Voice of customer surveys are increasingly used across the world. There are smart software vendors that have created easy to use platforms to deploy such surveys as well as capture the data from the feedback. The article talks about the various analytical techniques from data mining to Natural Language Processing based text mining, that can be used to analyse the data generated from VOC surveys to provide actionable insights to the business.
As the importance of social media increases, Voice of Customer (VOC) is increasing in importance. Word of mouth publicity, it can be a zero cost force multiplier for one’s marketing efforts, as well as a drag on one’s corporate reputation if the direction is negative.
Recognising this trend, companies are increasingly using insights from VOC to drive business priorities. VOC data is being collected both from targeted surveys as well as customer care center interactions. This data is often a mix of structured and free form information. While traditional analytic techniques can be used to mine the structured part of the VOC data, a variety of NLP based tools are required to lend structure to the free form text data that is often equally important and rich.
Key measures of CSAT from VOC data
VOC surveys usually track 2 key metrics – Top 2 Box Score and Net Promoter Score (NPS)
VOC scores are typically on a 11 point Likert scale (0-10), with responders being divided into the following categories
- Promoters – 8 , 9 & 10
- Passives – 6 & 7
- Detractors – 0-5
Top 2 Box Score is calculated as the percentage of the total respondents that have marked a rating of 9 or 10 while NPS is calculated as the percentage of promoters – percentage of detractors
Global benchmarks for NPS
- > 70% -> World class performance
- 50 – 70% -> Acceptable levels
- < 50% -> Improvement opportunity
The NPS / Top 2 Box are computed for the overall company, specific business units, functional areas (finance, support etc) and trends analysed for actionable insights.
Using the VOC data to drive analytical decision making
Directed data mining of the VOC data generates insights into the company’s strengths and opportunity areas
Use VOC to drive strategy across various business units and functional areas
- Differentiate using attributes where CSAT performance is at world class levels
- Improve in areas where CSAT levels are below satisfactory
- Maintain service levels in areas where performance is satisfactory , some areas with high impact on overall NPS can be selected for improvement
Identify trends in CSAT performance by analysing CSAT time series data
- Areas that are picking up over time
- Areas with historically high levels of CSAT performance that are showing signs of a down turn
Extensive deep dive analysis to understand customer feedback in greater detail
- Geographical cuts – are there trends specific to geographies that are different from overall trends
- Cuts by customer role within client organisation – Do executive sponsors rate us differently than practitioners? Does IT department rate us differently than marketing? Etc
- Simple questions on competitive performance included in the survey help the client benchmark their performance against key rivals
The data can also be fed into statistical models that can identify key drivers of growing CSAT or Top 2 box scores
- Using the NPS or Top 2 Box as dependant variable and the scores on various attributes as independent variables , Logistic Regression / Decision Tree / Neural Network models determine the key drivers that show a statistically significant impact on CSAT
- These are the attributes that should be focused on with top priority, as they give the greatest ‘bang for the buck’ when improvements are made
- Simulations are run to determine the impact of improving scores on key attributes on CSAT scores, as in the illustration below
Once key drivers and trends are established, mining the free form text provides insights into the exact issues facing the company. In the example illustrated below, we have mined through free text to isolate exact issues that bother / delight the client once performance in Global Support Services has been identified as a key driver of CSAT. This provides management a list of the actions that need to be taken to spruce up their CSAT scores
Analysing the statements above, we see that the two critical pain points in support are (a) too many layers and (b) Inadequate competence of support personnel. This becomes a focus area for the Functional Area head.
Try deep learning using MATLAB