Can machine leaning help us design better ideas, brands, products & services?



Back in 2017, Rare hosted a panel session at a conference in New York which was focused on the subject of targeting Millennials. The topic of discussion, with our panellists from Heineken and Gillette, was: Are brands relevant to today’s consumer?


A lot has been said about the so-called Millennial generation, but two areas stand out:

  1. Attention is short. They want something and they want it now (as the below video shows)

  2. They are a digital native generation where using technology is second nature and this is only increasing as they and Gen Z get older and the next generation beyond grow into adults and become a target demographic.


This desire for speed and technology is not just limited to Millennials or Gen Z. There are many people who, although defined by age as Gen X or a baby boomer, have a similar mindset, are just as likely to use technology and have equally short attention spans.


If you have read Sapiens by Yuval Noah Harari you’ll be familiar with his view that the reason why we are the only ‘Homo’ species to still exist is that our species developed the ability to transmit information about things that do not really exist - The Cognitive Revolution.


At that time, all those million years ago, it ensured our species survival. Today, with technology advancements rewriting the way information is being transmitted, some argue this is still the case.


For marketers, it is certainly true. Utilising technology is one of, if not THE hot topic of the last few years. AR, VR, Blockchain, AI, voice – the list goes on...

The Martech Landscape by Scott Brinker

I’m sure we are all familiar with Scott Brinker’s (above) infographic which shows that Ad & Mar-Tech is a booming sector with [too] many brands and agencies exploring how best to use tech to enhance what and how they can get their products and messages in front of consumers.


Solving the innovation challenge faced by brands and agencies

Back to 2017, one of our key discussion points on the panel was around agility.

Do businesses, especially large multinational ones, have the internal structure and the capabilities to compete with the influx of ‘start ups’ and get new products or services to market quicker? To quickly learn whether consumers will buy or to kill a launch before wasting too much money?


We also discussed whether agencies offer the services to achieve this?

It was at this point we had our ‘aha’ moment. Could we use AI and machine Learning to augment the creative process? Could we turn data into a decision-making tool thus creating unlimited possibilities and ideas which are focused on the consumer? Could we use AI to get to better designs and create better brands, products and services?


Let’s try, we thought!


We were conscious of not wanting to focus solely on technology so began to look at a process that combined human and machine. As Esben Ostergaard said in his presentation at the Thinkers 50 European Summit, on 26th October 2018:


“The human touch is needed. The question is how we can use machines to ensure faster delivery and create unique products”


Moreover, following many conversations with clients and industry peers, we have come to realise the process also offers a solution for a number of further frustrations involved in the creative process and getting from initial brief to concept to launch, these are:

  • Has the right creative briefbeen issued and executed against? (Do all stakeholders agree?)

  • Are we placing Positioning over Preference?

  • Can we reduce complexity?

  • Can we do this yesterday?

  • Do we really need post it notes?

We're augmenting creativity

The process is really simple (as this short video shows) and have five core stages:

  1. Data from selected sources is fed into our server

  2. Our algorithms analyse the data making lateral connections and associations

  3. A range of keyword queries are run based on the brief

  4. The output provides us with a quantifiable source of data and segmented imagery at our finger tips

  5. We develop these into opportunity platforms and concepts


Once the material is ‘read’ it is then in a quarriable format meaning the data will persist and we’ll be able to use it again - saving a lot of time and money for future projects as we already have the training data.


One of the analogies we often use to describe the process can be summed up in the below quote from a review of ‘Behold America’ by Sarah Churchwell. Financial Times, 17th May, 2018:


“Back when I was writing my history thesis, one of my great joys was wandering into the library stacks in search of a specific item – Once there, I invariably found that what I really needed was in a different book on the shelf above”


This process helps identify our book on that above shelf.

And the first results are in...

At the outset, we hoped that this process would give us the ability to push the boundaries’ of where we can take the design outputs and do so with the whole process (from data source to output) taking a matter of days not weeks and it does but since realised, it also offers our clients reassurance and confidence when presenting the outputs to the business.


We’ve run a couple of pilot projects. The first a project to understand how to frame a particular communications challenge for the largest UK FMCG brand and the second, a project to develop a new brand for the innovation incubator team at one of the world’s largest food and drink businesses.


Pilot #1


The first pilot took a total of 10 days. The initial data output, a list of hundreds of words, was narrowed down to a produce a consolidated list of positive and negative word associations based on the queries.


Once we had agreed which words felt most relevant to the challenge, we were able to source hundreds of segmented images via our process which we then translated into initial visual mood boards with the end output a series of opportunity platforms that included starter ideas for a campaign tone of voice.


Pilot #2


The second pilot also took 10 days from start to finish. As with the first pilot the initial data output gave us hundreds of words associations and segmented images. These were filtered into 3 territories which were created based the challenge.


For each territory we created an initial ‘mood & image’ board, brand identity and consumer facing pack design concept. The final output was a more complete creative platform designed for consumer testing, featuring a brand identity, pack design, adcept & a proposition statement.

Key Take Outs

We have been pleased with the outputs but more importantly, so have our clients. Being able to turn around projects quickly is in our DNA however where this process has come into its own (as we hoped) is to help is broaden the range of exploration to create platforms and designs that push the boundaries of conventional human thinking.


If we were to summarise the key benefits in this process, it would be the following:


It makes data useful

The more data the better. But we don’t have to solely reply on data sources from within the business. We can train the data using the national corpus with additional supplements from relevant external data sources. In doing so the associations are extremely strong and not restricted.


It’s Agile

The process allowed us to run a number of different queries and explore the different associations. Once trained, we could adjust our search queries and overlay this to the data we already had.


It’s Fast

The process is very fast. We were able to generate hundreds of word associations from our queries within 24 hours. Once we had refined our search queries and picked which associated words were relevant, we were then able to access an image bank containing hundreds of segmented images which could be downloaded with hours.


It’s GDPR compliant

Because we are using data that has either been provided to us with consent or supplemented by data that is available in the public domain, we are not at risk of becoming non GDPR compliant.


It can push boundaries yet offer reassurance

As stated above, a piece of feedback from our initial pilot was as they weren’t using it to replace traditional methods it allowed them to further enhance the positioning they were taking in the new campaign. Using this process had allowed them to present a strong business case with confidence.