Have you ever been asked to complete an NPS (net promoter score) survey which only asked you to rank how likely you are to recommend that company, service, or product? Did it leave you wondering why they didn’t ask for a reason you gave them that score? You wanted to provide feedback! And they didn’t give you the chance!! How could they?
Now imagine you’re that organisation. While the overall score will be useful to give a preliminary indication of brand health and perception, it lacks actionable insights into where to focus your attention to improve the result and provide a better customer experience – or know what to continue doing well to maintain it. Wouldn’t you want to know?
NPS continues to be one of the most popular ways of learning how your customers feel about your organisation. While many NPS surveys already include a follow up question asking why they gave the score that they did, unless you only have a small customer base – or a very large team of people to go through the answers – it’s unlikely you’ll be able to analyse the results, and get a solid understanding of the sentiment and drivers that sit behind the score.
If you read our blog ‘How to improve customer loyalty with sentiment analysis’ you’ll already know that sentiment analysis quantifies the attitudes, opinions and emotions in language. It can be used on almost any text-based data that you have access to – which includes those free-form answers included in your NPS survey.
Better yet, once set up it can do this analysis on large sets of data with little to no human intervention, providing you with meaningful, actionable insights into your detractors, promoters, and those sitting on the fence.
Why would I want to apply sentiment analysis to my NPS surveys?
Classic NPS survey results consists of a number from -100 to +100 indicating whether your customer base are promoters, detractors, or just kind of meh about the whole thing. It’s based on a simple bit of math subtracting the % of detractors from the % of promoters to give you that final score.
But there could be numerous reasons why you’re getting that particular number. Your customers could be receiving the most amazing customer service from one of your support staff, so they’re giving you 10s. Or your shipping is slow, and doesn’t keep customers in the loop about where their package is. In their frustration they're giving you a 0. On another day these same people might have given you a very different score based on different circumstances. Therefore knowing the reason behind the number will help you drill down into any issues there may be.
NPS isn’t an exact science, and everyone thinks and reacts differently. People will base their likelihood to recommend you on different reasons because they value different things. Two people can have exactly the same service or experience but give you a score that puts them into different NPS segments – even if their follow up responses are practically identical.
Sentiment analysis can therefore offer a much more accurate insight into whether someone is actually happy, or whether they just picked a number at random. People also tend to be more conservative with the number they give you, but a lot more transparent in the comments they write.
Ultimately, people are complicated. Trying to boil it all down to a number between 0-10 severely limits your ability to understand the nuances into what’s really going on, and thereby identify what you need to do to improve.
Understanding the drivers behind the sentiment
Applying analysis on the text portion of the NPS responses gives you an opportunity to dive into the reasons they gave you that number - also known as ‘drivers’. Through this you’ll learn that they gave you that 10 because they love your customer service, or a 0 because your shipping is slow and lacking in sufficient communication letting customers know the status of their delivery.
You can also segment on these drivers to get an understanding of the overall sentiment relating to that particular point, and whether this is the norm or an anomaly. Thereby you learn where to focus your attention on making changes.
For example: we applied sentiment analysis to the NPS responses of one of our clients where we had identified the drivers behind the score they gave. We then cross referenced this with how important it was to the customer.
The results showed that performance was strongest in the things least important to the customer, and vice versa. To improve their overall NPS (and customer satisfaction) our client would need to focus on improving their performance in the areas most important to their customer if they wanted to turn detractors into promoters.
The analysis showed that performance was strongest in the things that were least important to the customers.
Previous research showed that customers care about:
Time it has taken to sort their issues
Keeping them informed
If the focus was to drive performance higher in these areas, then there is the opportunity to shift detractors into promoters.
How do I get started?
As with most initiatives you first want to ask yourself what you’re trying to achieve – what’s the objective? This will not only help you focus how you initially structure your NPS, who you send it to and when, but will also influence the type of sentiment analysis you apply to your data and the insights that you’ll get out of it.
There’s a lot that goes on behind the scenes in terms of algorithms and machine learning to deliver the results that you’re after, so take some time to work through all the possible outcomes and information that you want to have in hand. If you’re a larger or more complex organisation where customers can have multiple touchpoints, you’ll want a more complex system to understand the drivers behind the sentiment.
Is there an app for that?
There are plenty of online resources, and even tools, that can help you get started with applying sentiment analysis to you NPS results. However, unless you know your way around code, or are happy with a more generic application of sentiment to the text, you’ll want to find an experienced data scientist to help you through the process.
They’ll be able to identify the most appropriate algorithms to use, how to apply any machine learning, and identify key words or phrases that will help identify trends and further segment the results.
And that’s not even the best part
Now that you have all this information at your fingertips, it’s time to do something with it. After all knowing why someone does or doesn’t love you is only half the battle.
Having identified the aspects of your organisation, products, or services that aren’t living up to customer’s expectations you can now look into why it’s not performing so well and how this can be improved.
For those whose items still haven’t arrived you’d look into the shipping process. Do you need to change suppliers? Do you need to set up an automated programme that sends text or email updates as the item passes certain milestones? Or are there internal processes that could be improved, like faster time to package and pass to courier? Because if The Iconic can deliver my package from Australia (and through customs) overnight – why can’t the shop just down the road get it to me at least within the same time frame? Setting up a matrix like we did for our client will help you prioritise improvements based on how important they are to your customer and how low you’ve been scored on it.
Understanding the reasons and sentiment behind the numbers – and then using it to drive improvements within your organisation to create a better customer experience – is the most effective way to gain value from a tool like NPS and truly drive business growth.