Predicting future, in many different fields, might be a hard task. If you play with stocks, for example, you might have realized that. Even though economists give you recommendations about best companies to invest, a single video that turns viral from a customer complaining about an experience might affect predictions, lowering prices of these stocks.
In the Customer Feedback Management field, the same thing happens. Even though we might have some control over the internal environment, external events might affect positively or negatively in ways that are hard to predict. Go Daddy, for example, might not be able to predict that one bad service would result in this video from our blog, in which hundreds of people will now have a different perspective about this company.
This is one of the reasons explained by our CEO Riccardo Osti in this video, to answer the question: Why is so difficult to predict the future?
If you like the content, feel free to subscribe to his channel here.
Do you prefer to read it instead? Have a look at the script below:
Hello everyone, I am Riccardo Osti and on a daily basis I help the worlds best brands become more profitable by investing in the consumer experience
In this video, I will tell you why today’s predictive technologies are not as precise as we may expect.
When talking to my clients I often get asked: “does your software have predictive capabilities?”. Then I go online and I see all kinds of tools selling “prediction” as you would sell ham or cheese. Imagine the data scientist asking you “I have cut two hundred extra grams of predictions….should I leave it or I take it away?”
Mmm….no, I would say that this is for sure an interesting HOT topic, but there is still a lot to do from an education standpoint…that’s why I decided to put this short video together.
Please let me know if you like it in the comments, and I can make some more recordings on this topic!
There are two basic kinds of predictions that people make: intuitive predictions, which rely on experience and intuition, and statistical predictions, which instead rely on data and algorithms.
Today we are talking about the second ones, which are obviously more interesting from a business perspective. The goal is to gain enough evidence, both from a qualitative and quantitative perspective, in order to make decisions and drive investments in the near future.
What I am saying is that predictive technologies usually read and interpret past events trying to understand how things will develop in the future.
There are three main factors that we need to consider to evaluate the reliability of any prediction:
Number one, the quantity of historical data. Number two, quality of historical data and Number three, the influence of external events
Quantity is key. Most of the predictive technologies are based on statistics and are nothing more than advanced regression analysis. This implies that if you have more data you have a more solid backbone to start with. In some categories, it is easier to collect a large amount of data: you can think of weather forecasts for example. In some other categories, such as text analysis, the availability of the data is much more limited
Quality is also key. The reliability and the accuracy of the dataset that we use to start the prediction is just too important. Let me give you an example. Let’s say that you are a product owner at Samsung and your goal is to improve your products to make sure customers like them better and sell more. You may use text analysis to understand what customers dislike and identify the problems of these products. You could then try to predict how much the satisfaction would increase if you could solve one of those problems. In this case, in order to make a realistic prediction, you should be able to start with a dataset that has been correctly analyzed.
External events are the deal breaker. The biggest barrier to generate an accurate prediction comes from the outside. In fact, all the events that are not directly influenced or controlled by us are much more difficult to predict. We have made the example of the product owner, who tries to predict how much he could increase customer satisfaction by improving one element of its product. What would happen if a competitor launches a product with better performances at a lower price point? Well, most probably the prediction would be completely invalidated. And the problem is that predicting these type of external events is extremely difficult, but common at the same time.
So… what’s the secret recipe for good prediction?
Use a lot of historical data, make sure the quality of the analyzed data is high, and most importantly….keep in mind that external events are hard to predict but extremely dangerous
I hope you liked this video. If so, please subscribe to my channel! If you want to see me talking more in-depth about this topic just let me know in the comments.