Business analytics
and forecasting

Identify the metrics you need to make a decision

Analyze past performance and forecast future trends

Build reports and visualizations for your team

Explore business goals

Business goals

Evaluate campaign performance

Identify customers at risk of leaving

Forecast demand

Recommend relevant products to each customer

How we approach analytics

Identify processes that could benefit from using data

Recommend ways to improve your products

Assess the financial impact of projects

Agree on which reports you need and how often

Build reports around the metrics you choose

What you get

A plan for using your data

Product development recommendations

An assessment of project costs and benefits

A reporting schedule

Reports built around your metrics

Project results

75% average reduction in store inventory

Objective:

Reduce inventory while keeping products available in stores.

In 97% of observations, actual sales fell within the forecast range

Solution:

The model forecasts a range of weekly sales for each product in each store, up to a year ahead.

Tools: R, Python

Algorithms: XGBoost, CART, k-means

10% maximum difference
from the expert forecast

Objective:

Automate financial planning at head office by forecasting regional spending.

Over 120 cases of fraud
detected in the regions

Solution:

We integrated a tool built with KXEN and SAP BI into the client’s systems. It uses historical spending and regional economic indicators to forecast costs, detect fraud and adjust the budget.

Tools: R, KXEN, SAP BI

Algorithms: KXEN time-series algorithms

At least 9 subscriber segments for targeted campaigns

Objective:

Group customers for targeted campaigns and identify subscribers who also use competing providers.

86% accuracy in identifying subscribers who also used a competitor’s services

Solution:

The system groups subscribers and describes each segment. It estimates how likely each subscriber is to use another provider.

Tools: R, Python, Spark, Hadoop, Hive

Algorithms: XGBoost, CART, k-means, k-medoids, t-SNE

8 data sources connected

Objective:

Identify and prevent product losses and detect declines in the quality of petroleum products.

Over 1,000 incidents detected

Solution:

We connected the data sources to a Hadoop platform and set up processes to extract, transform and load the data. We built analysis modules and a web application for analysts.

Tools: Python, JavaScript, Shell, SQL, Hadoop, HDFS, Spark, YARN, Hive, Tez, Hue, ZooKeeper, Luigi, Kafka, NiFi, PostgreSQL, Microsoft SQL Server, Oracle, PI System, Django, AngularJS, Bootstrap, CSS, Docker, CentOS/RHEL