Increase Customer satisfaction with Net Promoter Score
Net Promoter Score is a quick and easy way to measure customer satisfaction. Originally, the method was based on asking a single question. In this article I'll show how the original method works, and how it can be enhanced by adding two more questions. I'll also discuss some of the claims and counter claims about how useful the method really is.The original method is easy to use:
- Ask your customers a single question: "On a scale of 0-10, how likely are you to recommend us to a colleague or friend?"
- Count the number of 0-6 answers. These are your detractors.
- Count the number of 7-8 answers. These are your passives.
- Count the number of 9-10 answers. These are your promoters.
- Calculate the percentage of promoters and detractors.
- Subtract your detractors from your promoters to get your Net Promoter Score.
The result can be shown in a graph like the one above. A score by itself means little. The important thing is to track trends over time.
The following table shows (made up) raw NPS data:
I like to set up the spreadsheet so that the data is automatically color coded. It gives me a sense of where things are going while I enter the data.Here is how my spreadsheet handles the conversion from raw data to NPS score:
The Net Promoter Score tells you how satisfied your customers are, but it does not tell you what to do to increase customer satisfaction. Let's fix that!It would be interesting to know the distribution of our three customer categories. The table above tells us that, but it is better to visualize the data:
If we only knew what to do, we could use the staple diagram above to evaluate our actions. Let's add a second question to our survey:What is the main reason for giving the score you did?Now we can figure out why we get the score we do. I'd recommend doing some serious analysis and synthesis here, because you cannot take customer responses at face value. You have to put them in context. For example, remember Apple´s Time Machine that I wrote about awhile ago. If you did a survey on the time machine, and asked only professional system administrators, the score is likely to be low, but in this case, that only indicates you are targeting the wrong user segment.
Based on the replies to the second question, and the NPS Breakdown table, you can decide whether to focus on moving detractors into the passives category, or moving passives into the promoters category, and how to do it.
It is also possible to find out other things. For example, you might have a product which the users love, but still does not sell. In that case, your constraint is not in customer satisfaction, but somewhere else. You might even have a product which people really don't like, but which sells anyway. Information like this can be extremely valuable.
For example, the Southland Corporation embarked on a costly program to improve customer service at its chain of 7-Eleven stores. The goal was that every customer should be greeted with a smile, eye contact, and thanks. They succeeded in increasing courtesy, but afterwards they discovered that the service improvements had little effect on profitability. (From Hard Facts, Dangerous Half-Truths & Total Nonsense, by Jeffrey Pfeffer and Robert Sutton.)
If the Southland Corporation had bothered to analyze the situation before acting, they could have avoided a costly mistake. An NPS survey would have revealed the same thing the Pfeffer-Sutton study did, that 7-Eleven customers want fast service, not insincere greetings.
Oh, and you may want to tack on a third question:
May we contact you again?This is not possible in all situations, but it is useful if you have a steady control group, so that you don't get hit by random variation in small samples. The thing to watch out for, of course, is that if you have a steady control group, you may get more and more positive answers over time just because you keep asking people what they think. But I know you already knew that.
Before using it, you should know that there is some controversy surrounding Net Promoter Score. Fred Reichheld, who invented the technique, claimed to have found a strong correlation between NPS and corporate growth. His claim was affirmed by some researchers, and refuted by others. (There is a Wikipedia article that references material from both camps.)
From a Theory Of Constraints perspective, if customer satisfaction is the system constraint, then you will benefit from improving it. If customer satisfaction is not the constraint, improving it will yield little benefit.
In either case, tracking it to know whether it is the constraint or not, is useful.
Comments
When we started out 4 years ago to measure Net Promoter Scores for major electronics companies, we used exactly this technique. We made simple surveys, and each month loaded the results into a spreadsheet, and tracked NPS and customer comments.
It worked fine - we were able to make some significant improvements within weeks, and the score improved by 0.5 - 1 points every week.
I can definitely recommend this as a great place to start NPS.
But to make NPS work well, you need to do it on an industrial scale - sampling every transaction. And after three months with a spreadsheet, we started running into problems.
1. It took too long to update - we had to do a manual import - almost one day a month to do it right
2. The spreadsheet was growing - each time we made a new pivot table, the monster grew - eventually to 20Mb!
3. It was not real time - and impossible to keep track of customer comments - what had been actioned.
So after a while we decided we could create a better tool than a spreadsheet, and a better way of attacking NPS than the best known surveying tools.
We wanted a tool that would sample customers automatically, could provide tracking graphs by segment in real-time, and then help us answer and address customer issues.
So CustomerGauge was born and we launched it last year. It's designed to be simply the best NPS tool out there - simple to set up and easy to use internally, with graphs and real-time displays.
You can find out more on Net Promoter and engaging customers on our blog enGaugement. Hope you find it interesting.
Adam
1. You do not have to sample every transaction, only a statistically significant sample. Figuring out what a statistically significant sample is, can be a bit tricky, but you can always ask a statistician. (They need to make a living too.) Sometimes it is easier to just ask the whole lot of customers, but unless everyone answers, you can unwittingly introduce bias in the results.
2. When using automated software you get answers only from people who care enough to reply to the automated questionnaire, which tends to polarize results. You also filter out the people who are not comfortable using computers, or who won't reply to a web survey no matter what.
Automated questionnaires can be very helpful, but they are not a panacea.
Of course, manual sampling can also introduce bias. For example, an attractive female asking questions might get different answers than, say, me.
One way to guard against such bias is to use different methods to gather data, and see if the method affects the results significantly.
3. When doing a survey, it is important to have a good idea, beforehand, how the results will be used. I have participated in, and analyzed, a couple of surveys recently. All of them shared the same problem: the survey results will not be readily translatable into specific actions. That is one area where I believe the NPS approach can be of great help.
4. I like your blog.
My own survey of this blog (see a later posting) will introduce bias in the results. I expect the people who answer to be different from average management blog readers in several respects. In my case, I suspect the bias will not hurt my efforts - I am targeting a specific kind of reader, rather than trying to attract everyone who has an interest in management. (If I did, my blog would be quite different.)
Thanks for the well considered answer, and at the risk of this turning into comment table tennis I wanted to add one thought to your Point 1.
"1. You do not have to sample every transaction, only a statistically significant sample..."
Well, I agree of course. But actually a correct answer is only half the story with many of the companies I work with. The most surprising fact I learned about consumer electronics is that they often forget who the the consumer is. They erect many obstacles between themselves and the consumer in the form of retailers, distributors, out-sourced callcentres and so on.
At some stage, smart executives realise that this is the 2000s and we can do things differently - and permits employees to contact or even sell directly to the consumers.
But changing cultures is hard work. And so many managers are happy to hide behind stats, and generalisations.
It takes a really brave manager to listen to every customer comment, and a still more daring leader to make sure his team act on what customers say. There are a few of them out there, and they are skilled at balancing stats as well as individual comments.
I hope you are not offended by my stout defence of sampling everything!
Kind regards, and thanks for your nice comments.
My only problem with comment table tennis is that I enjoy it a bit too much.
I take no offense, and I do believe it is better to sample a bit too much than to sample too little. The zone between "sufficient" and "more than necessary" is large enough to accommodate most of us who are interested in gathering statistical data.
We do agree on the really important stuff: management needs accurate information about what the customers think about their company's products and services.
We also agree that the information must be gathered as quickly as possible, with a minimum of effort and pain. (Especially to the customers, which is a point some companies miss.)
And, we agree that the most difficult part often is using the data, not collecting it. As you point out, this may require a considerable change in corporate culture.
Interpreting data from customers can be...challenging. One of the things I like best about the NPS approach outlined in my article is that the data gathered is actionable, that is, it is possible to work out a plan of action based on it.
From my point of view, sampling everything or a statistically significant subset is a matter of economics and expertise. Sometimes one method is cheaper, sometimes the other. Sometimes you have a good statistician on hand, sometimes you don't.
In either case, it pays to find out about possible bias in the sample. For example, by investigating customers, we find out why people buy a product or service, but it will not necessarily tell us why people who went to the competition did not buy from us.
Since the number of people who do not buy a product generally outnumber the people who do, and we generally know even less about them than we know about buyers, it means we know least about the largest group of people of interest.
Just musing a bit:
If I could figure out a way to identify people who could have bought, say, my services, but did not, and then ask them about it without being annoying, then I might be able to use the information to increase my own revenue stream.
An automated tool could come in handy, partly because of the number of non-buyers, partly because it may very well be less annoying than I am. :-)
(My main business is getting my customers to ask the right questions, and then teach how to work out the answers, rather than just providing the answers to them in a neat little report. A consultant can't possibly be more annoying than that.)
Err, I suppose the above proves my point about enjoying comment table tennis a bit too much... :-)
Anyway, thank you for your comment.
The NPS spreadsheet is in Numbers format. It is easy to convert to Excel, but given the information in the article, you can easily construct an NPS spreadsheet of your own.
Since I am in the business of teaching, I would much prefer it if you did that. However, if you run into trouble, email me your spreadsheet, and I'll be happy to help.
(Instead, you are going to be annoyed because you'll never find out what the joke was. Believe me, you gained from the trade-off.)
-mel-