What Data Do You Actually Need?

The minimum viable dataset for elasticity modeling is simpler than most people expect. You need four fields per transaction: a date, a product identifier (SKU number, article code, or similar), the quantity sold, and the unit price. That's it for the baseline.

Optional but valuable additions include: customer identifier or segment, product category or group, promotion flags, and cost data (for margin calculations). Most ERP systems—whether you're on Visma, Fortnox, SAP, or Microsoft Dynamics—can export this data as a CSV or Excel file in minutes.

For the analysis to be statistically reliable, you need at least 12 months of history, ideally 18-24. The more price variation in the data (from annual adjustments, promotional pricing, or customer-specific discounts), the more precise the elasticity estimates will be.

The First 24 Hours: Data Ingestion and Cleaning

Once your data arrives, the first step is automated validation and cleaning. This catches the issues that plague every real-world dataset: duplicate transactions, missing values, currency inconsistencies, outlier prices (that €0.01 test order), and date format mismatches.

The system normalizes everything into a consistent structure, flags anomalies for review, and generates a data quality report. In our experience, about 2-5% of transactions in a typical SME dataset need some form of cleaning. The automated pipeline handles the obvious cases; edge cases get flagged for a quick human review.

By the end of this stage—typically 6-8 hours after data submission—you'll receive a data quality summary showing: total transactions processed, SKUs identified, date range covered, and any data gaps that might affect specific products.

Hours 24-40: The Elasticity Engine

This is where the core analysis happens. For each SKU with sufficient data (typically 30+ transactions with at least 2 distinct price points), the system runs a regression model that estimates how quantity demanded responds to price changes.

The model controls for seasonality (monthly and weekly patterns), trend (is demand for this product growing or declining over time?), and any promotional effects captured in the data. For products with customer segment data, it also estimates segment-level elasticity—because your largest distributor may have very different price sensitivity than a small end-customer.

The output is an elasticity score for each qualifying SKU, along with a confidence interval and a model fit statistic. Products that don't have enough data for reliable estimation are flagged as "insufficient data" rather than given a misleading number.

Hours 40-48: Your First Report

The final deliverable is an interactive dashboard and a downloadable report that includes:

SKU-level elasticity scores — sorted by pricing opportunity (low elasticity = high pricing headroom). Each product gets a clear classification: inelastic, moderate, or elastic.

Price change recommendations — specific suggested increases or decreases for your top-opportunity products, with projected volume impact and revenue impact for each.

Scenario simulator — an interactive tool where you can model "what if we raise prices 5% on these 30 products?" and see the projected P&L impact before making any changes.

Executive summary — a one-page overview suitable for sharing with your CFO or commercial director, highlighting the total margin opportunity identified and the recommended next steps.

From here, the typical workflow is to review the recommendations with your pricing team, select the changes you're comfortable making, implement them, and then track actual results against the model's predictions over the following quarter.