Artificial Intelligence customer experience Customer Experience Management Data Science games IBM Watson Studio Machine Learning

Using IBM Watson to Answer Two Important Questions about your Customers |

Using IBM Watson to Answer Two Important Questions about your Customers |

Buyer expertise administration (CXM) packages are essentially a quantitative endeavor, requiring CX professionals to decipher insights from a sea of buyer knowledge. On this publish, I’ll illustrate how you should use IBM Watson Studio to research one supply of buyer knowledge, buyer survey responses, to reply two essential questions concerning the well being of your buyer relationship: 1) what’s the present degree of satisfaction throughout the CX contact factors and a couple of) which of those contact factors is chargeable for making certain clients are loyal?

Buyer Expertise Administration (CXM) packages depend on several types of knowledge that come from quite a lot of sources. The preferred supply of buyer suggestions is surveys. Buyer surveys are used that will help you reply two broad questions concerning the well being of your buyer relationship:

  1. What’s the present degree of buyer satisfaction with totally different elements of your enterprise (i.e., product, ease of doing enterprise, technical help)?
  2. What’s the probability your clients will exhibit several types of loyalty behaviors towards your organization or model (e.g., probability to remain, advocate and purchase extra).

These two questions will assist you perceive how properly you’re assembly the wants of your clients and, extra importantly, perceive what you should do to enhance buyer loyalty. In in the present day’s submit, I’ll current a case research as an example how you need to use IBM Watson Studio to reply these two questions.

Why IBM Watson Studio?

IBM Watson Studio, an end-to-end analytics answer that will help you achieve insights out of your knowledge, was designed for knowledge scientists, software builders and material specialists to collaboratively and simply work with knowledge to construct and practice fashions at scale. Of the totally different deployment choices which are out there for IBM Watson Studio, I signed up for IBM Watson Studio Desktop and I will probably be utilizing this deployment on this submit.

IBM is utilizing the facility of its Watson Studio platform to increase the facility of AI to individuals who fall outdoors the realm of knowledge science, machine studying and AI specialists. IBM Watson Studio is an end-to-end analytics answer that will help you achieve insights out of your knowledge. Watson Studio completed this feat by offering a platform that will help you put together knowledge and construct fashions by yourself desktop utilizing their easy-to-use visible drag and drop instruments.

Earlier than diving into IBM Watson Studio, it’s necessary to offer some background on each the survey knowledge and the analytics behind driver evaluation.

Case Research: The Survey Knowledge

An organization conducts an annual buyer relationship survey to assist them perceive and enhance the well being of the client relationship. Utilizing a zero (Extraordinarily Dissatisfied) to 10 (Extraordinarily Glad) scale, the client relationship survey requested clients to point how glad they’re with the client expertise throughout eight enterprise areas or contact factors:

  1. Ease of Doing Enterprise
  2. Account Administration
  3. General Product High quality
  4. Buying/Invoicing/Supply
  5. Utility of Site
  6. Technical Help
  7. Communications from the Firm
  8. Future Product/Firm Course

Utilizing a zero (Under no circumstances probably) to 10 (Extraordinarily possible) scale, the survey additionally requested clients to point their probability of partaking in several types of loyalty behaviors:

  1. Advocacy Loyalty (e.g, probability to advocate, purchase once more, general sat)
  2. Buying Loyalty (e.g, probability to broaden utilization, purchase several types of merchandise/providers)
  3. Retention Loyalty (e.g., probability of renewing service contract, staying)

IBM Watson Studio Desktop

To reply the 2 questions posed earlier, I’ll comply with a 5-step course of for the present case research:

  1. Load Knowledge
  2. Add and Put together Knowledge
  3. Discover Knowledge
  4. Analyze Knowledge
  5. Talk Outcomes

To make use of IBM Watson Studio, merely comply with the directions to put in it in your desktop. After putting in and logging into the Watson Studio platform, you possibly can create a brand new challenge or handle an present undertaking. Within the screenshot above, you possibly can see that I have already got three tasks. I’ll click on on the New Venture button to create a brand new challenge.

Subsequent I’ll create a New Challenge through which I’ll attempt to reply the 2 necessary questions.

1. Load Knowledge

Step one is to load the info into Watson Studio. The uncooked survey knowledge are in an Excel spreadsheet, with the columns representing the totally different variables in your knowledge set. Utilizing machine studying nomenclature, consider the client expertise (CX) contact factors because the options of your mannequin and the three several types of buyer loyalty as your outcomes or standards.
I uploaded this knowledge set by clicking on the higher proper “browse for information to add” hyperlink and choosing it from my laptop computer. That knowledge set might be saved as a Knowledge asset.

2. Add and Put together Knowledge

After you might have uploaded your knowledge set to Watson Studio, the subsequent step is to discover the info set to raised perceive the variables in your knowledge set. Watson Studio makes use of a Modeler Stream (click on Add to undertaking button) that will help you with this process and different, extra in-depth analyses.

The Modeler Circulate is predicated on IBM SPSS Modeler that makes use of Nodes to carry out sure features in your knowledge. The varied Nodes might be discovered within the Palette window on the left aspect of the appliance.

Add Knowledge. On this step we have to first import the info asset to the Modeler Stream. Merely drag and drop the Knowledge Asset Node into the primary a part of the Modeler Stream window after which double-click the icon to seek out and join your knowledge set.

Put together Knowledge. Subsequent, you will have to organize your knowledge for exploration and evaluation. Merely drag and drop the Auto Knowledge Prep Node into the Modeler Circulate window. The Auto Knowledge Prep Node routinely prepares your knowledge for evaluation (e.g., altering strings to numerical values). Join the 2 nodes by clicking and dragging from the 2 circles on the respective Nodes.

Subsequent, to view the ready knowledge, merely click on the higher proper nook of the Auto Knowledge Prep Node and click on “Preview.” A brand new window will open with the brand new knowledge. I see that, though my uncooked knowledge have been initially in string/textual content format (although they have been numbers), the Auto Knowledge Prep Node reworked these values to numerical format. Along with mechanically getting ready your knowledge, Watson Studio lets provides you choices to conduct further operations, together with filtering and binning variables in addition to choosing samples.


three. Discover Knowledge

Subsequent, we will discover our knowledge by calculating some descriptive statistics for our measures. Merely add the Statistics Node to the window and join it to the Auto Knowledge Prep Node. Click on on the higher proper of the Statistics Node to open the window to pick the variables you need to embrace within the evaluation (Add Columns) and the statistics you need to see for these variables and click on save.


To view the statistics, click on on the Statistics Node and hit run. An Output panel will speak in confidence to the correct with an inventory of icons, every representing a specific evaluation you ran. Double click on the icon to view the outcomes (see under). You’ll be able to see the descriptive statistics for every of the variables chosen (scroll down if mandatory). You possibly can examine for outliers or invalid values to make sure your knowledge are clear for evaluation.

You may also view the values in graphical format. Add the Histogram Node to the Modeler Stream window. Once more, connect the Histogram Node to the Auto Knowledge Prep Node, then open the Auto Knowledge Prep, add the variable of curiosity. Save then run the node. The Output panel will open up once more with a brand new icon representing the outcomes of the evaluation. Double click on the icon to view the outcomes (see inserted image under for Advocacy loyalty). The info look good (all variables inside potential ranges (zero to 10)).

four. Analyze Knowledge

The subsequent step is to research the info. For our instance, to reply our questions, we have to take a look at two forms of analytics: 1) Descriptive and a couple of) Predictive. Descriptive analytics are used to point the present state of the world. In our case, descriptive analytics helps us know the present satisfaction degree of every buyer contact level and the present state of buyer loyalty. The extent of satisfaction is listed by a abstract statistic. Whereas totally different abstract statistics can be utilized, I exploit the imply rating (sum of all responses divided by the variety of respondents).

Predictive analytics are used to make predictions about future occasions. In our case, predictive analytics helps us understand how strongly/weakly every of the client contact factors is said to buyer loyalty. This info tells us the relative significance of every CX contact level in making certain clients are loyal. The connection between CX contact factors and buyer loyalty could be calculated by merely correlating the scores of the contact level with the client loyalty measure. This correlation is known as the “derived significance” of a specific contact level. If the survey has a measure for every of the eight (eight) contact factors, we’ll calculate eight (eight) correlations for every of the measures of buyer loyalty. Larger correlations (max is point out a robust relationship between the contact level and buyer loyalty (e.g., enterprise space is necessary to clients). Low correlations (close to zero.o) point out a weak relationship between the contact level and buyer loyalty (e.g., enterprise space shouldn’t be necessary to clients). I added the three loyalty metrics to the mannequin. As a result of analysis has proven that several types of loyalty are impacted by several types of CX contact factors, I wanted to construct totally different fashions for every sort of buyer loyalty.

In IBM Watson Studio, I chosen the Statistics Node from the Outputs choices to calculate the means for every variable and their correlations with every of the three measures of buyer loyalty. Double click on the Statistics Node so as to add the eight CX contact factors (predictors). You’ll be able to then verify the precise descriptive statistics you have an interest in analyzing (I added rely and imply). Scroll down in that panel and you may add the three loyalty measures to which you need to correlate the eight CX contact factors. Hit save. Then run the Statistics Node to view the descriptive statistics of all predictors and their correlations with every of the measures of buyer loyalty. See under. You possibly can scroll right down to view your complete output.

5. Talk Outcomes

Technique of CX Contact Factors and Their Correlations with (Influence on) Advocacy, Buying and Retention Loyalty.

Now, we’ve got the solutions to our two questions: What’s the present degree of satisfaction for our contact factors and what are the correlations with every of the measures of loyalty. You need to use the output in your reporting. I took the additional step of consolidating the outcomes right into a easy desk (See Desk 1).

With these knowledge, we will draw some conclusions concerning the well being of the client relationship. First, from our earlier evaluation within the exploratory part, we see that the means for the three measures of loyalty are: Advocacy: eight.three; Buying: 6.9; and Retention: eight.three. That’s, Advocacy and Retention Loyalty look like comparatively excessive whereas Buying Loyalty is comparatively low; this means that, despite the fact that the purchasers usually tend to advocate the corporate and stick with them for a very long time, they’re much less more likely to increase their relationship with the corporate.

Clients are most glad with Product High quality, Technical Help and Communications from the corporate. Clients are least glad with Utility of Website online, Buying/Invoicing/Supply, Gross sales Course of and Firm’s path and future merchandise/providers.

Second, we will determine the CX contact factors which have the most important impression on the several types of buyer loyalty. The highest 4 drivers of Advocacy Loyalty are Ease of doing enterprise, Communications from the corporate, Firm’s path and future merchandise/providers and Product High quality. The highest 4 drivers of Buying Loyalty are Firm’s course and future merchandise/providers, Product high quality, Communications from the corporate and Ease of doing enterprise. The highest 4 drivers of Retention Loyalty are Technical Help, Ease of doing enterprise, Communications from the corporate and Firm’s path and future merchandise/providers.

I observed that, whereas Site utility acquired the bottom scores of satisfaction, it additionally has the bottom influence on buyer loyalty. That’s, Website online utility doesn’t look like as necessary to driving buyer loyalty in comparison with the opposite CX contact factors. Enhancing Site utility will do little or no to enhance buyer loyalty.

To enhance buyer loyalty, this firm may initially think about focusing their CX enchancment efforts on these CX contact factors which have a low degree of satisfaction and have a big influence on buyer loyalty. Based mostly on the outcomes, it seems that Firm’s path and future merchandise/providers could possibly be an space the place there’s room for enchancment and it has a comparatively excessive impression on the three forms of buyer loyalty.


On this article, I illustrated how you should use IBM Watson Studio to research buyer survey knowledge to reply two necessary questions concerning the well being of your buyer relationship. In pursuing the solutions, I discovered Watson Studio Desktop was pretty straightforward to make use of. I referenced their on-line documentation a couple of occasions within the course of and located my solutions shortly, with hyperlinks to documentation and different nice assets together with IBM Watson Studio Group.

This publish was delivered to you by IBM Watson Studio. I acquired compensation to write down this publish however all opinions expressed are my very own.