To build & deploy a churn model (vs. 14 months!)
In months with the churn prediction and revenue assurance models in Dataiku
From fiber and transport to network connectivity and managed services, Zayo designs and builds solutions to help organizations connect their business — from edge to core to cloud. Over the last year, Chief Data Officer (CDO) David Sedlock has set out to overhaul the company’s data strategy to generate recurring business value for the company, faster — and Dataiku has been a catalyst for that success.
Upon joining Zayo, Sedlock knew there were changes to be made and pockets of operational efficiency to untap. The focus was creating standardized data ingestion and governance standards, data libraries for easy reuse, and business-ready datasets — optimizing core foundational data capabilities.
The data science team focused on “data quick strikes,” one- to four-week activities intended to leverage known data to impart a rapid impact. The end-to-end Dataiku platform enabled the team to build and deploy models quickly, delivering 11 data quick strikes in 2023.
The team is also focused on supporting large programs (i.e., integrated network inventory, service order management), a data product catalog with controls and data quality, and a data maturity index and model to measure how to add maturity to the data across their data analytics ecosystem.
Zayo’s first use case with Dataiku was churn prediction. Previous attempts to build a churn prediction model took 14 months. After creating and deploying the model, a full-time employee (FTE) was required to maintain the production model.
The ability and usability of Dataiku is great because you don’t need to be an expert. You have to be data cognizant, but it’s not like other tools and platforms where if you aren’t trained for three weeks, you can’t figure it out.David Sedlock CDO, Zayo
With Dataiku, the team built and deployed a churn prediction model in mere weeks, driving speed, agility, and faster time to value for the data science team. The team incorporated data from their acquired companies to make the model as robust as possible and leveraged causal prediction and uplift modeling in Dataiku.
By leveraging Dataiku’s automated MLOps capabilities, what used to take them an FTE to maintain now takes minutes a week of model oversight and maintenance. This frees them up to work on more use cases that will impact the business. With the churn model in Dataiku, the Zayo team realized millions in savings in 14 months.
A second use case enabled the Zayo team to again realize millions of previously unrealized revenue. They pulled in and scanned contracts and used data points on the contracts to correlate between the service orders, contracts, and billing.
For this use case, the team leveraged Dataiku’s text extraction and OCR plugin, reinforcing Dataiku’s extensibility and the ability to use pre-built plugins and solutions from the robust developer ecosystem.
A third use case was also a massive win, especially given Zayo’s complex architecture. For years, Zayo used Salesforce as its CRM and for service delivery, network monitoring, billing, implementation, etc., resulting in a whopping 66 open text fields inside Salesforce. With machine learning (ML) modeling in Dataiku, the data science team at Zayo created a capability to crawl the open text fields to pull the data back into Dataiku.
When Zayo received orders, it needed a scalable way to provide accurate predictions of when a given customer’s service would start. The data science team decided to analyze service orders 15 years old and younger, combining different data elements to create a dashboard for order delivery validation. Using the historical data analysis, Zayo can predict when service can start for their customers and what needs prioritization to realize revenue earlier.
Had the team not been able to automate their model maintenance with Dataiku, they wouldn’t have been able to even consider working on these use cases. Today, they actively work through their backlog of use cases and work with the business on what’s most important.
The economics of using Dataiku are exponential — both for scaling and having the power in our own hands. There’s no ‘this is the way we do it or else’ methodology.David Sedlock CDO Zayo
The data science team at Zayo is extremely intentional about embedding data science and ML in the business, pointing teams to centralized datasets and Dataiku as a centralized platform. They are focused on democratizing and extending analytics and AI capabilities to the company using common and set frameworks, models, and datasets that everyone can use so that when projects are running, every team can align back to the same numbers.
The team is keen to enable the business to achieve its operational metrics and be more successful. They prioritize projects that deliver cost benefits to the company and improve data projects’ direct correlations to business KPIs. They are optimizing data foundations and working on advanced ML and AI capabilities and, once those standards are established, plan to experiment with Generative AI.
Eighty or 90% of the time people shouldn’t know Everyday AI is happening, it should be that embedded in the business.David Sedlock CDO, Zayo
The issue of service downtime and failure (which can lead to significant revenue loss) is specifically prevalent amongst telcos, but this problem can be solved by using machine learning to predict what might cause service failures and prevent them from happening.
Read moreOrange has accelerated their data science practice with Dataiku, building a call load detection and triage model in just one month, with more use cases on the horizon.
Learn MoreHear how Engie's IT team partners with Capgemini and Dataiku to support business teams across the company on their data journeys.
Learn MoreBouygues Télécom uses Dataiku to mitigate risk by building, validating, and retraining their models.
Learn MoreSLB partners with Dataiku to drive improvements and save millions of dollars through the use of data and AI across the business.
Learn More
