PROBLEM:
Cluttered view due to increase in number of signals
Complicated process/calculation of scorecard, Lead-scoring and few other signal dependent tasks
CONCEPTS:
Before jumping into the solution, let’s understand the concept of “Category”
Category:
A group of signals which reflects the same theme or concept. This is explained in detail while orchestrating the solution.
Stacked Bar Graph:
A ‘Stacked bar Graph’ is a graph that is used to break down and compare parts of a whole. Each bar in the chart represents a whole, and segments in the bar represent different parts or categories of that whole.
SOLUTION:
Note: Our current representation of signals only considers the case of one signal mapped to one category (one to one relationship between signal and category).
In this document, we’re discussing a different case where multiple signals can be mapped to one single category.
We’ve made 3 small modifications in our current product:
Modification 1: Grouped multiple signals which have similar theme or meaning under one category (many signals to one category relation)
Modification 2: Representation of the ‘Signals’ graph has been changed from standard bar graph to stacked bar graph where the stacked bar has multiple signals which belong to one category and that bar reflects that category
Modification 3: In the ‘signal-analysis’ report, we created a super-set to signals which are categories, under each category, we’ve signals which belong to that specific category and under each signal, we’ve utterances that belong to specific signal
The document explains each modification in detail with screenshots of the modifications made.
BENEFITS:
This will help in reducing the number of signals as we display the graph of only categories
we comprehend information easily and gain insights better
This way of organizing & viewing data will also ease our work in terms of creating ScoreCards, Lead Scoring, generating insights and many other problems
For lead scoring model, we will end up using both category as well as signal name for finding weights to handle lead score
MODIFICATIONS:
Group multiple signals with categories
Let’s say we have a conversation where we
The below two tables represent the old and new scenario:
Now with this representation, one would be able to visualize the number of instances of loan as a category was caught and also the number of instances where each individual signal was caught.
Representation of the Signals graph in Dashboard & ‘Signal-analysis’ report
Previous Scenario:
In this scenario, each signal is mapped to a particular category. The view is cluttered due to more signals.
Proposed Solution:
Here, signals having similar concepts are combined together into a single category giving better readability of the ‘signal-analysis’ report.
Signal Coverage Stats in 'signal-analysis' report
In the ‘signal analysis’ report, we’ve added a super-set/parent of signals called ‘category’ where there’s a drop down against every category under which we can view all the signals corresponding to that category. And these signals can be further expanded to view utterances corresponding to those signals.
Previous Scenario:
Dropdown for signals
Proposed Solution:
Dropdown for categories, signals and utterances.
The below image has break down of category ‘Agent Performance’ which has 4 signals ‘agent harassment’, ‘agent negotiation’, ‘product knowledge’, ‘agent assertiveness’
The below image displays the utterances of one particular signal ‘product knowledge’ which was part of the above mentioned category ‘Agent Performance’
