DATA SCIENCE & ENGINEERING. ART
The data science team combines coding and math’s.
As a result you will take data driven decisions which will impact all parts of your business and deliver real growth.
WHAT WE DO
-
Data Science
Our data science team brings their programming knowledge and mathematical skills coupled with domain experience to solve business problems.
They delve in artificial intelligence, machine learning and other supervised or unsupervised models depending on the business problem.
-
Data Engineering
Our engineering team provides the data framework that enables our scientists and analysts to solve business problems.
We have made significant progress in this are over the past 12 months, investing in the analytics server and the new MarTech environment.
Check out some of our Projects below:
MLOPS
Machine Learning Operation is the combination of 3 items: DevOps + DataOps + Model Ops. It combines proper governance e.g. version control, logs and lineage of the main components of a Machine Learning powered application: Code, Data and ML models.
By using Databricks platform over Azure we were able to migrate all our ML powered applications to a fully governed and scalable framework using Pyspark and Mlflow.
MLOps compliant pipelines are easy to monitor and maintain, leaving the team able to develop more models to support revenue growth and cost reduction objectives.
Sentiment layout
Data is available either via direct connection to Databricks (internal use) or via public REST API and that allowed us to visualise this datapoint in combination with other useful information such as pricing.
Simply overlaying these 2 data points can give actionable data to our customers and internal analysts.
Data is currently used on email communications and internally via PowerBI reports.
Sentiment Backend in Databricks
Using Databricks were able to combine historical data (from the EDS tables) and the streaming data from the execution venue to calculate near real time sentiment on our markets for all Service Offering. Currently the data includes G2 and MT4. In Q3 FY2024 MT5 will be added into the live stream.
The data points are refreshed every 10 mins and combine trading and customer data so that sentiment can be segmented by Value Tier or regulatory classification.
ACQUISITION DASHBOARD
Provide end to end data from impression to trading so you can understand the ROMI of each channel.
We applied "last paid channel" attribution and the data model can give a comprehensive view of both organic and paid channels.
Per advertisement cost is also included in the Dashboard with a daily granularity.
Data is refreshed every day thanks to our scripts running on Airflow.
GCLID PROJECT
Thanks to our Airflow instance we are able to periodically (every 3 hours, planning to decrease to 1 hour) communicate to the Google Advertisement Platform (Google Campaign Manager) which customers went on to trade or fund.
This deep conversion data point is then used by Google to further optimize our advertisement.
PERSONAS DASHBOARD
Using unsupervised k-means clustering we identified 4 core groups of our customer base.
This allowed us to understand other characteristics such as demographics features. The data-driven study of Personas allowed better targeting in acquisition and also more customized journeys in CRO and CVM.
Persona description is available in the following systems: Google Analytics, Monetate, DWH and Salesforce
-
Segmentation
Segments are used to divide the target market into manageable groups to offer a tailored experience of our products and services
-
Marketing ROI
We evaluate the marketing touchpoints a consumer encounters on their path to conversion and link spend to revenue
-
Lifetime Value
Projecting lifetime value of a client is important for acquisition, retention sales and value management teams.
-
Propensity Models
Be it lead scoring for new acquisition or churn models, propensity models are key to provide a targeted service to a Retail client base
Meet the Team
-
Elettra Damaggio
DIRECTOR OF DATA SCIENCE & ENGINEERING
Doing anything that Anirban doesn’t want to do.
-
Sumeyra Karaca
DATA SCIENTIST
Working on NLP models. Projects: Text to Analytics.
-
Agata Plewa
DATA SCIENTIST
Working on Customer Behavior Forecasting. Projects: Life time value, Personas, Churn Prediction
-
Chitra Pun
DATA ENGINEER
Supporting the team into migrating in the new Martech Platform
-
Hela Momand
DATA ENGINEER
Supporting the team into migrating in the new Martech Platform
-
Cigdem Tuncer
DATA ENGINEER
Supporting the team into migrating in the new Martech Platform
“The analytics team help us identify various commercial opportunities and we are excited to see what they have in store for us in 2022!”
— Sixto Alonso, VP Americas, Retail
Core competencies
Machine Learning
Using supervised and unsupervised machine learning techniques to solve business problems. Examples of what we used so far: K-means clustering, Neural Networks and XGBOOST random forests.
Data Integration / Engineering
Building ETL pipelines to collect, transform and process data from our various data sources in cloud and on premises, supporting commercial reporting and ML modules. The Data Engineering team also builds two-way connectors to make data points available where is most useful for the business.
Languages / Platforms
We code in: Python, SQL, Big Query SQL
We use: PowerBI, AirFlow, DAX, Google Cloud, sklearn and scikit learn, REST API.
Automation
Data automation for end to end reporting via Airflow (task scheduler).
We can develop automated process for Data treatment and consumption.
Let’s connect to discuss how analytics can improve performance
Elettra Damaggio
Director, Data Science
elettra.damaggio@stonex.com