Self-optimizing pricing engine lifted gross margin 12 points in 6 months
A scale-up SaaS company replaced their manual pricing review process with an autonomous pricing platform — dynamic signals, real-time competitor inputs, and a feedback loop that improves with every deal closed.
Context
A B2B SaaS company at Series C was running pricing on a quarterly review cycle. A small committee set list prices and discount bands; sales reps exercised wide discretion within those bands and most defaulted to the bottom. By the time the committee reconvened, the market had moved, competitors had repriced, and the discounting patterns from the prior quarter were already stale — and margin was leaking in the gap.
The approach
The competitive intelligence layer pulls pricing signals continuously — competitor public pricing pages, review site deal data, product positioning changes — and feeds a streaming pipeline refreshed multiple times per day. The engine knows what the market looks like now, not last Tuesday. That freshness is what made real-time recommendations possible; a batch-updated competitive feed would have been describing a market that no longer existed.
The pricing model generates a recommended price and discount ceiling for each deal based on four inputs: competitive position in the specific segment, the buyer's historical price sensitivity based on comparable deals, current pipeline velocity, and margin floor guardrails encoded as policy rather than left to rep discretion. Sales reps see a recommended price with a confidence band and a brief rationale. Overrides are permitted and logged — the log is part of the feedback mechanism.
The feedback loop is what makes the engine self-optimizing. Every closed deal — won or lost — flows back into the model: the price offered, the competitor the deal was lost to where known, rep notes on price objections, and the margin at close. The model recalibrates weekly. By month three, win rate at recommended price had increased enough that discount override rates fell without a policy mandate. The incentive shifted because the recommendation got better.
Results
Gross margin up 12.3 points over 6 months
Discount override rate: 61% → 29%
Win rate at recommended price: 67% (up from 44%)
Time from competitive price change to engine recalibration: under 4 hours
Pricing committee cadence: quarterly reviews → exception-only
What made it work
The feedback loop was instrumented before the pricing model was deployed. Most pricing automation projects treat deal outcome data as a reporting concern — captured in the CRM, reviewed in quarterly analysis. This project treated it as a training signal from day one. Connecting closed-won and closed-lost outcomes directly to the model's weekly recalibration cycle, with structured competitive context attached, was the architectural decision that converted a pricing tool into a pricing engine. Without it, the system would have been accurate at launch and increasingly stale every month afterward.
More outcomes
Other engagements
48% reduction in procurement cycle time across 14 categories
A North American manufacturing firm deployed custom procurement agents across invoice processing, approval routing, and exception handling — eliminating the manual overhead that was adding 3–5 days to every purchase cycle.
AI onboarding agent: 3-day → 4-hour time-to-productivity for new hires
A financial services firm deployed an onboarding orchestration agent that coordinates HRIS, IT provisioning, and compliance training — reducing the administrative burden on HR and getting new employees productive the same day.
Built something similar?
Let's talk.
We work with a select number of organizations at a time. If this outcome resonates with what you're trying to build, we want to hear from you.