last frontier and because it’s so subjective and devoid of data it’s not automated and that’s what we’re trying to change. To give you some proof as to the fact that visuals are really important and that it’s kind of the final frontier in the automation kind of the hyper personalization of the marketing funnel. In 2015 ad spend on visually led platforms was about equal to ad spend on search. By 2020 ad spend on visual led platforms will be 75% higher than that on search. Right there that is enough kind of reason for people to believe that visually led marketing is the future and I know it’s a buzzword that everybody says it but it’s true. The ad spent is going through the roof in that vicinity and it will continue to do that. However in addition to ad spend on these visually led platforms there are websites, emails like I mentioned, your organic posts, display ads even packaging and a brick and mortar advertising like in stadiums or on billboards or on bus benches. All of these things have to do with imagery and for that reason in addition to ad spend on visually led platforms it is extremely important that we kind of take the final step in the hyper personalization of the marketing funnel and include data in how we create our images.


  1. Story 1: Blue Lights Decreased Crime By 9%

What does this mean for marketers, what does this mean for you guys? So I’m going to start high and kind of give you a lot of examples, a couple of examples at least of instances when Color Psychology has been used to affect how people behave. In Japan there was a city and specifically one area where there was a high number of people taking their lives higher than usual in other parts of the city. And so what the city decided to do was implement blue lighting, blue signs, just blue everything. Blue was a calming color and they thought let’s see if we can use color to affect how people are feeling in the day to day basis. Let’s see if there’s any way that adding these blue lights and blue signs and blue accents all throughout this part of town help with this really unfortunate problem. And other places than this has some more concrete data to go with Glasgow.


I mean Glasgow there is an area of town that had a very high crime rate and one of the things the city did same idea is they decided to put blue signs up, blue lights up, blue street lights and whatever else. In an effort to calm people down and really attempt to reduce crime without any massive measures like tons of police officers who are really just a lot of bright lights and cameras and all that stuff. And over the course of I think it was the year that part of Glasgow actually saw a 9% decrease in crime rates and that was without any other police added or anything like that. And that goes to show really that decisions are affected by color and something as simple as adding blue maybe convincing people not to commit a crime or made them in a better mood or stopped a road rage or whatever the case may be. And this is how kind of what we interpret through our eyes impacts our decision making.


  1. Story 2: Pill Colors In Pharmacy Trials

Another interesting example of where Color Psychology has been used is in the pharmaceutical industry. One thing that a lot of pharmaceutical companies have been criticized for is using Color Psychology color theory in their trials. In this slide you can see blue and a red pill. Typically like I said, blue is is kind of a relaxing color, more calm in a pharmaceutical term will be more of a downer whereas a red pill a stimulating kind of impulsive. And again in the pharmaceutical term would be an upper. And a lot of these pharmaceutical companies were accused effectively or are not accused but there’s been some controversy around whether or not they use color to influence trials. So for example if they had for whatever reason if they desired for people to experience some sort of upper reaction with the red pill or with any pill they would make that pill red because it is associated with kind of that upper feeling that intense high octane impulsive feeling that is associated with an upper. Where as the blue pill if it was meant to be a downer or not would associate… They would take the blue pill and it would be used to kind of calm them down and really play the part of a downer. Both of those examples have to do with just simply Color Psychology affecting the human body on a literal physiological level. There’s a funny example in the pharmaceutical industry that has been brought up a few times and it has to do with Viagra. A lot of people have basically complained that they took on the little blue pill thinking that it was a downer of some sort and we’re kind of surprised with an upper that they were not expecting. Forgive the the dead joke.


  1. Big Brands Are Using Color On You Right Now!:

There’s a lot of different places where we can already see our psychology being used. And one of the most obvious is logo. Companies that want your data, that want your money are typically going to use again this blue color that is trusting, calming and makes you feel at ease. This is not a one size fits all. There’s obviously exceptions to every rule. But you see companies like Facebook and JP Morgan these companies use blue because they want you to feel safe and at ease with them. Companies on the other end of the spectrum that want you to be a little more impulsive will use red. A funny example I like to use is Seamless. The food ordering app. I’m sitting on my couch on a Sunday and I’m deciding between cooking a breast of chicken and brussel sprouts or I can flip through my phone and see a Seamless app and all of a sudden before I know it there’s a delivery man on my doorstep with Indian food. The red Seamless app inspires impulse gets you to open it and says I want a more unhealthy dish through it. I’m going to go away from the chicken and Brussel sprouts and I’m going to be impulsive and have the meal that I really want. So those three examples are very kind of broad examples of how Color Psychology and how we as humans react to color and how it affects our decision making.


  1. Does Font Affect Trust? Yes, Use “Baskerville”:

With that said it’s not just about color. It goes way beyond that. One example that we can use for this is what’s called the Baskervilles study which was done by the New York Times. And what they did was put out an article to over 40000 unique individuals and they had in a bunch of different fonts. And the article the images and everything about the entire piece was identical with the exception of the font. And what they used this for was to see does fonts change the way that people trust the New York Times, trust the author, interpret the article as factual or non factual? Can font really affect what people are doing when they’re reading an article? Or is it just the words that mean anything and the fonts kind of useless. As it turns out the font is not useless. Baskerville hence the name of the study was found to be the most trustworthy, honest and valid if you will font whereas not surprisingly Comic Sans was the least trustworthy font that The New York Times used to write this article. This really just goes to show that it’s not necessarily about the color. It’s not about the shade or anything like that. It can be as simple as the style of text, the style of font that makes people react in a kind of have this reaction a very visceral reaction to a piece of text despite the fact that all of the letters, all of the words, all the periods everything is identical.


  1. A/B Testing & Subjectivity:

Another place that we’ve seen this play out is A/B testing. A/B testing its scale is really really challenging because a true A/B test is two identical assets with one variable different. As you can see in this slide. President Obama had two landing pages to get email addresses and ZIP codes. One of these and I don’t even know which one. One of these performed nine times better than the other. That’s it. That’s a really… 9X it’s a really high number. But as you can see there are dozens of different variables in these two landing pages. One you’ve got Michelle and his daughters. One black and white. There’s signs and one is crowned one the saying is different. Get involved or says Change We Can Believe In. So there’s always different variables. And although I’m sure his team found this to be a very interesting insight. The reality is you can’t learn a whole lot in the long term from this A/B test. I’m sure that they took the one that performed 9X better and they put it out to the people and promoted it on as many social media platforms as possible and got a ton of emails and zip codes. But the reality is when it comes down to it and if they came back and said what within these two landing pages affected such a performance boost? They wouldn’t be able say because there’s so many variables. So although A/B testing has its values and it’s a very important thing to do.

It’s really hard with imagery because it is so time consuming to create images, videos, creatives that have only one variable differed from the previous one and then to put all those out and promote them it’s just so complicated and frankly impossible to do for large scale organizations. And so that’s kind of the issue with A/B testing and why we’re trying to move beyond that. A funny example as you can see on screen here which look is more likely to lead to a better outcome. This is actually our CEO Anastasia and she is an extremely data driven person. And one of the things that she thought… I wonder how my appearance affects outcomes when I take meetings whether that be for fundraising from venture capital firms or business meetings or who knows even dating life does glasses change how people react to me? Was her question. And so usually on this slide I do a show of hands and half the crowd comes up with one thing and then maybe a quarter comes up with another and a quarter of the people just don’t feel like raising their hands typically. But the point I make here is we don’t have answers to this question as not a statistically significant amount of data, amount of meetings that she went to to say whether or not conclusively that wearing glasses leads to a better outcome. Wearing no glasses leads to better outcome. But the point that I illustrate with this slide here is that every time I get somebody to raise their hand and say yes she does better with glasses or not she doesn’t do better with glasses.

That in and of itself is an opinion and just goes to show and demonstrate that we as human beings have reactions to images that infect and kind of impact our decision making. Every single person in the room is going to have a different answer and a different story as to why they believe glasses work, glasses don’t work and that’s what’s so unique and that’s what’s so hard about imagery because every person’s reaction is different and visceral and it’s really hard to capture that in a data way in a statistically significant way.


  1. How Netflix Optimizes Icons:

Netflix has been an industry leader in figuring out which visual cues increase the metrics they care about. Probably clicks on shows, how long people watched the videos. All of these different things. And so what they’ve done here you can see six thumbnails that’s been put on their homepage or their landing page. And for the Unbreakable Kimmy Schmidt on the bottom right hand corner the one with the green arrow, that was the most popular one and I don’t have access to their data as to why that was most popular or maybe it was because only wanted to had both the lead and the second lead in the show. Whatever the case may be what they’ve done is they’ve put out all these thumbnails with different visual themes and seen how they impact performance whether that be like I said clicks through to the show, how long people watched the show. How many episodes people watch depending on the time of day. Is it better to put something that’s black and white at night versus in the afternoon. Of all of these different data points they’ve put out hundreds of thumbnails to figure out exactly which ones work best with which audiences and from there they’re able to kind of optimize because they have a ton of data on all their users. Be able to optimize the thumbnails by time of day, by gender, by age, by race. All these different things to get people to click and watch the shows that they’re putting billions of dollars into making. And so Netflix has done a really good job at that and they’re an industry leading in trying to figure out at scale how can we change people’s reactions to imagery and how can we leverage that.


  1. Using Tech To Analyze Imagery:

So getting onto a kind of more scientific part of this. And when I was talking about the hyper personalization of the funnel, I started at the top with very broad strokes when it comes to Color Psychology. Now we’re going to get into the nitty gritty a little bit more about how technology is used to take this to the next level. On my screen you can see here… This is how people see an image. Here we’ve got a burger with fries and ketchup, Pepsi in a wooden table with two hands. Some fingers popping out of the corner on the right end corner and every person is going to have a different reaction to this image. I personally had a small watch so I’m hungry when I see this image. I might prefer sweet potato fries or muster. All of these different things are very subjective and have to do with the mood we are in as humans, how we were raised. All of these different things. Our reaction to images is not scientific, it’s very visceral. It’s very feeling based. The point I’m trying to make here is that as humans when we take this very subjective look at an image that’s not helpful to kind of data driven marketers like myself. However when you see an image through technology this is actually how technology sees an image. And on average when we put an image through our system at Picasso labs there are hundreds of data points that we can pull. Here you can see that there are people present. That’s true. We see the hands. Is there a female present? That’s going to be false.


We don’t know if there’s a woman here or not because it just looks like hands. What’s the beverage type? Is anybody making eye contact? Are there multiple people? Is there a logo present? All these different things that the technology sees time and time again almost instantaneously that we as humans don’t even pay attention to. Furthermore not only do we have these creative attributes that technology can see within an image so quickly but we have display ads. Excuse me, but we have performance data that can be overlaid on top of this created data to correlate the two. So we have total ad spin, ad recall rate, conversion rates, cost per acquisition, cost per click all these different things in these very specific metrics. And when you put all this data into one image and then you get hundreds of images into a pool you’re able to see over time that maybe sweet potato fries does reduce the cost per click. Maybe having Pepsi increases the ad recall rate over having just a glass of soda. All of these different things which are almost impossible to see with just the naked human eye are visible almost instantaneously through technology. And that’s really what we’re doing at Picasso labs.


  1. **HOW TO: Using Excel**:

What I want to do is kind of take a step back and say how can you take advantage of this and how can you get your toes wet and see especially if you’re a smaller business with not a ton of resources. How can you do this at the start to make sure that you’re at least taking a different eye when it comes to social marketing. The first leg is the Excel way. It’s pretty simple. You put assets into an Excel spreadsheet. You put binary things like are the people in the image? Yes or no. Is there text in the image? Yes or no. And you put performance data into this Excel spreadsheet. Click through rate, ad recall rate, cost per acquisition whatever the case may be. Once you have all this data in an Excel spreadsheet and even then pivot tables and figure out which specific attributes affect performance best which don’t and kind of run a slightly in-depth look into which of your images are working and why they’re working. If you have a small business you only put a couple of images up per month or on average per week or a couple images per week and they’re not that complicated this is a great way for you as a marketer or you as a business owner to get started and really take your visual marketing to the next level.


  1. HOW TO: Outsourcing:  

Unfortunately, a lot of companies have a lot of images going out and don’t have the visual eye to see all the different attributes in an image. So if you think this basic Excel way isn’t quite as complex as you’d like it to be. There’s another way. Way number two is outsourcing visual recognition. This is kind of a hybrid between a full service company and doing it yourself. When you outsource a visual recognition you can do it to companies like Google or clarifier Microsoft. And what’s unique about this is their software is that a very inexpensive price will generate a ton of tags is what we call tags in the image which are things like I was talking about earlier the creative attributes, the logo present, is their face present, are the people present all of these different things and these companies can do that for you in a pretty short amount of time at a pretty cheap rate and you get that information back and you’ve got images with 50 different tags in it and then you can re-plug it into the excel spreadsheet that I was showing you earlier. And from there you’re going to have to add your performance data. And at that point you can now start running pivot tables and regression models and see exactly which attributes affect performance based on audience and all of these different things.


  1. HOW TO: Using Software:

If you have a lot of image and a lot of tags Excel spreadsheets get really complex really quickly. And so finally there’s more of an end to end solution and this is what we do at Picasso labs. And so we’ve broken it down here into kind of three different buckets if you will. One of them is ad targeting one of them as social media optimization and one is guideline compliance. When it comes to ad targeting this has to do primarily with audience data because you have audience data when you’re targeting on Facebook, Instagram whatever else you’re able to overlay that data with performance data as well as the creative attributes just kind of three layers there. And we’re able to see that as this example states close up pictures of food over perform for 35 to 44 year old males in the US. That’s a made up stat that’s not necessarily true. Like I said it’s not one size fits all. Depends on every brand and they have different audiences with different preferences. But that’s the kind of information you need with the kind of targeting available in advertisements. Social Media Optimization. This is taking a look at your organic post. Take a look at your organic posts put them through a system and say what in this image performs well, what increases engagement, what gets more likes, what gets more shares, what gets more comments. All of these different things.


A good example of this is one of our clients at Picasso labs, one of our clients found that street style despite being the type of fashion that is the least expensive to produce for them for their Instagram posts was the… Had the highest engagement rates and so they said if it has the highest engagement rates and the lowest production costs, we should really ramp up our production of street style because it’s cheap for us and people like it the best. They didn’t think that they thought the runway was the absolute way to go. They were the most impressive images. They were the most stunning and they had these preconceived notions about what worked and what didn’t and technology went on to disprove that. Another funny thing they found out is that people with their heads chopped off in the image actually perform way better than when their heads were present. We don’t know why this client’s audience likes that or why they don’t. But it’s statistically significant facts that are only available with hundreds of thousands of data points that it just isn’t computable by the human brain at least not quickly. And finally guideline compliance. This is really relevant to larger client of ours. That larger organizations when you have multiple brands across all the continents putting out different creatives, you’ve probably got guidelines that you want met no matter the geographical location or the team that’s putting them out or even the brand. For example, Nike. If they wanted to create a guideline compliance or best practice if you will. I would imagine they’d want the swoosh in almost all of their assets. So what this guideline compliance does and what technology can do better than any human or at least faster than any human that’s for sure. As you put an image in it it finds the swoosh and says check. That’s one best practice.

If it doesn’t have a swoosh, it will spit it back out and say nope that best practice was not met. And so with this technology you can put tens of thousands of images into this technology and in a very short period of time it’ll tell you exactly which guidelines were met which weren’t. And it’s actually able to point people in the direction of which teams are best performing which ones are using guidelines. And also maybe even most importantly which guidelines work and which don’t. A lot of the times people make these guidelines based on their subjective feelings. I don’t know CMO might have a daughter with red hair and he or she says I like red hair because my daughter has it. So let’s make the model have red hair. Well that unfortunately doesn’t fly.


  1. Conclusion:  

And I want to wrap up with kind of one final thought. Going back to the hyper personalisation of the marketing funnel. Like I said the when, the who and the where those are all pretty automated. But we as humans crave kind of these feelings and interactions that influence decision making and I’ll give an example. We can debate… And I debate this with many people many times whether or not brick and mortar retail is dying. Some people say it is. Some people say it’s never going to die. The argument that of the people who say it’s never going to die is usually that you have these one on one interactions with the employees at these establishments. I personally I have many examples of this in my life. I get my coffee same coffee from the same guy every single morning because we have that one on one interaction. That inspires emotion and convinces us to buy there if it’s more expensive than Starbucks about you know whatever the case may be. I still do it. This is really important and we believe that if we’re going to kind of continue this hyper personalization of the marketing funnel and really take it to the next level it compete with this one on one interaction that we find in brick and mortar stores with other human beings is extremely important that we take data of the imagery. Make sure that the imagery we’re producing is speaking to our audience in a way that inspires emotion and can convert or can help convert those people into lifelong customers and really brand advocates going forward so that your brand is not just throwing random beautiful assets at them but is rather throwing creative assets at them that really speak to them and get them to engage and achieve the result you’re looking for.


Leave a Reply