en

ADNOC: From Data Silos to Customer Intelligence

See how ADNOC uses data-driven solutions — including Dataiku — to more accurately predict fuel consumption & drive 11x ROI by hyper-personalizing marketing campaigns.

85%

Improvement in inventory loss

Millions

Incremental gross profit from ADNOC Distribution’s customer engagement use case

50%

Reduction in time to market for data solutions

 

ADNOC — Abu Dhabi National Oil Company — is the fastest-growing brand in the Middle East and the first United Arab Emirates (UAE) brand to surpass $10 billion in value. As a diversified group of energy and petrochemical companies and a major contributor to the GDP of the UAE, it’s no surprise that the company aims to infuse efficiency and accuracy in every project they take on. 

How does that carry over to analytics and AI? A significant portion of this work involves optimizing customer demand at ADNOC’s vast network of over 500 service stations across the Middle East. The challenge was twofold: 

  1. Ensuring a consistent supply of all fuel grades at the stations, and 
  2. Untapping new ways to energize and enhance every customer journey at the stations to send the right offer(s) at the right time and on the right channel(s).

Advanced Fuel Demand Prediction With Dataiku

Ensuring a consistent supply of three different fuel grades across such a large network of service stations is no easy feat. The AI and analytics team was previously working with a basic model averaging forecast for predicting fuel demand, which lacked accuracy. 

This inaccuracy led to a loss of business opportunities, as competitors were able to capitalize on the gaps in ADNOC’s fuel supply chain. Further, it had a negative impact on the company’s reputation, highlighting the need for a more efficient and accurate solution. As a result, ADNOC gathered a small team of data scientists, engineers, and business leaders from retail operations, logistics, and finance to build a new predicting data workflow. 

This is where Dataiku came into the picture. The ADNOC team enjoyed Dataiku’s quick learning curve, which enabled their data scientists to be productive in just a few days without long or tedious product training. They also favored Dataiku’s visual data flow that is easy to understand, from syncing to the data and preparing it to developing the models and generating new outputs. 

Further, Dataiku’s explainable AI features enabled the data scientists and other stakeholders to understand the factors that influenced model predictions through interactive reports for feature importance, partial dependence plots, and individual prediction explanations. With automated model and project documentation, the team was able to save time on maintaining docs while providing consistent records of the project for compliance.

ADNOC found Dataiku to be the leader in developing fast and efficient AI solutions. Awad Ali Head of AI & Analytics, ADNOC

3 Models for Maximum Fuel Demand Impact

As a massive improvement from their previous average-based model, the new fuel demand prediction system includes three separate models:

  • The hourly prediction model runs every hour to predict seven days ahead, providing visibility to the operations team for planning the following week. Seasonality plays a key role in this model, as it accounts for low consumption during the night and high consumption during the day. This ensures that trucks deliver the right quantity of fuel.
  • The daily prediction model runs monthly to predict 45 days ahead. Its purpose is to plan the retail workforce distribution in each station for each shift during the next month. This model optimizes resource allocation for the 10,000 staff members of the retail operations unit based on real demand.
  • The monthly prediction model provides an 18-month sales volume forecast for fuel, enabling finance and retail planning management to forecast profit and loss for the year ahead across stations and cities. This helps them to allocate budget and resources across departments more efficiently.

As a result of this three-pronged approach, the entire fuel delivery system is now automated using a command center. A master dashboard is updated in real time to display traffic lights to indicate the status of each station (i.e., green signifies that the station is operating perfectly, yellow indicates that it may run out of fuel within the next eight hours, red warns that the station has only three hours left before running out of fuel, and black means that the station has already run out of fuel).

In addition to the market-leading command center, the fuel delivery process itself has been automated, ensuring that fuel is delivered from the nearest fulfillment center — streamlining the delivery process and reducing the wait times for fuel replenishment. On top of the automation, what did the new prediction system mean for the enterprise?

  1. Tens of Millions in Revenue Gain: The newly built models reduced inventory loss from 0.8% to 0.12%, the scale of which translates to tens of millions in revenue gain. This improvement is tracked on an automated dashboard, which is critical for monitoring the reduction in business opportunity loss.
  2. Operation and Retail Staff Loyalty Improvement: Staff optimization has been implemented across all service stations, involving 10,000 staff members. Previously, it was a real challenge to move staff dynamically without the aid of scientific data and models. Now, staff loyalty and happiness have increased as they benefit from more accurate planning and no longer need to manage deceived customer expectations.
  3. Stronger Customer Experiences: The customer experience improved globally, as customers are almost guaranteed to find all fuel grades across each station anytime they need it. This resulted in better interactions with retail staff and increased customer loyalty. 
  4. More Accurate Financial Planning: Budgeting has also seen significant advancements, with more accurate forecasts for target and future sales for each division in the retail sector. This has allowed for stronger financial planning and resource allocation across departments.

When the ADNOC team was evaluating a vendor for this project, we mentioned above the crucial role that Dataiku’s advanced explainable AI capabilities played, along with the low barrier to entry and visual data flow. Beyond that, Dataiku enabled a significant reduction — 50% — in time to market for data solutions.

With traditional methods, development time may have taken more than eight months. The ease of using Dataiku, combined with its comprehensive feature set, enabled us to cut development and productionalization time in half. Awad Ali Head of AI & Analytics, ADNOC

From Challenge to Customization: A Marketing Department Overhaul

Dataiku has helped ADNOC drive not only improvements in their core business, fuel distribution, but also in their marketing department. ADNOC Distribution — the UAE’s leading mobility retailer with a network of more than 570 service stations and 350 convenience stores — has a goal of being the global mobility partner of choice, an enabler of sustainable mobility, and a provider of exceptional customer service (i.e., through smartly generated personalized promotions, offers, and upsell/cross-sell opportunities). However, they faced many challenges pertaining to technological solutions and platforms that prevented them from driving efficiency and accurate to achieve their objectives:

  • Customer data was decentralized on many sub-systems and managed in silos.
  • Lack of a 360-degree customer view by cluster/segments/region to understand revenue patterns, demographics, customers categories (i.e., VIPs, EV charging customers, caffeine addicted, digital natives, etc.)
  • Unavailability of data science solution to perform day-to-day analysis and build machine learning (ML) model(s).
  • Risk of losing competitive advantage in the future due to the absence of key advanced technologies, such as AI and advanced ML, a personalization engine, etc.
  • Lack of sharp data insights, which led to a low ROI on marketing investments.
  • Low customer engagement due to absence of intelligent, relevant, and sharp customer targeting.

The team in ADNOC Distribution’s marketing department adopted a phase-wise structured approach on their data analytics and AI journey. The key milestones included:

Creation of a Single Source of Customer Information by Building a Customer Data-Mart

ADNOC Distribution team created the customer data-mart, which includes more than 300 customer attributes for over 1.7 million loyalty customers of ADNOC Distribution. Some examples of customer attributes are:

  • Calculations-Based: Avg. monthly spend, monthly station visits, etc.
  • Data Mining-Based: Primary station, weekday/weekend preference, etc.
  • ML-Based: Customer segmentation, propensity to purchase, etc.

Data analysts use the customer data-mart to provide customer trends to support business decisioning. Marketing analysts use the data-mart for dynamic segment creation for marketing campaigns. Data scientists use the customer data-mart variable as input features in ML models.

customer data-mart from ADNOC

Building ML Models to Meet Customer Engagement and Experience Vision

ADNOC Distribution developed customer segments based on customer historical purchase behavior, demographics, and loyalty enrollment and digital channels activity data. The team leveraged Dataiku’s drag-and-drop recipe UI and SQL coding interface to come up with segmentation schemes. The team utilized sampling, hierarchical clustering, and K-nearest neighbor ML techniques to perform segmentation.

The segmentation schemes are refreshed every six months and segmentation mapping against customers is pushed to the customer data-mart. Customer segments are used to run segmented marketing campaigns against specific business objective(s), drive segment-specific desired behavior changes, perform SKU planning across stations, etc.

ML models in Dataiku with ADNOC

Execution and Operationalization of ML Models

ADNOC Distribution understands the importance of consuming ML model outputs for customer-related decisions. The team has performed necessary integrations between different technology solutions in achieving real-time consumption of model outputs. The model scores along with business rules are used to deliver hyper-personalization and Next Best Action / Offer (NBA/NBO). 

Integration of ML Models (From Dataiku) With Their Marketing Technology Platform to Deliver Personalization and NBA/NBO

Hyper-personalization and NBO/NBA delivery requires a combined data science and marketing CRM solution. ADNOC Distribution has built a series of ML models that help decide the right product, offer, or communication to send customers across various stages of their customer journey.

Gross Profit Driven, Improved Campaign ROI, & Data Scientist Productivity Boosted

All in all, these marketing campaign optimization initiatives aimed to increase ADNOC Distribution’s customer base, generate more revenue through customer engagement and cross-sell/upsell and repeat business, and ensure customer stickiness to their brand and services. They successfully did all of that and more:

  • Incremental gross profit: In the first half of 2023 alone, the project delivered an incremental gross profit of tens of millions USD. ADNOC Distribution’s F&B business has seen significantly impressive category growth, 40%+ conversions from core fuel business to non-fuel business, and a 30% increase in customer participation on targeted promotions.
  • Improved marketing campaign ROI: ADNOC Distribution has recorded its highest-ever ROI from a marketing campaign in 2023 as a result of these initiatives. Data-led promotions have delivered up to 11x marketing promotion ROI in recent quarters. The campaign response rates improved by more than 30% (basis prior to solution launch).
  • Growth of loyal customer’s spend value: Average spend per customer is up by 18% — one of the important factors has been the precise customer targeting based on the data modeling.
  • Better customer insights: The project has provided ADNOC Distribution with better understanding of its consumer profiles, their mobility needs, shopping mission, etc. The team is able to identify opportunities at a micro-segment level and take targeted actions.
  • Strong data support to organization initiatives: Everyday AI and analysis capabilities enable business stakeholders to have visibility around customer trends, analyze different scenarios that pertain to a business case, and drive better decision-making.
  • Speed of decisioning: The team saw a significant improvement in the turnaround time on descriptive analysis and hypothesis testing to support new business cases — including up to 40% lesser time taken on complex analytics.

When discussing those who were involved in the marketing initiatives at large, Dataiku had a positive outcome on both technical and non-technical experts. First of all, data scientists experienced productivity improvements due to quicker ML outputs based on the ability to identify best-fit algorithms with Dataiku’s UI-based ML features (in addition to hyperparameter selection, feature engineering, and results interpretation with statistics and charts). On the flipside, Dataiku’s GUI-based analysis features enabled non-technical users (without SQL or Python skills) to perform day-to-day analysis within the platform. 

Next, Dataiku enabled the team to seamlessly consume data from multiple data sources and perform concurrent analysis, thus saving valuable time and data preparation efforts. Finally, the team integrated Dataiku and the Salesforce CRM personalization engine, which ensured consumption of ML model scores in real time to deliver hyper-personalization and improve campaign ROI.

SLB: Putting Data & AI to Work for Energy

SLB partners with Dataiku to drive improvements and save millions of dollars through the use of data and AI across the business.

Read more

Vestas: Propelling Sustainable Energy Solutions With Dataiku

Though the savings generated by the express shipping recommendation model will only fully materialize over time, the tool when globally implemented is estimated to reduce express shipment costs by 11-36%.

Learn More

Ørsted: Monitoring Market Dynamics With LLM-Driven News Digest

Ørsted uses Generative AI to ensure its executive management has a more aligned understanding of market dynamics, for a 100% time savings over a manual approach.

Learn More

SLB People Analytics: Optimizing the End-to-End Talent Lifecycle With Dataiku

SLB's People Analytics team uses Dataiku to better equip its talent management teams globally and improve talent retention.

Learn More

Showroomprivé: Putting ML-Powered Targeting in the Hands of Marketers

Showroomprivé leverages Dataiku to innovate across their business, including for machine learning-based targeting to build marketing campaigns that are 2.5x more effective.

Learn More