I have always used Excel in order to build the prediction model for customer churn. Although I am not an expert in Tableau, Qlik or similar platforms, the tests I have done with them made me conclude that they are excellent tools for data representation, but data need to be ready for such representation, what in fact makes using Excel prior to representing its data and results.
I would also add that these tools make sense depending on the nature of the problem and your database size and structure. Furthermore, if you want to just predict the churn, Excel is probably enough, but if you want to find the key reasons that lead to that churn, and how they impact in the actual rate - so that you can do something about it -, Tableau or Qlik can be great answers to your needs.
Hope it helps. Regards
Hi Jose Luis, thank you so much for your reply. I will be looking into Qlik also but I agree with you in using Excel.
I have worked with a company called Sagacity who have a clever system QTOX. If you want to know more please contact me directly.
Hi Paul, how can I contact you?
Jean-Pierre I have been involved the last few months with a company: www.adazza.com , introducing them to Telco’s in Africa. They have a cloud based, subscription model platform, which would deliver what you are looking for very quickly. They are totally focused on the Telco market though. If any use I’ll introduce you to the CEO.
That’s interesting, can you tell me more about the features, pros/cons? Thanks, Paul.
Depending on the size of your data, Excel may be a bad choice. You need to work with your DWH or MIS teams to make sure that the summarized tables have the information you need before you go into a tool like Tableau. I’m using a lot Power BI these days and find it more user friendly to use than Tableau. Once you get the data represented, it’s easy to drill down and filter and that way understand how your subs behave before they churn. The advantage of these tools is that once the model is created it will run by itself as data is updated in the tables (as long as you use relative dates in your main query). Power BI is free unless you need to publish but again, if your dealing with more than 2million rows in your data-set, than Tableau would be a better choice.
Thanks, Luis. I will have a look at Power BI.
Hadoop BDA stack and using a proprietary optimised combination of Data Cleansing, Classification and Regression algorithms to accurately predict churn amongst Mobile Subscribers.
Thanks for the insights, Felim. I will definitely research Hadoop
On my experiences, predictive analysis are best build under R, matlab or python. It needs heavy coding but worth to try.
Im software engineer myself so im more comfortable using python with its library or matlab to build prediction models.
Anyway, there are lots of program out there with more user friendly tools such as:
- Anaconda enterprise
- Great tools should be azure machine learning
I have experience with Tableau and Anaconda, thanks for your answer, it is helpful.
Jean, we have successfully utilized sales force.com to analyze churn predictions based upon various changes to accounts over time. Furthermore, we have leveraged our internal IT department to write code against our legacy customer database systems to predict churn when implementing changes to pricing strategies, contract terms and product platform changes. Depending upon the size of the base plus predicting new customer acquisition will determine what platform is best for you.
expanding a bit on Yoedi Hariadi’s answer before I’d like to mention that A.I. probably is a prime tool for this sort of problem. Much useful code is premade and available in Python 3 from anaconda.com. The price that one needs to pay is that you’d need a fairly extensive record of previous churn/no churn situations to train your model.
Hi Eduard, I looked into Anaconda and I am definitely interested in this method of using the premade code. I think traning it will be time-consuming but well worth it if we can achieve good predictions.
In my experience I used SAS Visual Analytics with some insight perfomed by Knime platform in a big data environment. It’s not just about the choice of a platform but rather the complexity of the model you try to best fit with your data. Usually, churn is characterized by heavy tail distributions, very far from uniform ones, therefore slight changings of the more sensitive variables you analyze can cause a steep breakdown in the extent outcomes. Custom work based on refining of your model can be tuned via R language or similars. Kind regards. Giacomo
I agree, it’s about the amount of data and we’re dealing with a pretty significant sample. Thanks a lot for the response.
Hi Jean-Pierre—My personal favorite is “Looker” with my second favorite being Excel. I do like Excel for simple data mining, but when it gets more complicated, then I’ve used Looker to get the information that I need. I have to agree with Luis, for the setup time and time required, I think that Excel is probably your best bet, but if you’ve got a lot of churn that you’re mining and predicting for, then I’d recommend at least taking a look at Looker. It’s a pretty easy platform to navigate and is VERY user friendly. They’ve got 99% of the bugs worked out of it (in other words, I’ve never had a problem with it, but I have heard of some, but that was with huge data projects). For customer churn, I think that either of these solutions would work for you.—Les
I haven’t tried Looker - I don’t mind the time for implementation if it will be helpful over time. Thanks a lot for the recommendation, Leslie.
Fast track projects with limited KPI’s - Excel and Power BI. Expect 95% results; For extended and quite large data projects, use more powerful tools like Tableau and Anaconda, they can get you 98% of results. With more than 300 kpis’ and millions of items to process, make sure you have a nice model first. The complexity and volume of Big Data will need specific solutions and humans to help building the model. My recommendation: Make sure your model first knows your Client live cycle model. If you need a hand, drop me an e-mail (email@example.com)
This will mark this comment as best reply and close your question.
Are you sure?
This will close your question without a Best reply.
Are you sure?
This will report this content as inappropiate to the moderators.
Are you sure?