I have a project with data of sales field officers who visit their customers and enter the progress details. Visit can be an order or any kind of customer interaction.

Let's say one sales guy has around 1000 customers, It's only natural that he might skip some customers which can result in loss overtime. So I have the data for the visits done by him since the customer was onboarded. What model should I use to check the old frequency of visits done by him and factors which are stated below so that I can say like:

"Do you want to get in touch with "this" customer. ?"

So basically by checking his previous interaction rate, In case if he forgot to visit, I need to recommend like that.

Data points are like:

  • date/time - date when visit done
  • remarks - what was discussed
  • there is a status which is related to internal
  • and some other customer details.

So what model should I use or which technique do you think is best suited for this problem. I'm fairly new in machine learning and kind of learning it by doing.

  • 1
    $\begingroup$ It is not clear to me what the exact objective is. Do you want to prevent customer churn? Or maximize some return? $\endgroup$
    – Peter
    May 17, 2019 at 18:21
  • $\begingroup$ Yes in turn to reduce that. $\endgroup$ May 20, 2019 at 3:55
  • $\begingroup$ This is a very clear question, only the optimization goals could be stated more clearly. If you want to minimize customer churn, it should be quantified somehow (by the sales team) and be part of the data set, I guess. $\endgroup$ May 23, 2019 at 14:31

1 Answer 1


What I understood is, since your sales person skipped the customer meeting you want to display details or send a notification like:

  1. Last date of conversation
  2. Discussion points and so on.

such that sales person decide her/his priority based upon the importance of discussion.

If so, I'm not sure how much it's possible with ML. But one thing is possible, where you can trained RNN(Recurrent Neural Network) based time series model. This model will predict the next meeting time, which is based upon the pattern of historical meeting data.

In that way you can book sales person calendar along with the last meeting Key-Discussion points or send an update in case sales person hasn't attended that person.

  • $\begingroup$ Thanks I will look into this! $\endgroup$ May 20, 2019 at 3:56
  • $\begingroup$ Do you have a working example or any article I can follow? not sure how to train a model like this? $\endgroup$ May 27, 2019 at 5:55
  • $\begingroup$ I have pushed a sample example on github. github.com/vipinbansal1/multipleparalleltimeseries. Its just a basic example for PoC. What I did here is, I feed the series of date to my model and trying to predict the next date. $\endgroup$ Jun 5, 2019 at 10:17
  • $\begingroup$ Please dont go with the results as I just feed a very small sample. $\endgroup$ Jun 5, 2019 at 10:19
  • $\begingroup$ thank you for the answer! can you please add comments at some lines. Sorry I'm trying to understand how to use tensorflow. $\endgroup$ Jun 17, 2019 at 6:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.