Data Science: Custom predictive modelling

A data scientist takes raw, mostly unstructured, messy data and turns it into valuable insight. But Data Science is more than applying the scientific method to discover what’s hidden in an organization’s information. Acting upon that insight and bringing a predictive model into production is the business goal.

For example, a superior recommender model may prove the viability of automated recommendations, but a recommender system in production is what makes an organization money. Our data scientists are business minded, communicative people, with scientific backgrounds. They’re also fluent in programming languages and excel at working with database systems and Big Data technology.

Predictive models

Our data scientists are not one-trick ponies. They’re capable of building predictive models on top of any data set. Our past models include:

Recommendations and online personalisation

It doesn’t matter whether you’re an online retailer, private bank, airline or brick-and-mortar shop with an in-shop mobile site. People visit your site because they’re looking for something. If they fail to find anything they need or like, they’ll leave.

Recommendation systems have one objective: present potential customers with products or services they need and like. Real world recommender systems take millions or billions of signals into account and make complex trade-offs between strategies and product selections.

We have experience building multiple, custom recommender systems, both in the physical world (based on point-of-sale transaction data) and in e-commerce (using clickstream and product catalog data).

Computer vision

Do you want one of those apps where a user takes a picture of something and the app then looks up all similar products in your catalog based on the image? We have the technology to build one that works. Ever wondered about correlations between your customers' favorite color profiles and their geography? We’ll analyze your product photography in combination with clickstream data and purchases to tell you what those correlations are. We’ve created an image based search for complex product catalogs and worked with many other computer vision models on product photography and more.

On-site search optimisation

The days when an on-site search was only a glorified point of access to a product database are gone. Does your search engine learn about synonyms automatically from user behavior? Does your search engine automatically correlate search queries with the most likely candidate products? Does your modeling autocompletion automatically change with user behavior? We’ve worked on search implementations that could do all these and more. Improving search through optimizations takes your users from landing to spending more quickly.

Financial transaction analytics

In large organizations and financial institutions, the flow of money tells its own story. When you analyze financial transactions individually (as opposed to in aggregate, which is accounting), you discover patterns that may point to new business opportunities, fraud, and/or cash flow issues. We’ve worked with multiple financial institutions and large corporations to reap information and find hidden opportunities in financial transaction data through data mining and automated discovery.

Text mining

What are the most common phrases and terms in e-mails from customers who are about to cancel their accounts? How common is a complaint? By processing large bodies of unstructured text data using text mining and natural language processing techniques, we gain these and other interesting insights for businesses. We’ve worked on automatically recognizing patterns and categorizing text content in massive, multi-terabyte contract and CRM data sets.

Ready to challenge us? Send us an (anonymized) sample of your data sets along with a description of the business opportunities you hope to gain. We’ll give you our initial assessment of feasibility, free of charge.

Technology

We combine proven technologies (like R Studio, Python, Pandas, Spark, among others) with the latest innovations to create custom predictive models. And if for any reason, an organization cannot monitor and optimize their predictive models, we’ve got a solution for that, too. Our Data Science Box, is a product-as-a-service that runs and monitors the performance of a predictive model.


“Thanks to GoDataDriven we automatically produce and provide the right quantity of bread tuned to the predicted shop demand.

Hein Boersma
ICT Manager, Bakkersland