The full revenue management cycle — monitor, diagnose, simulate, optimize, recommend, measure, and learn — automated with machine learning. Here it is, module by module.
Most pricing and promotion decisions are made on intuition, or justified with numbers no one can verify. The number that actually matters — how much extra you sold thanks to the promotion — is invisible: it requires knowing what would have happened without it.
For every SKU we reconstruct the baseline — what it would have sold without intervention — and compare it against actual sales. That's where the three effects that matter come from.
The extra sales during the promotion, above the baseline.
The dip afterward, once the customer has already stocked up during the promo.
The sales you took from another product in your own portfolio.
Every stage runs on machine learning and shares the same financial logic. Each period's result feeds the next.
Sales, units, price, cost, and promotions, against last year or last month. The macro picture, ready to drill into.
Is the drop about volume or price? The system drills from category down to brand, product, and the exact mechanic.
Project a promo before launching it, fatigue curve and cannibalization included. The same logic as the postmortem.
The optimization engine builds the combination that maximizes your goal — volume, revenue, or margin — under your constraints.
The specific mechanic that works best, with its confidence level and the reasoning behind it.
The causal postmortem on what ran: real incremental, net of pull-forward and cannibalization.
We compare the prediction against reality and the model retrains. The more you use it, the more data it has and the better it predicts. ↻ Every cycle, more precise.
The full macro view: how total sales are trending against last year and last month, whether units, price, or costs went up or down, and which categories, brands, or mechanics explain the change.
Reconstructs the counterfactual baseline by SKU, store, and period, and measures the real incremental effect of every promo — net of pull-forward and cannibalization — with its confidence level.
From the aggregate down to the root cause: category → brand → product → mechanic. Finds exactly what stopped working, not just that something's off.
Projects the outcome of a promo before you run it, with the same financial logic as the postmortem, including promo fatigue and cannibalization.
The optimization engine builds the optimal grid for your goal under real-world constraints: budget, minimum margin, exclusions, price indexing, and psychological rounding.
Not just “lower the price”: the specific mechanic that fits each SKU — discount, 2-for-1, bundle — with its confidence interval and the reasoning right there.
A red number on the dashboard is the start, not the answer. Here's how Revux drills from the aggregate down to a concrete decision.
There's no one-size-fits-all recipe. We train several models — time series, causal models, gradient boosting, regression — and for each product and category the system automatically selects whichever predicts best, measured over real periods with walk-forward validation. Proven accuracy, case by case. And every recommendation stays reproducible and traceable: you know which model generated it, which variables mattered, and what limited it.
Enterprise-grade security from day one. Your data is never mixed with anyone else's, never sold, never shared.
Your data travels and is stored encrypted, in transit and at rest.
Every client lives in its own space. Your data never crosses paths with anyone else's.
It's never sold or shared. Export or delete it whenever you want.
Aligned with Chile's data privacy and protection law (Ley 21.719).
Revux doesn't ship with a fixed catalog: it works with the mechanics your business already uses, on your own data. These are a few examples — the system adapts to whatever you have.
We'll show you, on your own data, how much incremental was left on the table.
Revux