There can never be too much data

image

They say you can never have too much of a good thing. In my view data is a good thing – but can you have too much of it? Clearly many of my fellow marketers do – recent research by IBM found that more than 70% of CMOs are put off from doing anything with their data due to data overload. However I think data needn’t be challenging – by following a few simple tips it can benefit every marketer.

Should content still be king?

There is so much content being created at the moment that marketers’ messages are often drowned out. Just in the last minute 72 hours of youtube video, 571 new websites and 100m new emails were created. How do you cut through all of this content? You need context. If you are not getting the what, where and how right then all your effort spent on content is wasted. So content is still king, but context is its partner.

Think about when

The trick is to start small and to think about the different phases a customer is going through. For example, if you are a mobile phone company – one of the most important dates is obviously the renewal – but the renewal date varies from customer to customer and their situation will be different. Understanding the nuggets of gold in your data, which in turn lets you determine nuance along with understanding which channel works best in that specific situation will deliver cut through.

Don’t forget slow data

Initially, when trying to react to your customer behaviours, you start with “if this – then that” type messaging. For example, if a person browses a product or downloads an article then you send them a certain communication. I call this “fast data” – but there is a much bigger opportunity with “slow data.” This deeper analysis, allows you to find new patterns in the data, which adds colour to your messaging.

How much talking is too much?

There is a fear of sending too many messages and alienating the reader. But this doesn’t need to be the case – it all depends on the context. You can talk a lot to your customer if it is relevant.

Be careful how much you say

Many of us will already be familiar with the perhaps mythical story from Target in the USA. Target used patterns in their data to understand the 25 products an expectant mother would buy. By assigning the products with a score they could calculate how far into a pregnancy a mother was. They were so accurate that they were apparently able to deduce a particular teenager’s pregnancy before her own father could. Whether this actually happened is not the point – the example demonstrates the need to respect your customer and their need not to feel watched.

It all comes down to responsibility – there is a huge amount of data available, but by understanding what you are trying to achieve and by communicating upfront on how you are intending to use the data you can successfully use it to further your content marketing.

Riaz Kanani

COO @ Profusion 

21st July 2014

Not just a pretty face? Amazon and the art of secret signals

image

This week, surely in anticipation of my upcoming trip to Normandy, Amazon Mastercard cleverly sent me an email to ask where I was jetting off to. Normally, I’d be inclined to send this sort of message straight to spam, but I couldn’t resist opening this one given its sheer timeliness.

image

I was pleasantly surprised to find a nice reminder and easy step-by-step guide on how to inform the bank of my travel plans via SMS to avoid my card getting blocked. This information, combined with a crisp visualisation, and I was sold.  A picture’s worth a thousand words after all. And because I was already checking emails on my phone, it made sense to send a quick text with my plans then and there.

image

But this email was so much more than a pretty face. Of course I was struck by the copy and creative, but its Amazon’s ability to integrate its products and services with its customer data that really caught my attention. Not only was I being subtly encouraged to use my credit card abroad, if I decide to use it, Amazon will gain access to data on my travel destination, dates and holiday activities. Just think of the countless recommendations I could receive on my Amazon account in preparation for my trip. Sun cream and hats for the beach? Hiking boots and a light jacket for a wander in the French countryside? A few light beach reads and a Lonely Planet for my Kindle? What about a Kindle Fire for the flight?

But why stop there? With information on my holiday activities and shopping behaviour, Amazon will quickly discover my penchant for funky jewellery and all things beauty. Soon, I’ll start receiving emails with suggested products to re-stock my bathroom cabinet or a list of Amazon’s best shop-within-a-shops to browse for a cool new bracelet. Of course, they won’t forget to remind me that I’ll earn double points on my credit card when I shop on their site and, consequently, my Mastercard will remain my first port of call long after the holiday glow has disappeared. Top that all off with the fact that I’ll be able to buy all my necessities in one place and have them delivered to my door once I’m stressed and time-strapped again, and how can I say no?

Admittedly, this level of integration is easier for the well-established digital players, but any business working in the digital space should consider taking a leaf out of Amazon’s book. Mapping out your digital strategy, be it specific to email, social, search or mobile, within the framework of your customers’ complete journeys (both online and offline) is the way forward.  Consumers want personalisation and convenience and this holistic approach means businesses can tap into specific needs and apply the right secret signals at the right time to keep their customers satisfied at any stage of the journey. As a consequence, they’ll keep their brands top of mind and increase sales across entire portfolios.  

We all know that for today’s consumers, convenience is king. If your business can ease our lives by solving problems before we’ve even identified them, you can almost guarantee we’ll keep coming back time and time again.

Yariella Coello

Consultant

14th July 2014

Jul 8

How we use trees* to predict customer behaviour: random forests in action (*not real ones)

image

Discovering business insights requires understanding how customers will behave, which means decoding whatever information we have on a customer to predict their needs and ultimately whether they will buy (see article: Are there such things as digital buying signals?). How does one draw out from all the available information a clear signal and filter out irrelevant information? Intuitively, we understand that a salesperson is able to classify prospective customers based on cues such as the clothes the prospects are wearing, their apparent age and other such factors. However, how do we formalize these heuristic methods in such a way that a computer could do this? And how do we let the data reveal which features or combination of features have the greatest predictive value for say buying a new car or booking a holiday to the Bahamas? The random forest algorithm solves both of these problems.

The phrase random forest might evoke the image of collections of trees popping into existence at random locations, however it is actually the name of a very powerful algorithm used in data science. For understanding how this algorithm works, it is helpful to think of the two player board game Guess Who? popular in the 1980’s. In this game, there is a collection of known individuals represented by cartoon images and their names. Each player selects one individual from the population and then by taking turns asking yes or no questions one attempts to guess the identity of the opponent’s selected individual. Whoever manages to correctly guess the identity first wins the game.

image

Source: Tynker

The game is very simple to understand and children quickly pick up through experience some of the more subtle nuances. For example, although all questions have a binary outcome of yes or no, some questions will be more valuable than others. You could start out the game by asking “Is this person Susan?” and if by some stroke of luck your opponent’s selected individual is Susan then you would have immediately won the game. However, more often than not that will be a terrible strategy since a negative answer would only allow you to eliminate one person from the pot of all possible identities. A wiser initial question might be “Is this person blond?” No matter what the answer is you will be able to either eliminate all people who are blond or all people who are not blond. For the sake of argument, let us assume that 20% of individuals are blond and 80% have some other hair colour. If one is lucky, the opponent’s individual is blond and then one only needs to sift through the remaining 20% of people left. If one is unlucky, then this wasn’t a very good question to start with since there is still 80% of the data left to go through. It quickly becomes clear that the optimal question is one whose answer would split the data evenly into 50% yes and 50% no.

The second subtlety is that the order in which the questions are asked matters and the answers revealed along the way will have an impact on what the next optimal question will be. Imagine that in your previous round you have asked whether person has brown eyes and the answer is yes. If there is a 50-50 split between brown eyed people who wear glasses and brown eyed people who do not wear glasses then asking whether this person wears glasses as a next question would be a very good choice. If on the other hand the answer to the previous round was no and among non-brown-eyed people only 10% of them wore glasses then asking whether the person wears glasses is a sub-optimal next question.

This method of repeatedly splitting data and finding the optimal splits at each stage to do so is known as a decision tree. Visually speaking, the metaphor is that all the data starts out at the trunk of the tree and then every split in the data represents two branches of data going off into different directions. The meaning of optimal split is a bit different when our goal is classification. The optimal split in this context is the split that divides the data as purely as possible with regards to whatever it is we are trying to predict. Thus, if we are looking for an algorithm that predicts whether or not someone is going to buy a certain product, we find the variables that are best able to correctly split our training data into buyers and non-buyers. Using a decision tree is simple but is unfortunately prone to over-fitting. This means it will do an excellent job at classifying the data you train it on but will do poorly when trying to make predictions on new data. The random forest algorithm remedies this by creating many decision trees each one of them trained on a subset of all the available data and each one only able to use a subset of all possible features in the data to make its splits. When new data is presented, each tree in the forest gets to vote on which class it believes a given data point belongs to and whichever class gets the most votes is the prediction made by the random forest algorithm as a whole. The beauty of the method is that it is conceptually easy and it tends to perform very well.

Thus we are better able to understand each individual consumer and know what they want – perhaps even before they do.

Henrik Nordmark

Data Scientist

8th July 2014

Don’t fall at the last hurdle - the essential rules of copy

image

We are well rehearsed in the important elements of digital marketing today. As we discussed last week data science makes it entirely impossible to send relevant, timely and personal emails. However all this hard work is often lost when customers are still not acting upon the messages that are sent. Why is this?  Marketers are often falling at the last hurdle. So what is this hurdle and how can we jump it?

Recently I sat down to a concert at my son’s school. Before it started the Headteacher gave a welcome talk. In it he remarked that if he took a photograph 100 years ago of a waiting audience it would show men in overcoats and pipes, with women in headscarves. He then went on to (half) joke that a photograph now would show just the top of people’s heads as they look down at their smartphones and tablets.

So what are we doing when we look down? Playing games, sending texts and of course, reading the latest email that has come in. Whereas in previous times all correspondence used to be read at the breakfast table with no external distractions we are now reading messages in all sorts of places – usually under time constraints – and that means we are just scanning messages while being easily interrupted.

So after relevancy, timing and personalisation the last hurdle is, of course, copy. How do you make your message engaging to these easily distracted customers? These 8 points make up the essential guide.

1. Make a compelling subject line

A quick google search will reveal so many articles on this subject but the importance can’t be emphasised enough. This is your first and most compelling headline – get it wrong and your customer won’t go any further

2 .The pre header should be used as an opportunity

There is an argument that the pre header is the new subject line. Customers are becoming immune to the subject line and, if available, will scan the preheader to decide if they should continue. This shouldn’t be underestimated and persuasive detail should be included.

3. The opening paragraph should keep the reader’s attention

This should clearly and concisely explain the purpose of the email and what benefit the customer can gain from reading it. Don’t fall into the trap of setting the scene or selling your company – if the customer doesn’t see a benefit they will move on to their next communication.

 4. Be concise

Remember, on the initial read your reader is scanning, not reading. A copy heavy email is just too challenging for a customer to sift through.

5. Divide email into sections

Readers will lose interest if we put demands on their attention span. By breaking up the copy into sections you allow the reader to pause and absorb information between points. The use of visuals can really help here.

6. Use bullet points            

Obvious, but often overlooked! Bullet points make it easy for the reader to pick out what they want to read about.

7. Don’t be too technical

It’s all too easy to forget that what is everyday language within your business may not be for customer. Technical terms can sometimes creep in. If in doubt, keep it simple.

8. The CTAS should be clear

Clearly the CTA must stand out in the email – but the copy on the CTA is equally as important. The language should be clear and action orientated.

I know, reading through these it all seems obvious. But it’s amazing how these principles are not applied. Even the most relevant, personalised, timely communication is useless if it just doesn’t get read.

Sam Killip

Consultant

30th June 2014

 

 

Pioneering principles still apply in the world of personalisation

 image

What can we learn from the ‘madmen’ of the 50’s and 60’s in today’s world of personalised digital marketing?

I recently attended a talk by Andrew Cracknell in which he brought to life a series of examples from the Madison Avenue golden age of advertising. Andrew spent his copywriting career on both sides of the Atlantic before getting out of the advertising game.  It was a fascinating nostalgic talk but far removed from today’s world of highly personalised marketing. But it got me thinking – the advertisers of yesteryear were pioneers of the bold and the brave in advertising and perhaps we can still all take inspiration from that.

Data science makes it entirely possible (although it would be a stretch to call it easy) to send relevant messages. We can look at declared data – such as information filled in from on-line or off-line forms. We can combine that with observed data such as customer clicks and web browsing. We can layer in transactional data, plus draw in environmental information such as the weather or the latest fashions. We can also send timely messages – if John Smith regularly opens emails at 8am that’s probably because 8am is when he prefers to do so. The huge amount of data that we are provided with also enables personal messages to reference customer names, products they own and what they like – so personal messages can be achieved through any digital channel. But how do we make our messages engaging?

This is where Andrew Cracknell’s book ‘The Real Mad Men: The Remarkable True Story of Madison Avenue’s Golden Age’ comes http://prfsn.com/jWQ  The example that really stuck in my brain from his talk was how the advertising men and women of Madison Avenue began to break down the barriers of how messages were transmitted to customers thanks to pioneering work from David Ogilvy and particularly Bill Bernbach.

Often cited as the greatest ad campaign of all time, DDB’s ‘think small’ campaign for Volkswagen was the first to push the boundaries of what print advertising could be.

Previously, hand drawn, idyllic scenes of car ownership bliss were accompanied with copy written at customers - often promising impossible fantasies of the American dream.

image

DDB ripped up the rule book and launched with this

 image

The ad’s genius lies in its subversion of all the previously held rules of the era. By using self-deprecating humour, photographed product imagery, and leaving white space this ad succeeded in fulfilling perhaps the toughest brief in history, selling Hitler’s car to post-war America.

The success of this campaign helped pave the way for a golden age of innovation, but is there anything to learn in 2014? Firstly, we need to succeed in the challenge of making our messages relevant, timely and personal that - but that’s only the start – making our messages engaging is the real challenge in 2014.

Technological advancements now mean that we’re in an enviable position to identify and personalise messages to our customers.  However we mustn’t forget what is as the heart of our industry. If we can learn from some of the advertising pioneers and create bold, brilliant, impactful imagery and copy within our messages than we really can create amazing customer experiences.

Simon Farthing

Head of Consultancy @ Profusion 

Are there such things as digital buying signals?

image

There was a time when the only person you could sell to was the person standing in front of you. Sellers were trained to ask the right questions and read the buying signals. Car sales people, for example, make a judgment about how ‘good’ a buyer you are based on the car you arrive in, the clothes you’re wearing and countless other smaller ‘signals’.

Making those judgments is a lot trickier if your potential customers are sitting at home – whether they’re online, watching TV, or reading a magazine.

I grew up in print media and we traded on selling our audience. We tried to create an image of who our audience were based on demographic information (ABC1 Housewives aged 25 – 34 were particularly in demand!) and survey responses from data sources such as TGI (http://prfsn.com/TGI). This type of audience profiling was so central to our pitch that my old boss even had a picture of himself on his desk with the caption “Like my audience, I am broken down by age and sex”.  An entire industry grew up around modelling postcodes and socio economic groups – and I still hear demographic profiling discussed now, in the digital era. But how relevant is this now, in the world of modern data science?

I think of that type of profiling as being an (often poor) substitute for really understanding who your customer is – and what they want. In the digital realm there is little excuse today for not reading the signals your customers send you – and acting on that insight. The demographic data we’ve always had might provide a thin pencil sketch – to get to a richer personal portrait you will want to hear:

  • What your customers tell you about themselves (declared information)
  • What behaviour your customers show about themselves (behavioural data)
  • What your customers reveal through past purchase behaviour (transactional data)
  • Where your customer is – and at what time of day, or day of week (contextual data)

Declared data

This will produce a snapshot of who your customer was at the time they answered your questions. Some of these will be true long term, some of them will change quite quickly. But it’s a good starting point.

Behavioural data

Often this is where the most valuable information will sit… after all, in a sense it’s the building block that has made Google the behemoth it is today. Understanding that someone is actively interested in buying something from you is the best indicator you could wish for – and we can read that interest through the web pages your customer browses on your website, the way they interact with your email communications, the items they search for.

Transactional data

In many circumstances, your customers are creatures of habit: a lot of brand loyalties can tend to be difficult to shift, so past buying patterns can predict future decisions… think about trying to persuade a mobile phone subscriber to upgrade their phone. A high-end habitual iPhone user is much less likely to be interested in the newly released budget android phone.

Context

We can quickly tell where someone is, and what sort of device they’re using, when they open an email or browse a webpage. This information can be gold dust. One example that springs to mind is the company who want to know when a customer opens their email on a desktop: at that moment, that customer is likely to be at their desk and available to take a phone call. Our client’s chance of success here is enormously greater than otherwise, just because they stand a chance of getting through.

It boils down to the art of all good storytelling: timing. When your customer approaches you, they’re telling you that the timing is right. But this can cut both ways. Retargeting banners from holiday companies are just an irritation when you have already booked your holiday. An insurance quote is spam if it comes the day after you’ve renewed your policy.

We have the knowledge of how long customers take to decide to buy things – car purchases can take days or even weeks to decide on, with car insurance you probably have a few hours at best to persuade. So let’s be more smart and sophisticated with our timings.

There is still a place for traditional modelling to help provide an initial sketch – but as marketers we want a single – or, better, a true customer view – and these clues are helping us get there. 

 Mike Weston

Managing Director  

Jun 9

Does the key to a successful campaign lie in our subconscious mind?

image


We all know that in today’s market attracting customer attention is more difficult than it has ever been. Our customers have become more sophisticated rendering traditional marketing methods ineffective. Having a superior product or offering the cheapest solution does not guarantee a sale.

So what is the solution? It just might be a scientific breakthrough called Neuromarketing.  So what is it?

Studies have shown that traditional marketing methods only appeal to 5% of the brain – ignoring the reptilian, or ‘old’ part of the brain which is responsible for most of the decisions we make. According to Harvard marketing Professor Gerald Zaltzman “95% of our thoughts, emotions and learning occur without our conscious awareness”. Zaltzman is not the only expert to believe in this theory, the 95% rule is widely accepted when estimating subconscious brain activity.

Our brain is made out of three distinct areas which act as separate organs with different structure and functions.

  • The new brain thinks – rational and analytical
  • The middle brain feels – processes emotions
  • The old brain decides – its primary goal is survival so processes information by asking questions such as “Is it familiar?”, “Do I have to act now?” “How do I benefit from this?” and “Is it safe?”

The challenge presented to marketers is how can they address the “old” brain which is 450 million years old? Neuromarketing involves the study of brain activity, eye tracking and skin response to advertising and brand related messages.

A leading scientist in the field of neuromarketing has discovered that the “old brain” responds only to six stimuli:

1. Self centred

The old brain is self centred and only worries about itself. Traditional marketing techniques have focused on how great a company is – the survival instinct of the old brain has no use for this type of information all it cares about is what a company can do for them.
So put your customer at the heart of every conversation.

2.Contrast

The old brain loves contrast; it won’t make a decision unless it sees before/after, risky/safe, with/without or fast/slow. Without contrast, the old brain becomes confused resulting in either a delayed decision or worse still, no decision at all.
Define a clear contrast between yourself and a competitor and increase the likelihood of a sale.

3. Tangible input

The old brain relies on tangible input; it is not qualified to process written language. It can’t process concepts such as “flexible solutions”, “agile environment” or “scaleable solution” without hesitation or effort.
Use simple ideas the old brain better understands.
 
4. The beginning and the end

The old brain pays most attention to the beginning and the end, it does not retain the energy to grasp what is in the middle. In a presentation scenario, if given the choice always present first, particularly as people find it easier to assess content to be worse than better.
Be the first in a presentation type scenario to attract the most positive attention.

5. Visual stimuli
 
The old brain is visual; our optic nerve is forty times faster than our auditory nerve, therefore we see before we hear. Neuroscience has proved that if you see something that looks like a snake you will associate it with danger before the brain has even identified it as a snake. 
Visual stimuli will lead to a connection with the decision maker.

6. Emotion
 
Finally the old brain responds to emotion which in turn helps us to remember. A few of our primal emotions include fear, joy and anticipation.We tend to have a clearer memory of things we were emotionally attached to.
Use emotion to help your customer remember your message.

Conclusion

The truth is customers don’t know what they want.  As Steve Jobs once stated, “It’s really hard to design products by focus groups. A lot of times, people don’t know what they want until you show it to them”. By using neuromarketing we can address the six stimuli of the old brain to understand the customer and ensure they make the right decision.

Nerkesa Pignatelli - 09.06.14

Jun 2

"A Stock Take - Where Have We Been And Where Are We Going?"

Mike Weston

Welcome to our new blog where we share all our deepest thoughts and feelings on – and some of the amazing things we can make possible because of – data science. But before we get to the future we need to understand the past. How did we get where we are?

Once upon a time we were an email production house. We turned our clients’ conversations with their customers into beautifully crafted emails – and we were very good at it too. But that was not enough. We realised we needed to tell our clients how to make it work better.

So we started thinking: what recommendations could we make that could improve communications to our clients’ customers? Of course we considered best practice – but soon realised that led to mediocrity and boredom.

It came to us that if we really wanted to understand how customers think we needed a bigger picture. So we started to look at more and more sources of data – some obvious and some less so –to work out what our clients’ customers did, how they think and what they wanted.

We needed help with all of this so we employed some amazing data scientists and invested in the latest technology to help us unravel all the secrets data can reveal. Now, instead of waiting months to find out how much impact our campaigns have had, we can pretty much predict how customers will behave.

But this is just scratching the surface. There is so much to talk about – surprising things we have found out and the difference they can make, common myths that are just not true – patterns people believe that are just not there. We are discovering things all the time. So I hope you will share with me our regular blog and explore our exciting world.

Mike Weston

Managing Director

2nd June 2014