Many companies collect vast amounts of customer data but simply don’t know how to derive actionable, meaningful insights from it. Is this your company? The key word being meaningful, as insights can be plenty but unless you can put them into the right context it will be meaningless for your business.
Getting actionable insights from your customer feedback analysis
Whether you’re using a chatbot, NPS a survey, or another means of collecting customer feedback, you will need a reliable solution to be able to analyze the data you collect to be able to decipher insights in the right context. Verbatims, or free-text responses, can be notoriously hard to analyze, which is why Natural Language Processing (NLP) methods are used to ensure correct analyzation. At Thematic, we use NLP and (Machine Learning) ML in addition to our own AI which we’ve developed to analyze verbatim feedback responses.
We’ve talked a lot about AI technology here, but don’t worry, we still need humans! In fact, it doesn’t actually matter how good an AI technique may be, only a real person can effectively decide what’s actionable and what’s insightful for the business (applying the right context with the correct historical knowledge).
What are actionable insights?
When it comes to making sense of data, getting actionable insights is the holy grail. But when is a finding an insight? When is an insight actionable? Can data analysis deliver them? Let’s get to the bottom of this by looking at some examples.
The difference between insightful and non-insightful data
Non-insightful data is everything that’s old news to you. Something that you already knew was an issue. For example, the fact that some students struggle with too many exams on at the end of the semester.
Insightful is everything that you did not know. Or, you may have had a hunch or a suspicion. Insights are findings that contradict your knowledge, confirm your suspicions or quantify the importance. For example, if the analysis reveals that 90% of students in a college struggle with too many exams, this is insightful and worth re-thinking.
An insight is a finding that contradicts your knowledge, confirms or denies your suspicion, or quantifies the importance.
Actionable insights lead to either adaptation and action or confirm the fact that no action is required. Actionable findings are those that translate into concrete actions. Companies need to ask themselves: What can be actioned on? What hasn’t already been actioned?
How actionable are your insights?
As an exercise, try to find examples of actionable insights in your business. We’ve come up with a few examples below. Which out of these would you say are actionable insights?
- Our NPS score this month dropped by 15 points
- Passengers complain at missed flight connections
- 20% of customers talk about price
- Buyers say that clothes sold by a competitor are better quality
- People talk about our brand more positively following a ban on plastic bags
- Twice as many Detractors talk about Product’s ease of use
In fact, the first 3 bullet points are not actionable for the business, and not very insightful.
3 Examples of actionable insights
Insight > Adaptation > Action
You will need critical thinking to turn insightful findings into actions. For example, you could solve the lack of parking on campus not by providing more parks, but by working with the city council to improve public transport options. Or, to give a positive example, if students say that they love the environment and the campus, the action could be to use this finding in the marketing material to attract students who care about this.
Insight > No Action Required
Not everything is worth measuring, but data analysis can validate one’s assumptions. The analysis can lead to insights that aren’t necessary actionable but are just as vital. For example, you may think that class size is an issue, but if students don’t mention it, no action to fix it is needed.
Insight > Rethink strategy
Data analysis can also help validate if a strategy implementation is working or not. Let’s assume: last quarter students complained that university staff wasn’t helpful. After taking measures to change that, this quarter’s results should demonstrate if the measures worked or need further thought.
If your customers are saying that your competitor makes better quality clothes, that is a key insight that you can action. You can further enquire why (by asking for feedback through a survey) and drill down into what exactly the factor that they like about your competitor. If your company is a supermarket chain, and you ban the use of plastic bags for a couple of your franchises, this can have a positive effect whereby customers feel you’re doing something positive for the environment. As an action, you can enforce this change nationally. If your detractors are getting frustrated that your product is not easy to use, that’s something you can action and do something about straight away.
Can today’s software find actionable insights in data?
In my opinion, despite all the promises, none of the today’s solutions can ingest data and spit out actionable insights.
Why? Because separating actionable and insightful findings from other types of insights (non-actionable/insightful, non-actionable/non-insightful and actionable/non-insightful) would require two types of knowledge:
- objective knowledge of difficulties associated with different actions,
- subjective knowledge of what’s old news and what’s genuinely insightful.
Eventually, AI agents may be able to learn the objective knowledge by reading materials published by people over the years. And they may even build up the subjective knowledge over time by working alongside its user. Unfortunately, we are still far from the inventions of science fiction.
What AI can deliver today is the ability to sift through the data more efficiently. When it comes to making sense of people’s comments, NLP algorithms can turn people’s comments into themes that can be analyzed just like numbers. Following that, when it comes to making sense of structured data, data visualizations help understand differences, uncover correlations and detect trends.
When evaluating an AI solution, use the following questions to test it:
- Will this solution tell you things about your business that you don’t already know?
- How easily will you be able to separate signal from noise?
- Will it be able to identify trends in data without having to specify them in advance?
At Thematic, we help companies to find insights by digesting customer and user feedback in easy-to-use data visualizations.