OUR EXPERIENCE IN ESPOCRM MIGRATION AND BUILDING ML MODELS

29 Oct 2019
Ihor, back-end developer
We channel all our efforts into building a trustworthy relationship with each customer.
all news

The key point that makes any cooperation successful is confidence. Confidence in the product, in the partner and the client’s confidence in the company. However, building trust is a pretty tough task that requires much effort and time. M.Water co. – is a big water delivery service that proved to be trustworthy among its clients. They provide 15-liter water bottles to organizations, factories, and normal households on a weekly basis.

Initially, to have all the data and processes arranged at hand, the company used SAP because it is a popular CRM. However, it has its drawbacks that M.Water wanted to change. The license and customization cost a pretty penny. The company must purchase the software and hardware necessary to run the programs company-wide.

They needed to replace legacy call-center CRM with a customized one. Although it still needed modifications, EspoCRM turned out to be a great choice. M.Water addressed SapientPro’s developers to help them with it.

There are several advantages of EspoCRM for the Thai water delivering company:

  • It is an open-source CRM with own community;
  • Flexible solution for any business project;
  • No need for 500 k payment per year for SAP license;
  • Super fast and intuitive in the working process.

Our first test task was to prove to our client that EspoCRM is scalable and can handle big databases. We had around 11 million records to migrate from SAP quickly.

During our presentation, we showed that CRM is upscaling to a large array of data like that. Then the development for M.Water started. Besides simple data migration, we had to customize the CRM to meet the client’s needs.

The call center consists of around 200-300 people working there. And the number of clients overcomes drastically these figures, so the editing of the delivering schedule or processing any other client’s order takes a lot of time.

The system makes it easier for the company by automating the whole process. In such a way it becomes easier to decide who takes the task, who closes it, who assigns the task to other workers. The operator can use a comfortable interface and conduct all the operations much faster.

The system tracks productivity on each step of the working cycle from assigning requests to the delivery process and generates detailed reports based on collected statistics. These reports help to determine employees with higher and lower productivity. It is also handy to detect various problems and find ways to solve or prevent possible issues.

After developing that project our team was asked to do another captivating R&D task. In spring we started working with Machine Learning that became interesting and important experience for us.

What we had to do is to build an ML model that would analyze a great deal of data and make a prediction of possible loyal clients and those, who are likely to finish the cooperation with the water delivery company.

ML technologies are very popular nowadays and there are a lot of them. For implementing the project we chose Support Vector Classifier (SVC) because it has the probabilistic classifier – a classifier that is able to predict the likelihood of something happening.
The main challenge in this project was that we had piles of data but didn’t have any specific parameters to use for calculating the client’s loyalty.

The parameters were chosen in several stages:

  1. We selected which parameters were relevant by cooperating with the customer. As all the data is in the Thai language, they helped us to translate the necessary parts;
  2.  The analysis of each parameter and dropping those that don’t influence customers decisions;
  3.  Grouping of each parameter – building models (choosing from different types of models available and customizing them);
  4.  models testing – if it does not predict with the desired accuracy, we go back and make some changes into the previous stage until the final result completely satisfies the customer.

Sure, the reasons why a client may stop having water delivered from our customers are numerous and some situations cannot be foreseen. However, with the help of this model, we could achieve the needed 91 % accuracy of prediction which customers are about to leave the company. For the company, it became a great advantage to see where the changes should be done so that their clients continue to cooperate with them.

The work on the project was thought-provoking and engaging because it was something new. The challenge was that nobody does that, as it’s not a direct engineering task – it involves knowledge of economics. The algorithm isn’t mentioned anywhere.

We were provided with a lot of technical documentation –and thanks to an experienced PM, who explains the material very clearly, we didn’t have much trouble with that.

Thanawat Laohadtanaphorn:
SapientPro was able to follow a strict agile process effectively, ultimately allowing for weekly deployments. They offer an economical price model, professional approach, and a team of highly skilled developers… I believe that SapientPro’s clients are in good hands.

 

We’ve been working with M.Water.co for 2,5 years now and the reliability of our partnership is constantly growing.

Our DevOps team is working to support their server, we’ve already done design and e-commerce projects for our Thai partners. The trust is still growing and we are ready to do more. Currently, we are planning a new version of CRM with new features and hope this project to be a great success.

share this article in:
HOW FRONT-END HAS EVOLVED
Max Tatarchenko, CTO
HOW FRONT-END HAS EVOLVED
read more
Back
© 2019 SapientPro. All rights reserved.
career design typing career design typing Ресурс 1