INTRO
In today’s fast-paced business environment, consumers’ low tolerance for a bad product or service experience quickly leads to a decision to choose from a wide variety of competing products. This increases the need for real-time feedback analysis methods and automatic role-based insight distribution.
Integrate Customers into the Decision-Making Process
Becoming a more customer-centric company is at practical level a process and the tools to put that process into use. Some tasks are only done once or when things change (like the overall process and mapping the customer journey), others require continuous monitoring and a large number of often small corrective actions.
Map the Customer Journey
Customer experience is a set of encounters between the customer and the company. This includes everything from evaluating a product, buying it and having a support center interaction. This set of events is called the customer journey and the actual delivery points are called touchpoints.
Most companies map the customer journey touchpoints from top down as an intellectual management project. That is a good starting point but it is also interesting to see what customers think about the touchpoints. After all, customers have the best understanding of their journey and experience.
Incorporate Customers Into The Service Design Process
Just like in customer journey touchpoint mapping, customers’ insights should also be incorporated into touchpoint specific service design. In order to make sure that the service changes are positively received, the touchpoint service experience should be continuously monitored with e.g. transactional NPS (TNPS).
Create A Holistic Customer View
More and more industries are offering their customers personalized communication, products, and services. There is no other way to deliver a personalized experience except by understanding as much as possible about the customer.
Try to Make The Information as Rich as Possible
Tying NPS surveys into a transaction and thus a customer profile will provide the richest possible information, but with feedback forms and social media, you might not be able to deduce the actual customer (profile) or the person commenting is not even your customer (yet!).
Many companies are already capturing customers’ Twitter handles and Facebook names and enriching the customer profile with comments from those systems. Social media is a powerful source of people’s values, opinions and interests.
7 Preparation and Planning Steps
1. Understand Customer Loyalty
The main cause for customer dissatisfaction is the failure to deliver the expected experience.
Your company communicates with customers by delivering messages via advertising, social media, website, direct marketing etc. Based on these messages and what they have heard from their friends, customers form a set of expectations in their minds. These expectations are often emotional. Therefore, it is important that you understand both the tactical marketing campaign messages and the strategic long-term missions (e.g. price leader, quirky brand, best service) when designing your Enterprise Insight Process.
It is important to know who the loyal customers are and why. We have learned time and time again that even loyal customers are unhappy about something. Loyal customers:
1. Stay longer.
2. Buy more and they buy more often.
3. Cost less to serve.
4. Insulate from price competition.
5. Act as brand ambassadors.
6. Provide honest, high-quality feedback.
And, loyal customer are typically much more profitable than other customers.
2. Get Top-Management Commitment
The importance of getting top-management commitment cannot be emphasized enough.
No KPI’s – No Candy
Before you propose the project to the top management make sure that you have decided which Key Performance Indicators (KPIs) to use. One reason for CX project failures is the fact that top management doesn’t understand what you are trying to accomplish, and this is because top management can’t manage what they can’t measure.
It is crucial that you find the metrics (NPS is a good start!) and ensure that, whatever measure you choose, it ends up in the top-management dashboard and is viewed and followed with enthusiasm (don’t forget to market your project!).
Strategic Goals
Most, if not all, large companies have a strategy. However, too often CX programs are put in place to solve only operational problems. They fail to track the strategic initiatives laid out by management. Such a gap in your CX program then makes it impossible to track whether your strategic initiatives have the desired impact with both customers and employees. Make sure that, whoever your feedback analysis company is, they can extract the strategic issues you are tracking from customer comments.
Your job as a CX professional is to incorporate strategic themes to be part of the Enterprise Insight Process.
Customer Proposition
Some companies are cost leaders while others go for luxury or excellent customer service. Whatever the core customer proposition is, you need to be able to extract the relevant aspects from your customer feedback data so that you can track and improve upon it. It really helps if you can quickly see whether your customers view your company and your products in the same way as you have defined in your customer proposition.
Company Values
A cleaner environment, happy employees, active in various charities and local community involvement might be important values for your company.
For example, if your Corporate Social Responsibility plan states that your company should be a frontrunner in environmental issues but your customer’s voice doesn’t reflect that, you are doing something wrong (most likely on the communication side).
It is important that the feedback analysis solution you choose is also able to capture value-related feedback categories.
3. Engage Employees
Employee feedback is usually an underutilized source of enterprise insight. Large companies have thousands, even tens of thousands of employees, and they are often the true customer experience experts.
Many of them are working in the customer interface and quite often will be able to tell you what works and what don’t work.
Employees also should be aware of the strategic themes and be able to monitor their success where rubber hits the road.
Listening to your employees serves also another purpose: it makes them part of the decision making process instead of them feeling like just another peg in a massive machine (which purpose they may not understand and associate with).
By mapping the employee feedback back to the strategic themes, management shows it is using the feedback to measure how the company is doing against the strategic themes.
Create a Process and the Culture Will Follow
Some people say that the right kind of company culture automatically delivers an excellent customer experience. I disagree. Culture is important but without processes there is no systematic scalable way to deliver the experience and introduce new employees to it.
Employee Pulse
Happy and motivated employees, who understand the link between strategy and service design, create better customer experiences. Continuously measure the employee pulse by giving employees the opportunity (an app) to continuously tell how are they feeling and what should be improved.
Human Resources department does and should run the Employee Pulse process.
Employee Suggestions
Employees have good ideas on how to improve products, service design or processes and create a more efficient and pleasant working environment. Why not to listen to them?
Customer Comments Via Employee
Employees who interact with customers hear complaints, praises and development ideas all the time. Employees must have a fast and easy way to share these insights with stakeholders like product managers or service design team.
4. Create a Listening System
Here we come to the core issues: you cannot understand the customer by just analyzing their past purchase behavior.
● First, it is history;
● Second, it misses the question why.
Why do they shop at your store (or some other store)? Why do they buy what they buy? “Why” is emotional. If you can understand the “why”, you have opened a six-lane highway to their hearts.
Traditional Market Research Doesn’t Work
Surveys serve a purpose but the only way to understand customer behavior is to listen to what they say about your and your competitors’ products. Traditionally market research has served this purpose but market research is not delivering the true customer’s voice due to survey fatigue and the changed consumer communication habits.
There is a strong trend in the survey market. One might call it a disruption point. According to Opinionlab 80% of consumers have abandoned long-form surveys.
Structured surveys have three problems:
1. They are too long;
2. You cannot learn anything new from structured surveys; and
3. The sample isn’t necessary representation of the intended target segment. Sampling doesn’t work anymore because only certain kind of people answer surveys and participate in panels (People who have time. They participate because they get rewards).
You have to go out there with a blank paper. The future is in opt-in listening and making as many feedback-giving opportunities easily available to the customers.
Net Promoter Is Good But Don’t Forget Customer Effort Score.
Given the opportunity customers will give spontaneous feedback or answer short context-sensitive surveys (e.g. transactional NPS – TNPS),
Relationship NPS is more suited to track strategic themes, marketing success and corporate values.
It is important to create both relationship and transactional surveys. NPS is a good format for both of these but when it comes to a pure support service experience, whether as a self-service or from a service agent, Customer Effort Score is a better measure.
Feedback Is Mostly Open-Text
Short open-text comments match the way consumers are already communicating. They all send text/WhatsApp messages, most of them post messages on Facebook or Twitter, and a lot of them, especially young people, chat all the time.
Some companies complain that they receive too little feedback and that feedback is skewed (to those who have the loudest noise). This problem is temporary.
The trick here is not to try to get people to learn or behave differently than they do everywhere else. Offer them a natural way, one that they are used to, and make it available in any channel and media.
This will increase the volume and decrease the loudmouth bias. When you combine the feedback from all the channels and maximize the volume, you will have created a powerful listening system.
5. Create an Analysis Framework
Because feedback is mostly open-text you need to figure out how to analyze it. The key question is whether you have in-house competence, and time to develop, optimize and continuously maintain your own analysis and categorization scheme.
In-house: If you have in-house competence and real-time analysis isn’t important in your business, you can choose a tool in which your team develops the analysis model and the categorization scheme.
What we have learned, however, is that most companies prefer not to create and maintain a linguistics team. In that case you need to find a computational linguistics company that does it for you.
Feedback Analysis Requirements
Before you decide on the vendor, please make sure that you have set your requirements for the analysis. As said earlier, most of the feedback is open text, and text analysis is a murkier area than (structured data) statistics.
Here are the requirements for high quality feedback text analysis.
1. Consistent: All feedback (keywords, phrases and brands) is always analyzed and categorized the same way in order to recognize patterns and do correlation and other statistical analysis.
2. Accurate: Analysis accuracy must be on high level so that you can trust the statistics.
3. Relevant: Feedback must be categorized in such a way that the analysis results can be immediately without post processing or additional interpretation used for decision-making. Categorization needs to reflect the way the customer experiences the service and the way the company is organized.
4. Encompassing: Whatever service you choose it must be able to categorize relevant keywords, phrases and brands automatically and in real-time.
5. Continuous tuning: It is important to understand the nature of feedback text: each industry, company and feedback channel has its own “language”, which evolves over time. Therefore, the analysis system needs to be continuously tuned. Make sure that either you or your vendor has the resources to do this
On Feedback Analysis Methods
There are quite a few methods for turning unstructured open-text customer feedback data into structured information. Some ways provide statistically relevant information, others don’t. Here some pointers what to look for in a feedback analysis system.
What doesn’t work
Some methods will not give you the desired result because they are inaccurate, don’t provide statistical information, or the information is on a too high or low level in order to make decisions based on the analysis results
1. Extracting brands (Named Entity Recognition) is useful information when it comes to social media and media discussions but fails to provide actionable information when it comes to your own feedback data analysis: it doesn’t tell what people are talking about.
2. Extracting whole comment sentiment based on the number of good and bad words in the whole comment and how far are they from the brand mention is quite easy but it fails to provide information about the topic of discussion: What people are talking about.
3. Extracting keywords and phrases doesn’t work because keywords cannot be turned into statistical information: there are an infinite number of them and the topic of discussion is statistically distributed “too widely”.
What works
Mapping all relevant keywords and phrases into “contextual baskets” works. We call these baskets topics or categories.
You get the most value out of this method if the categories are industry-specific because of the nature of language: different keywords and phrases have different meaning in different industries. Certain industries need the keywords and phrases to be mapped differently because of the way the industry works and the companies are organized.
To add another layer of information, the sentiment for each topic mention should be detected. This allows you to see in detail, which aspects of your product or business get positive or negative ratings.
The example here shows the feedback topics, topic volumes and topic sentiments for 6 supermarket locations:
This process turns unstructured open-text data into statistical information. This information can be handled like any structured — statistical — data: you can detect patterns and trends, do correlation analysis etc.
This is the type of structure that should come out of text analysis:
• Company Specific Touchpoints
o Capture all Touchpoint specific sentences (not whole comments), keywords and phrases.
o Typically there are less than 10 of these. These are often related to the customer journey: Product research, Price Evaluation, Purchase, Support, Finding Information, Pricing, Marketing, Web etc.
• Company Specific Topics
o Capture all Topic specific sentences (not whole comments), keywords and phrases.
o These are the things or themes customers talk about.
o Typically there is about 50-100 of them.
o Topics can be combined, split, renamed or closed.
o They are mapped to Touchpoints
o There are anywhere from one to over 100 keywords and phrases “mapped” to a Topic.
• Long Tail Topics
o Sometimes things out of the ordinary happen. You need to be able to capture those events because they might be relevant. These should, combined with the Company Specific Topics, form a “whole” worldview.
• Brands
o Capture and map brands and brand nicknames into Topics.
o Typically there is 10-20 of these.
• Sentiment
o Sentiment should be tracked at Topic, and not whole comment level.
o Some companies use multi-scale sentiment system but we believe that negative, neutral and positive is enough.
o Sentiment should be tuned to customer feedback: e.g. a missing item from a retail store selection should be considered a negative.
6. Enrich Data
Enriching data is probably the most difficult part of the Enterprise Insight Process planning and preparation phase. Unfortunately it is probably also, besides the feedback analysis, the most important step.
Enriching customer feedback data is nothing more than linking CRM or ERP data that you already have, to the feedback analysis results. For example in a retail store this includes data about purchase behavior (=line items)
• What customers buy,
• How much they buy,
• How often people buy,
• How far they live, and
• Gender, Age…
This is information that you already know. Now you just have to link it. Transactional or event-based surveys like TNPS are excellent for this: because you are sending the survey to a person you know, you don’t have to ask them any background information. You just link the customer record to the survey answer. Even better, store the survey answer to the customer record so that it is there to inform sales, support and marketing teams.
The results of a Transactional Net Promoter Process should include at minimum the following information: Score + Comment (and the Feedback Analysis Results)+ Age+Gender+ Average Purchase Basket+How Many Times/Month They Shop+When Was Their Most Recent Purchase etc.
7. Integrate Systems
An Enterprise Insight Process requires some level of integration between the various systems.
Please keep in mind that a listening and survey system might include many subsystems (survey tools, customer forum platforms, feedback forms, social media, Net Promoter etc.).
Insight distribution might also include many services like email, excel reports, reporting tools, visualization tools and case management software.
Feedback analysis should always be a single service but if the work is done manually, then the work should be distributed based on metadata. Advanced Feedback Analysis Services can automate most of the insight detection.
Case Management should be one system but depending on the size of the company, there might be more than one.
Each of the systems can run independently but in order to get to the real-time issue or trend detection, these systems should be integrated. Data flows only in one direction. That makes integration quite simple. Interfaces should be built by using web interfaces (REST and SOAP).
5 Continuous Process Steps
There are five steps in the continuous cycle.
1. Listen
Operationalize the listening system that you created during the planning phase.
The important thing is to continuously develop new listening channels and mediums. Especially when new technology makes it easier. NFC tags enable businesses, which have physical outlets, to gather feedback.
I don’t understand why every restaurant doesn’t want to hear what their patrons are saying. This would be simple: just put a NFC tag or text message number at the corner of each table.
2. Analyze
Operationalize the Feedback Analysis system. This, like the analysis system, should also evolve all the time. This is especially important because the world changes all the time
• Create new brand names, if new products are launched or product names changed;
• Amend strategic themes, If the strategy is renewed, the new themes should be included in the analysis;
• Tune touchpoints, if the management structure or the way you offer services is changed; and
• Continuously tune Topics and improve the keyword and phrase mapping to Topics.
• Periodically test the sentiment analysis accuracy and push your vendor to make the syntactic and semantic rule corrections required to improve sentiment analysis cross the board.
Please make sure that your Feedback Analysis vendor takes care of this, unless you are ready to dedicate a team for this task!
3. Distribute Insights
Improving operations and pinpointing mishaps in marketing or strategic initiatives is only possible if the insights are distributed to stakeholders who can take action.
Different organizational levels and roles require information in different formats. For instance, in “the field” information needs to be as detailed as possible, whereas on management level it needs to be measurable (KPI’s!).
Don’t forget to use the background variables (enriched data) in the insight distribution rules.
Task Lists
A Customer Feedback Analysis service should be able to automatically generate a task. This is possible because open text is turned into statistics. And, in statistical data it is possible to recognize patterns based on thresholds. This task should automatically create a case in a Case Management System.
Tasks can be divided into two categories:
Location- or product-specific issues: fixing one or very few problems locally (store or location). These tasks should be as concrete as possible. They could even have the actual phrases that customers or employees used to refer to a specific issue; and
Systemic issues: Feedback Analysis System should automatically generate alarms based on recognized patterns (trends and thresholds). And these alarms are automatically turned into cases.
A case administrator should decide who needs to get involved and what steps are required to resolve the issue.
In its simplest form a case management system can be a list of issues, people involved and proposed resolution as a row in a excel file.
The best way to do this is to use a Case Management System.
Research Tools
Quite often a more detailed analysis of the feedback is needed, for instance for root cause analysis, or to test certain hypotheses that management would like to investigate further.
It can be useful to find weak signals and emerging trends that may not yet show up in the regular reporting. This requires that your feedback analysis solution can provide advanced filtering and reporting features to be able to drill deeper into the data.
Analysts should be able to use tools they are most familiar with. These tools include visualization platforms like Tableau and Qlik but also statistical analysis tools like Excel, SPSS and SAS. Therefore your Feedback Analysis Service should be able to export to any database format.
Management Reports
Insights need to be delivered to the management in a form and manner digestible to them. Middle management wants more detailed information, even actual customer comments. Top management preference is for dynamic dashboards with limited number of metrics.
It is not enough to give management just metrics but also how the issues detected from metrics are resolved or what is the plan to get them resolved. This often requires manual work or at least retrieving information from the Case Management System.
4. Act
Everything from step 1 to 8 should lead to this most important point. This whole system is useless unless you:
● Fix loyalty affecting issues,
● Come up with service design and product ideas,
● Benchmark your marketing messages and campaigns,
● Ensure that your strategic messages are understood and appreciated, and
● Fix operational problems.
We tried to outline in this graph, the different steps and roles in running the whole system.
Please, note that the case management system can be just an excel sheet or an email chain. This graph is just about the operational issues. Management dashboarding (metrics and visualization) should also be taken care of.
Detect And Fix Loyalty Eaters
Most of the loyalty eaters are quite obvious: price, bad customer experience, change in a product feature, bad customer service experience, can’t find information. But it is not straightforward to detect these issues from open text (one issue can have many words and phrases to describe it).
You need to also know how many people are complaining and what are the details of complaints. If the feedback is not anonymous you can also know who is complaining.
Extract The Root Cause
Statistical information is valuable when pinpointing trends or emerging issues but the root-cause for problem is “hiding” within the text.
Create A Service Recovery Program
Companies are familiar with the most common service failures. They should implement a program in which there is a clear process and rules in each failure scenario.
Catching an unhappy customer who is just about abandon your company at the right moment can actually be more powerful than if they never had the bad experience. This is called service recovery paradox.
Know What Competitors Are Doing
One often-overlooked source for competitor information is customers and employees. Social media and product discussion forums can be a powerful source of competitor insight.
An Enterprise Insight Process should include competitor benchmarking – what trends – marketing campaigns – features and learn from competitor’s successes and mistakes.
The insight analysis system should incorporate a Named Entity Extraction and the crawled social media and product forum comments should be cleansed.
5. Communicate
Closing the loop by telling customers what you have done based on their comments is extremely important. You can also extract positive comments for use in marketing and to motivate employees.
Customers will notice that you have listened to them and made corrections and improvements. But most of the time the changes are so subtle and slow that the only way to let your customers know that you care about them is to tell them that. Social media is an excellent communication channel for this.
Please remember, you don’t choose the channel that they listen to, they do. Therefore you need to be in many virtual and real places to get the message through.
10 DESIGN PRINCIPLES
1. Action oriented mindset – ensure that information flows to the right people, who can take action to resolve the issue!
2. The listening system must be aligned with the customer journey!
3. Incorporate strategic themes to the Feedback Analysis and Insight Distribution!
4. Don’t forget the employees!
5. Educate staff to proactively resolve CX mishaps!
6. React to emerging trends – don’t wait for problems to grow!
7. Turn text into statistical information that you can directly use in decision-making!
8. Let analysts use the tool they are most familiar with!
9. Integrate the systems for minimum manual work and errors!
10. Have fun and the process will turn into a culture!
