Relevance of customer feedback
The core behind the performance of liCili is an algorithm, which we have developed and improved in more than 3 years. The relevant and visible thing for you is the clustering of the evaluated customer survey including the single view, which has already been described on the previous pages. Behind this there are several calculation schemes and researches, which ensure that you immediately recognize which topics are relevant to the customer and the market. The individual layers are explained and shown in the following diagram.
Clustering consists of customer and market relevance, which enables you to quickly gain an overview and assess the feedback received.
The weighting is the decisive factor when it comes to giving a realistic and objective result and decision. By different weighting curves all considered factors are weighted among each other.
The search volume for all words and word contexts is researched to find out how high the interest in the individual topics is.
Competition is also an exciting influencing variable. The aim is to find out how many companies are active in this field and how strong they are.
An important basis for the algorithm is the recognition of existing word correlations and their respective ways. This allows important topics and statements to be understood as a whole and not taken out of context.
The research searches the visible web on the basis of the words and word contexts found. More than 25 factors are researched, four of which are presented in the following layers.
By researching trend indicators, it is possible to determine whether this topic is currently experiencing a hype or whether there are sessional fluctuations, for example.
The degree of novelty indicates how new a feedback is and how important a reaction to the analysed evaluation is.
Degree of novelty
Types of words
The first and very basic layer is the recognition of the respective word type. This is necessary to identify the relevant and significant words from the feedback.
At the beginning of 2015, the idea for the algorithm was born and development began. Several theses were developed, evaluated, tested and confirmed or disproved. Especially through a scientific investigation in cooperation with the Change Management and Innovation Institute of the Hochschule Esslingen in 2017, the last pieces of the puzzle were brought together. Since then, approx. 100,000 data have been evaluated using the algorithm and checked for correctness. This was carried out in cooperation with the respective companies and experts. They all agree that the much faster algorithm was in no case worse than the more complex human evaluation. A total of 50% see the evaluation even better as carried out by human hand.