7 ways to benefit from machine learning

Tommi Vilkamo

“The future is already here, it’s just not very evenly distributed.” This quote is from science fiction author William Gibson from sometime in the early 90’s. Now, at the verge of widespread artificial intelligence, this is particularly true in the business world.

Applying machine learning is the most important trend of 2017, an explosively growing target for investments and a way in which the financial growth of a country can be doubled.

But have you already applied artificial intelligence and machine learning and predictive analysis in your work?

If you haven’t don’t worry, you are in good company.

In the majority of companies and their different functions machine learning and AI have not yet gotten very far. Therefore it is of critical importance that you now start taking your first steps regardless of whether you want to specialize yourself to become an expert organization in AI or just collect quick wins and preserve your competitive edge. 

Below are seven everyday examples from which it’s good to start. This way you can achieve early results.

  1. Predict and prevent customer churn

    Telecom operators have already long predicted the risk of customer churn on the individual level (of individual clients) and targeted their marketing efforts based on this. Research shows that a small decrease of only 5% in customer churn will typically improve the financial results of a company between 25 - 95 precent. Therefore managing customer churn is, for many of us, the fastest means to reaching tangible results in our business. 

    The necessary data is often already available. For example for a bank we were able to predict customer churn with up to 85% reliability with just a few days of work without new investments in technology. 

     
  2. Optimize your marketing campaigns

    According to a customer survey done by Forbes magazine 86 precent of companies have succeeded in raising their return of investment (ROI) for marketing efforts by using predictive analysis. 

    In the best cases the results can be massive. The American retail giant Tesco is said to have increased their results by 675 precent and saved 600 million dollars with the help of a data oriented marketing program and by relinquishing their special offer campaigns that had been proven useless.

    We are presently launching a similar project with a long established company in the Finnish food industry. I eagerly await the results. 

  3. Predicting supply and demand

    Our own customer projects show, as do publications (industry releases) and competitions in this field, through machine learning you can improve your forecasting significantly compared to before. And most importantly these better forecasts have provided supply chain business benefits in the range of up to 20 - 50 percent

    Market leader Relex’s CEO Mikko Kärkkäinen reminds us of the risks and speed blindness (loose sight of the goal) related to this topic. However when you look behind the surface I do feel that we are of the same opinion about in which direction things are developing.

  4. Machine assisted pricing

    Airlines are the pioneers of machine assisted pricing. Likewise the hotel and hospitality industry have achieved substantial results. Next generation host Airbnb has even gone so far as to share their pricing algorithms as open source. At the same time the company commented on their success by saying that every time the host set their price no more than 5 precent above what the machine suggested they increased the likelihood of getting a booking by nearly 4 times.

    In some situations machine assisted pricing is simple. As an example you can take the prototype we built for realtor Kiinteistömaailman for optimizing apartment pricing and realtor fees. Through this prototype we learned that, at this time, the professionals in the realtor world still beat the best machines nevertheless the estimates set by the machine are still credible and likely to benefit seller, buyer and realtor.

  5. Predictive maintenance

    Preventive maintenance is financially profitable but too often done according to a pre-determined maintenance program. This simultaneously leads to both unnecessary breakdowns and overly cautious service interventions. 

    By using real-time sensor data and other historical data you can use prediction/forcasting modals to predict/anticipate breakdowns and significantly reduce service breaks, maintenance costs, and accidents

    Despite the excitement and media attention around IoT data many, surprisingly, have not yet applied machine learning models to interpreting and utilizing that data and rely only on predetermined maintenance rules. 

  6. Understanding textual material

    Email spam filters and social media sentiment analysis are maybe the most typical applications/usage areas for text analysis. Even lawyers were able to apply it effectively in seperating the wheat from the chaf when analyzing the emails related to the Enron scandal back in the day. 

    The areas where you can apply text analysis may be surprising. One government agency who were visionary enough (from the beginning) asked us to classify 12 million a year textual transaction material. A few weeks later the machine had learned to do the classifications based on only a few hundred classification examples provided by subject matter experts.

  7. Suggestion engines

    Try going to your company’s webshop and buy, for example, a wood burning stove. Does your webshop suggest that you buy sauna stones and a smokestack so that you can build a sauna? 

    If not, now is a good time to take action. Amazon.com, the pioneer in this area, makes up to 35 percent of their sales from product suggestions. Even a small investment in suggestion engines usually leads to a revenue increase of 5 - 15 precent. Implementing a suggestion engine used to be a quite difficult and expensive endeavour but now the technology has matured enough to be within everyones reach.

    AI and machine learning will become mainstream and will change almost every company and profession over the next ten years. The more routine-like the nature of your work is the larger the opportunities and size of the change will be. 

    For example, healthcare analytics is already a 7 billion euro industry. And it’s growing at an incredible pace/with incredible speed. More and more actors are using the revolutionary potential of AI and machine learning in clinical patient work. 

    While admiring IBM’s Watson’s ability to diagnose cancer and other breakthroughs it is nonetheless easy to forget that clinical patient work is only one potential area for applying analytics and machine intelligence. With the financial, operative and administrative sectors the potential is even bigger in the near future.

    Some of us are starting to use and utilize new tools in a very pragmatic way to get better results in our work and some will fight the change. Because you decided to read this article all the way through to the end you have probably already chosen sides. The question that remains is, what will be your first step?

Archive