Revenue growth in the age of connected decisions

Revenue Growth Management (RGM) has evolved from a commercial discipline practiced in spreadsheets to a data-science-intensive capability that sits at the intersection of consumer demand, retailer dynamics, and operational supply constraints. For CPG, FMCG, and retail companies, it represents one of the highest-leverage applications of advanced analytics: small improvements in pricing accuracy or promotional effectiveness translate directly into hundreds of millions of dollars of incremental revenue and margin.

 

Agentic AI changes the premise. These systems pursue objectives such as planning, reasoning across data, and executing multi-step workflows without waiting to be asked at every step, instead of simply responding to prompts. Gartner predicts 40% of enterprise applications will embed task-specific agents by the end of 2026, up from less than 5% today2. Yet Gartner also warns that over 40% of agentic projects will be cancelled by 2027 due to unclear business value and poor governance2. Deployment strategy, not deployment speed, is what separates the companies extracting value from those writing off the investment.

 

This blog details how Sigmoid applies machine learning across the RGM value chain, from foundational price elasticity models to reinforcement learning-powered dynamic pricing engines, and how enterprises can architect an integrated RGM platform that delivers measurable revenue lift.

Why Revenue Growth Management is a strategic priority

The RGM Imperative: In an era of persistent cost inflation, shifting channel mix, and empowered consumers, RGM has become a board-level priority. Companies that deploy advanced analytics across all four RGM levers outperform their category peers by 2-3 percentage points of annual revenue growth.

ML Model Details: Pricing and Promotion Optimization

The table below provides a comprehensive breakdown of the ML models Sigmoid deploys across the RGM value chain, including model types, key input features, outputs, use cases, and implementation libraries.

 

Deep Dive: Price Elasticity Modeling

Price elasticity is the foundation of any RGM analytics stack. The core challenge is not computing a single elasticity coefficient. It is computing reliable, granular elasticities across thousands of SKU-channel-retailer-segment combinations while accounting for confounding effects such as promotions, competitor actions, and macroeconomic shifts.

 

Sigmoid’s Elasticity Modeling Approach

 

• Hierarchical Bayesian Models: Estimate elasticities at the individual SKU level while pooling information across similar products in the category. This is critical for new products with limited price variation history.
• Double Machine Learning (DML): Separates the causal effect of price from confounders using orthogonalization, producing debiased elasticity estimates that hold under competitor response and promotional correlation.
• Cross-Elasticity Matrix: Beyond own-price elasticity, the model estimates cross-elasticities between SKUs to quantify cannibalization and halo effects, essential for portfolio pricing.
• Dynamic Elasticity: Elasticities are re-estimated on a rolling basis (weekly or monthly) to capture consumer behavior shifts due to inflation, private label growth, and channel switching.

Technical Specification: Architecture: Hierarchical Bayesian (PyMC3) + Double ML (EconML) | Features: 40-60 per SKU | Granularity: SKU x Channel x Retailer x Week | Retraining cadence: Monthly with drift detection | Latency: Batch (overnight) with API serving for real-time queries | MAPE: Typically 8-14% on holdout at SKU level

Deep Dive: Promotion Response and Uplift Modeling

Promotional effectiveness modeling is fundamentally a causal inference problem: we want to know the incremental impact of a promotion, net of baseline trends, seasonal effects, and self-selection bias (promoted SKUs are often chosen because they were already trending up).

 

Causal ML for Promotion Measurement

 

• Meta-Learners (T-Learner, X-Learner, R-Learner): Estimate heterogeneous treatment effects, i.e., how promotion uplift varies by store cluster, shopper segment, or time of year. The R-Learner is preferred in settings with strong confounders.
• Difference-in-Differences (DiD): For natural experiments (e.g., test-and-control market rollouts), DiD with synthetic control provides clean incremental measurement.
• Bayesian Structural Time Series (BSTS): Constructs a synthetic counterfactual sales trajectory for promoted SKUs, enabling rigorous pre/post comparison with credible intervals.
• Promotion Decomposition: Splits observed sales lift into baseline growth, display/feature contribution, price reduction contribution, and interaction effects, enabling mechanic-level attribution.

 

Promotion Calendar Optimization

 

Once individual promotion effects are estimated, the optimization layer determines the optimal promotion calendar subject to constraints: retailer category constraints (number of features per quarter, display availability), brand investment thresholds (minimum spend per SKU tier), inventory and supply chain feasibility, and competitive response assumptions. Sigmoid solves this as a Mixed-Integer Linear Program (MILP) using Gurobi or OR-Tools, incorporating the causal uplift estimates as objective function coefficients.

Results Benchmark: Across 8 CPG clients, Sigmoid’s promotion optimization engine delivered an average 18% improvement in promotional ROI and a 22% reduction in trade spend associated with non-incremental volume, compared to prior-year baselines.

RGM Platform Architecture

Sigmoid’s RGM platform is built on a modular, cloud-native architecture that integrates data from multiple commercial and supply chain sources, runs ML models at scale, and surfaces decision recommendations through intuitive front-end tools.

 

Customer Transformation Journey

RGM maturity is a journey, not a single deployment. Sigmoid has worked with over 30 CPG and retail clients across varying stages of RGM maturity. The 5-stage model below reflects the transformation arc we navigate with clients from reactive pricing to autonomous, AI-driven commercial optimization.

 

Representative Client Transformation: Top-5 Global FMCG Company

Client Profile: Global FMCG company with $4.2B in annual revenue across 80+ markets. Initial state: pricing managed in 14 separate Excel workbooks; no causal attribution for promotions; $380M trade spend with unknown ROI; demand forecasting error of 28% MAPE.

 

Sigmoid’s 12-month RGM transformation program:

 

• Phase 1 – Data Unification (Months 1-3): Unified POS, shipment, and retailer portal data onto a cloud data lake. Implemented automated data quality monitoring. Established a single source of truth for commercial KPIs across all markets.
• Phase 2 – Elasticity and Attribution (Months 3-6): Deployed price elasticity models for the top 500 SKUs covering 80% of revenue. Built promotion attribution engine using DML/BSTS; identified $47M in trade spend generating negative incremental ROI.
• Phase 3 – Optimization (Months 6-9): Launched promotion calendar optimizer; deployed mix and assortment engine for retail category reviews; implemented trade spend MMM for annual planning cycle.
• Phase 4 – Dynamic and Real-Time (Months 9-12): Rolled out dynamic pricing for D2C e-commerce channel; implemented real-time price monitoring with competitor feed integration; launched revenue simulator for scenario planning.

Conclusion

Revenue Growth Management represents one of the highest-ROI applications of machine learning in the consumer goods and retail sectors. The combination of causal inference, constrained optimization, and real-time ML models, unified on a modern data platform, enables commercial teams to make better decisions faster, with quantifiable impact on revenue and margin.

 

Sigmoid’s RGM practice brings together deep commercial domain expertise and cutting-edge ML engineering to build platforms that are not just analytically sound but commercially actionable. Our models are designed to integrate into the planning workflows that commercial teams actually use, turning insights into decisions, and decisions into results.