Original article was published on Artificial Intelligence on Medium
Transform Customer experience with the analytics and automation powered by DeepSupportDesk AI
- Customers are happy if they don’t need to go to support for any issues with the product/services. We all know that is not the reality.
- If a customer reaches out to the support, the deepest desire for a customer is to be listened to, understood and appreciated and of course get the resolution of the issue.
- In such a scenario, the deepest desire of a business is to provide the best customer service at an optimal cost while growing the business.
Conversion of hypothesis into metrics
Customers do not like to go to support. But if they need to, they would like to experience followings:
- Listened promptly and carefully — This can be measured by initial response time (IRT) and quality of first response
- Understood accurately — This can be measured by average number of the replies needed to close an issue
- Get the issue resolved with minimum efforts from the customer — This can be measured by one-touch resolution (OTR), number of replies, average resolution time, the average number of reopens, NPS / DeepNPS(customer sentiment), number of times a ticket has exchanged hands during its life cycle
- Business can achieve operational efficiency by reducing the time and length the tickets travel. This means, shifting support to the left (shift left strategy)
- Self-service %
- (L1 / L2 / L3) %
- Agent cross-trained score (how many agents knows how many ticket types)
There are several metrics as described below which can support the hypothesis towards improving customer experience. Some of the important ones are below.
- Average Initial Response Time (IRT)
- Average Resolution Time (ART)
- Quality of Solution (QoS)
- One-touch resolution % (OTR)
- Net promoter score (NPS)
- DeepNPS (customer sentiment from the ticket content)
- The Average number of replies (replies)
- % number of reopens (reopens)
- The Average number of assignees (assignees)
- Volume (created, solved, backlog)
- % of tickets self-served
- % of tickets solved at first line(L1)
- Agent expert metrics
Initial Response Time
What should be measured?
- IRT Trend
- Breakdown by the support agent and ticket type
- Quality of response — re-stating the problem, next steps, estimated time to fix
What should be a target?
- Though it depends on the nature of the business, statistically, 15 minutes of avg IRT is considered as good practice
How can it be achieved?
- Measure the current baseline
- Identify the breaking hours, type, where IRT target/SLA is breaking
- Determine if there are enough agents to handle the incoming volume by each ticket type
- If not then it can be fixed by a) cross-training if the issues is at ticket type level or b) adding more agents if the issues is at the overall volume level
- Iterate the process every week
Average Resolution Time
- Measure resolution time (time elapsed between created and solved)
- Identify the ticket type, a feature that is taking the most time
- Do time-motion study on the ticket type with the top performer and low performer
- Identify the steps that are taking the most time for both -> simplify them by either created simplified KB or by automation
- Cross-train bottom performer on the steps that are taking longer for bottom performer only
- Iterate the process every week
The Average number of replies / OTR
If one does not get the solution right at the first touch, then he/she creates unnecessary work for others.
- Identify which ticket type, feature, agent, environment, days of the week, the customer has the most iterations
- Check if all the information required to solve such issue is captured in the input form
- Check if there is a step by step document to solve such issues
- Check if a new agent can solve such issue by following documents
- Automate any ambiguous step of the process if possible
Net promoter score (NPS)
- NPS Q2 is a very powerful tool for continuous improvement
- Identify the main pain point of a customer from NPS Q2 responses
- Correlate primary metrics such as IRT, ART, Number of replies to NPS and NPS Q2 to narrow down the top 3 issues
- Fix the top three issues every week and iterate the process
Deep NPS (Customer sentiment from the ticket conversation)
- Generally, NPS response rate is around 20~30% which is not adequate data for continuous improvement
- To increase the coverage, technologies such as ML/AI can be utilized to read and understand the sentiment of the customer from each interaction
- It is an extremely powerful tool as you get the sentiment of the customer from the moment when a customer is interacting with customer support organization
- Identify the top 3 pain points by analyzing key phrases, word, ticket type that is bringing unhappiness to the customers
- Fix the pain point by simplifying self-service document, KB, product feature or automation
- Iterate the process every week until every interaction brings happiness to the customer
Quality of response / solution
- It is the quality of the solution is a core ingredient of the highest level of customer experience
- At no time the initial response should not be just dummy robotic acknowledgment
- In the initial response, problem should be rephrased to make sure the agent understood the problem correctly
- It should have next set of steps to be taken to solve the issue
- It should also include the timeline by when issue will be resolved
- In the solution, a summary should be provided that should include, problem statement, data/entity involved, solution provided and instruction in case if the same problem is seen in the future
- Manual QA/ or AI power QA can be used to gauge the accuracy of the solution
Agent expert metrics
- Agents are the most important stakeholders in any support organization. Building experts for each ticket type is a must for the highest level of customer experience
- Having everyone trained on every ticket type reduce the turnaround time provides better redundancy and thus better customer experience
- A real expert is not who solve most tickets but the one who WOW the customer at every single interaction
- Expert metrics should build on the based on the various parameter such as the number of tickets solved, NPS, DeepNPS, IRT, ART, quality of the solution
Shift left (% ticket solved at self-service, L1, L2)
- Lesser the ticket travels lower would be the cost of support and better would be the customer experience. Businesses can utilize the latest technology to move the help to the left, closure customer.
- Technology can identify the ticket type/topic cluster that does not have any dynamic data so it can move to the self service
- AI can detect the entities and intentions from the request to make contextual search from the KB or trigger a specific workflow to shorten the effort and time to resolve the issue
- AI can also detect if the information is to be retrieved from the external systems and get all the information required to solve the ticket to the agent.
- Technology can also predict the ticket type that can be moved from L2 to L1 based on the various parameters such as IRT, ART, number of replies, and number of hands changed, etc
- AI can also be used to detect anomalies, ex: some feature is broken, and support starts receiving more than the normal ticket flow.
- System issue, ex: system not moving tickets from one state to another and accumulate tickets in the backlog
- Besides, to measure and continually improving the key metrics, AI can be used to predict the escalation, customer churn
- AI can also predict the volume at various times of the month and year. It can be used for preparation such as peak times and off times of the businesses
Most companies are striving to focus more closely on their customers. With the progress of the web, social media, customers have near perfect information. Only companies that put the customer at the center of their operation can successfully compete in such a world.
Customer service is where the AI will have perhaps the greatest impact for an organization in the years to come. It will reshape the traditional customer service with natural language processing, machine learning, automation and predictive analysis.
But it’s not just about the technology, it’s about what can you do with it to solve customer problems.
- How to use AI based analytics and predictive analysis to provide better customer experience
- How to make AI work behind the scenes to help employees deliver better customer experiences?
- How to make AI work for cost-effectively increasing agent efficiency and customer satisfaction and deliver business value to the organization?
To conclude, analytics and automation based on the AI/ML technology brings limitless opportunities. Organization those who adopt them early and accurately will benefit the most.