Causality: Turning Predictive Insights into OpEx Savings and EBIT Growth

Causality harnesses Microsoft Azure Machine Learning to forecast future states with remarkable precision, whether modelling macroeconomic conditions or forecasting product demand at scale.

Engineered for Financial Control

We recognise that a forecast is only valuable if it drives a better financial outcome. Causality is built around one core financial metric: the Cost of Forecast Error (CFE).

Beyond Accuracy: Unlike generic tools that only chase statistical accuracy, Causality optimises models to minimise the actual financial cost of being wrong, whether that's the cost of excess inventory or overstaffing (over-forecasting) or the cost of poor customer service, lost sales and expedited freight (under-forecasting).

Actionable Intervention: The solution integrates with your existing data sources, using Microsoft Fabric to streamline storage and data flow, while Azure Machine Learning automates model selection and optimisation. But our true differentiator is identifying causal drivers so your teams can intervene and influence outcomes, not just predict them.

Enterprise Ready, NZ Focused

Results are delivered through clear, intuitive Power BI visualisations. Causality also enables interactive what-if analysis, allowing teams to simulate different scenarios and explore the impact of key variables before making critical decisions. Built on our evidence-based five-step data science methodology, Causality transforms uncertainty into quantifiable and actionable insights.

Available now on the Microsoft Marketplace, Causality offers a scalable solution to elevate your planning. Contact us today to discover how Causality can help you Predict, Plan, and Act with Confidence—whatever your business needs.

Call Us: 64 272 555 000

Mail Us: hello@virtualblue.co.nz

Most executives will tell you that when shaping business plans and strategy, forecasts can serve as a great counterweight to gut feelings and biases. Most will also admit, however, that their forecasts are still notoriously inaccurate.

- McKinsey & Co.

The Predictive Value Framework™

Every forecasting problem sits somewhere in a three-dimensional space. Understanding where determines whether machine learning will help or waste your investment. Our framework is engineered to eliminate the wasted budget often seen in generic AI rollouts.

The Three Dimensions

1. Forecastability: Can it be predicted? Entropy measurement identifies if a time series contains exploitable signal before you invest in modelling. High entropy means buffer with safety stock. Low entropy means ML can add significant value.

2. Causal Leverage: Can we intervene? Pattern recognition tells you what might happen. Causal inference tells you why and whether you can change it. We identify controllable drivers (pricing, promotions, operational levers) so prediction becomes intervention. This is why our solution is called Causality.

3. Cost of Forecast Error: What's at stake? We quantify the financial impact of over-forecasting versus under-forecasting for your specific business then optimise models for your cost function, not just generic accuracy metrics.

Insights to Action in Five Steps: Our proven data science framework delivers predictive modelling solutions through a systematic five-step process, typically implemented over six weeks from initial discovery workshops through to production model deployment on Microsoft Azure.

01. Discovery & Problem Definition

We clarify the commercial question, not just the technical one. What decision will this forecast inform? What's the cost of being wrong?

This is where we define the CFE that shapes everything that follows.

02. Data Preparation and Forecastability Assessment

Before modelling, we perform an Entropy analysis to assess forecastability. This identifies which series have exploitable signal and which are dominated by noise. This prevents wasted effort and sets realistic expectations.

03. Feature Engineering and Causal Analysis

We engineer features that capture causal relationships, not just correlations. Macroeconomic indicators, promotional calendars, and operational constraints are integrated where they genuinely drive outcomes.

04. Model Development and CFE Optimisation

Models are optimised for your cost function, the financial impact of over-forecasting and under-forecasting, not generic accuracy metrics. Azure Machine Learning automates model selection; we ensure the right objective function.

05. Implementation and Ongoing Optimisation

Production deployment on Microsoft Azure with Power BI dashboards. Automated retraining as new data arrives. Ongoing monitoring for model drift to maintain CFE targets.

The Power of Prediction

Forecast Accuracy

50%

Reduced forecast error from machine learning

Increased Sales

7.5%

Increase in sales due to improved stock availability

Reduced Disruption

40%

Reduction in business disruption impacts

Reduced Inventory

75%

Reduction in excess inventory holding

Less Emissions

15%

Decrease in value chain carbon emissions

McKinsey & Co., 2022

Predict, Plan, and Act with Confidence

For Your Executives

STRATEGIC PLANNING

Scenario: Generate data-driven scenarios for strategic planning sessions
Benefit: Make informed decisions based on predictive insights

RISK MANAGEMENT

Scenario: Model and identify potential risks before they impact operations
Benefit: Proactively mitigate risks and protect business value

RESOURCE OPTIMISATION

Scenario: Forecast resource requirements across different business scenarios
Benefit: Optimise allocation of capital and resources

For Your Finance Team

CASH FLOW FORECASTING

Scenario: Apply machine learning to anticipate future cash flow demands
Benefit: Ensure adequate funding and optimise working capital

BUDGET PLANNING

Scenario: Generate accurate revenue and cost predictions
Benefit: Create more reliable budgets and financial plans

PERFORMANCE TRACKING

Scenario: Monitor actual versus predicted performance in real-time
Benefit: Quickly identify and respond to deviations from forecasts

For Your Operations Team

DEMAND FORECASTING

Scenario: Predict future demand patterns across products and services
Benefit: Maximise sales, optimise inventory and resource allocation

PROCESS OPTIMISATION

Scenario: Model and predict operational bottlenecks
Benefit: Proactively address efficiency challenges

CAPACITY PLANNING

Scenario: Forecast resource requirements across different scenarios
Benefit: Ensure optimal staffing and capital resource levels

Causality is the key to prediction; randomness is just a symptom of missing causes.

Judea Pearl

The Science of Prediction

Predictive modelling is concerned with finding a function that optimally maps input data to a given output with the goal of making accurate predictions — this principle underpins time series forecasting, where historical data is analysed to identify patterns and predict future outcomes.

At its core lies the concept of the data-generating process (DGP), which refers to the underlying mechanism that produces the observed data. By understanding the DGP, businesses can develop more accurate models to leverage trends, seasonal behaviours, and other time-related insights. From planning inventory to anticipating economic shifts, this technique provides a structured framework for addressing uncertainty in dynamic environments (Hyndman & Athanasopoulos, 2021).

Methodology

Our forecasting methodology integrates both machine learning models and statistical methods to extract meaningful insights from time series data. Statistical methods, such as ARIMA and exponential smoothing, are particularly effective for identifying trends, seasonality, and random fluctuations, allowing us to build robust mathematical models. Machine learning models, on the other hand, excel at capturing complex, non-linear relationships in the data. By combining these approaches, we develop flexible predictive models that accurately project temporal patterns into the future (Makridakis et al., 2020).

This foundation is further enhanced by advanced feature engineering, incorporating external factors such as macroeconomic indicators (e.g., GDP, interest rates, CPI), social media sentiment and themes, competitor actions, customer demographics, and proprietary datasets. These external influences refine the models by accounting for variables beyond the historical data, further improving forecast accuracy (Petropoulos et al., 2022).

Ensemble Methods

Building on these principles, we leverage ensemble forecasting techniques to develop robust predictive models. Ensemble methods combine multiple forecasts to enhance accuracy and reliability by utilising the strengths of individual approaches (Oliveira, 2015).

This multi-faceted methodology enables machine learning to automatically identify complex relationships between time series behaviour and exogenous variables (e.g. macroeconomic indicators), maintaining robustness against market noise while ensuring precise estimation of both short-term fluctuations and long-term trends.

Conformal Prediction

To ensure reliability, we employ conformal prediction, a method that provides dynamic confidence intervals, ensuring that forecast ranges align with specified confidence levels. For example, in demand forecasting, conformal prediction not only highlights the most likely outcomes but also offers a clear range of possibilities, empowering decision-makers with actionable insights (Vovk, Gammerman, & Shafer, 2022).

Model Validation

Ensuring the robustness and generalisability of our models is central to our methodology. To achieve this, we rigorously evaluate performance against holdout datasets, which simulate real-world conditions by testing the models on unseen data. This validation step helps us prevent model overfitting and ensures the reliability of our forecasts when applied to practical scenarios (Hyndman & Athanasopoulos, 2021; Petropoulos et al., 2022).

By combining statistical rigor, machine learning innovation, ensemble forecasting, and conformal prediction, our models transform uncertainty into quantifiable probabilities, enabling you to Predict, Plan, and Act with confidence.

References

Hyndman, R. J., & Athanasopoulos, G. (2021). Forecasting: Principles and practice (3rd ed.). OTexts.

Makridakis, S., Spiliotis, E., & Assimakopoulos, V. (2020). The M5 competition: Results, findings, and conclusions. International Journal of Forecasting, 36(1), 54-74.

Oliveira, M. (2015). Ensembles for time series forecasting. JMLR: Workshop and Conference Proceedings, 39, 360–370.

Petropoulos, F., Apiletti, D., Assimakopoulos, V., Babai, M. Z., Barrow, D. K., & others. (2022). Forecasting: Theory and practice. International Journal of Forecasting, 38(3), 705-871.

Vovk, V., Gammerman, A., & Shafer, G. (2022). Algorithmic learning in a random world (2nd ed.). Springer.