We at Vertical Leap are big advocates of automation – seeing it not as something to be feared as a job stealer, but instead a way of liberating workers from boring, repetitive and manual tasks. It won’t take jobs, it’ll make them more interesting and creative.
The reason this automation is needed is because of the mind-blowing data proliferation of recent years. Want proof? We create 2.5 quintillion bytes of data every single day. As such, 90% of all the world’s data was created in the past two years.
This means that 75 per cent of a typical marketer’s productivity is taken up by analysis alone; time that could be spent much more effectively, especially if there are little robots helping.
We asked people about their data challenges
Our services manager Steve Masters gave a talk on this exact subject alongside Smart Insights called How SEO techniques and technology will change in 2016. A survey conducted during the talk found that 37 per cent of people viewing were either ‘not sure where to start’ with their data analysis, or found it overwhelming.
A further 40 per cent struggled to integrate their data, whilst only seven per cent said they were ‘totally all over it’.
Meanwhile, 72 per cent said they were currently using no algorithms to monitor their search marketing efforts.
Questions from the webinar
With so much confusion still pervading, there are countless questions about automation and how it can be put into practice in the real world. Here are just some of the most pertinent, as well as Steve’s response to them:
1. How does human analysis fit in with automated analysis of data?
Steve: “The biggest difference between human and automated analysis is breadth and depth. One person can only do one thing at a time, and to analyse a wealth of data is time consuming. Bots, on the other hand, can analyse many thousands of records very fast – so they’re able to analyse far more. The human difference is creativity and intuition. Until we have true artificial intelligence, a bot can only analyse what we teach it to; a person can apply more abstract thought to the process.”
2. Is there a danger in trusting your own set of collected data, as the criteria are always put together by humans? Will there still be the need for ‘confirmation’, ‘double check’ or something else?
Steve: “Your data can be accurate, but can also degrade over time. If my wife uses my computer to look at shoes, I keep receiving remarketing banners from shoe shops. Similarly, if I download an eBook on fly fishing, the publisher may correctly assume that I am interested in discovering more on the subject. That doesn’t mean I will be interested in it forever, though. Data may be accurate when you gather it, but you can’t assume it’s always going to remain as such. You need to collect more data and cross reference it all the time to weed out the rubbish and any false positives.”
3. Will it be possible for us to ever truly understand our big data when we always find more sources?
Steve: “We can get as close as possible but we constantly need to develop the technology to incorporate new sources which crop up all the time. If you take your focus off this, even for a short while, then your understanding will be inferior to the true picture. In some areas, such as weather, we never really get a true understanding – we’ve been gathering data on weather for many years, but the forecasts still aren’t wholly accurate – although they are a lot better now than they were 30 years ago.”
4. How would you use TensorFlow to curate content for a ‘personalised’ content blog experience?
Steve: “TensorFlow is a powerful AI system that was used to develop Google’s RankBrain, part of Google’s Hummingbird. It uses machine learning algorithms, enhanced occasionally by engineers, to identify meaning in search queries. If Google sees a search that is unfamiliar, RankBrain tries to understand what it could mean – leading to better search results.
“Now that Google has released TensorFlow as open-source software, it has opened the door to a world of AI inventions that we will probably see trickle through over the next couple of years.
“You could create a personalised blog experience in a number of ways. You could automatically re-order articles based on user behaviour, time of day or engagement. What if you create an algorithm to watch reader behaviour, so it automatically tests different layouts or different ad placements, and keeps track of the user behaviour? The machine would learn from its own tests and devise new tests to learn from, so it could constantly improve the user experience.”
5. How do you know the right data sources for your business?
Steve: “You can gather unlimited data on anything. What matters is knowing what information to pull from that data. The best thing is to start with what you want to know, then look for correlations. Take the example of Tesco Clubcard vouchers. Tesco could start by saying it wants to identify what percentage discount is most likely to encourage people to spend more money. Then, it can run some tests to gather data and analyse the results to identify buying patterns.”
Got any questions yourself?
If you have any questions about data analysis, automation or digital marketing in general, tweet us and we’ll be more than happy to reply.