Machine learning in marketing

Marketing was always dealing with big data - sales results, campaign results, clients’ surveys, et cetera. With the rise of machine learning new methods became available which can help marketers to grow their results and predict with greater accuracy campaign results.
ai, big data, machine learning, marketing,

How to run a successful marketing campaign online? There is no simple answer. Campaigns tend to be complex, run on different channels, the way they are measured can involve multiple metrics and source data may not be compatible. Moreover it is hard to adjust them live, without knowing a potential outcome of a given change.

 

Problems and opportunities

 

There couple of basic problems in marketing which can be approached with machine learning. In this subsection we list some of them and then in the next we will cover some of the already implemented solutions.

 

Automated data visualisation. Data visualization is a basic tool for every marketer. You want to see your data organized, so that you can reason and act upon it. However it is not always clear how you should structure your data, which metrics to choose, how to measure a success of a given campaign. Building predictive models becomes easier by implementing ready-to-use solutions. As an example take NeuroX which can be built-into Excel to help you with sales predictions.

 

Content analysis – text, images, videos. With growing machine-understanding of data online, one is able to direct precisely potential customers (market segment). Language analysis helps in understanding emotions in each post (anger, humour, etc) and act accordingly. You can identify the emotional language, find the structure to it and in general understand your audience. Of course this can be done by experts on individual level, but with machine learning you can take that to macro level and apply it to every one of your customers at once.

 

Real-time analytics is what allows making changes on the go. You don’t have to stop the whole campaign, build another model from scratch and re-run it. With real-time analytics you can tweak parameters while running the campaign or – even better – allow the machine to do that for you.

 

Media buying was automated already some time ago. No online campaign runs without a demand-site platform to process ads to multiple platforms and the right market segment.  

 

Using Internet of Things (IoT). IoT is growing, there are more products on the market which are connected to the web and thus are ready to process data about users. This provides a perfect source for marketing strategies. It is still in development.

 

There are couple of problems with marketing online. One of them is ad fraud, especially bots directing traffic to fraudulent websites. It was estimated that advertisers lost around 7.2 billion $ in 2016 on ad frauds caused by bots. In order to avoid it, marketers would need to use accurate statistics and with the big data, that’s impossible to do by humans. That’s why machine learning techniques are necessary. There is also a problem of monitoring active campaigns online 24h/7 (especially if they run in different time zones). Only by real-time analysis one is able to address both these problems.

 

Already existing solutions

 

There are already plethora of companies providing help in optimizing your marketing strategies by using artificial intelligence. Let us give a couple of examples.

 

Albert.ai is an example of an artificial intelligence platform designed to automate marketing campaigns. It is multi-channel, optimizes while running, and performs content analysis.

 

MediaMath is a company for media buying and an example of a demand-side platform. Other DSP is provided by Codewise’s Voluum.

 

For a multi-purpose product check DoubleClick by Google, which provides services from DSP to campaign tracking tools.

 

For more examples, see a list in Forbes or a list in Business Insider for smaller companies.

 

Beyond obvious

 

Automating marketing processes doesn’t end with optimizing existing campaigns along different channels. Machines are definitely better than humans at handling numbers, but can they be creative? Can ad creation be also automated? This will be the next step in bringing machine learning to the ad world. Bosaz from the Coca Cola company has spoken at the conference about automated content creation – writing the full ad scripts, creating images and music. It is not unimaginable as already music creation can be automated.

 

In order to create the whole campaign, one needs to understand the market segment one is addressing and its needs. That includes choosing good emotional tones (context analysis). This is often a subtle task – a hard one, but not out of reach for machines in coming years. There is definitely a large room for developments in the near future.

ai, big data, machine learning, marketing,