Solutions/Dynamic Floor Pricing Real-time floor optimization

Floors that adapt in real time.

Floors solved per segment in real time, across every connected demand source. Every change carries its reasoning. Pricing stops being a spreadsheet exercise.

The problem

A static floor leaves revenue on the table.

Demand moves constantly, and differently by geography, device, and audience. A fixed floor prices an average that never occurs: it caps revenue when demand spikes and loses fill when it softens.

01 Cadence

Priced for an average

Review cycles run on a human calendar. Bid landscapes do not. By the time a floor is revised, it describes a market that has moved on.

02 Granularity

One floor, many markets

One site-wide floor averages segments with very different willingness to pay. Revenue is left on the table at the top, fill is lost at the bottom.

03 Feedback

No record of why

When a floor change moves revenue, the reasoning lives in someone's memory. Nothing to audit, reverse, or learn from.

Real-time repricing

Price the segment, not the site.

Every segment gets its own floor, solved against a reconstructed view of what demand was actually willing to pay, across every connected source at once.

Method / 01

Bid landscape reconstruction

Win and loss signals build a demand curve per segment, so the revenue effect of a floor is estimated before it is applied.

  • Win-rate curve per segment
  • Bid density estimation
  • Counterfactual revenue modelling
Method / 02

Segment-level curves

Geography, device, placement, and context each carry their own floor curve. A high-value context is never priced down to match a low-value one.

  • Geo & device dimensions
  • Contextual segment inputs
  • Per-placement overrides
Method / 03

Real-time reprice cadence

Floors are re-solved continuously and written back through each SSP's own API. The price in market tracks demand instead of trailing it.

  • Continuous solve and write-back
  • Daypart-aware adjustment
  • Native SSP floor updates
Explainable decisions

Every change carries its reasoning.

Every write records what it saw and what it expected. That is what lets the loop be scored, tuned, and reversed — and what keeps it moving toward maximum monetization instead of drifting.

Trigger What opened the decision: a win-rate shift, a bid-density change, a pacing signal, with the window it was measured over.
Alternatives considered Every candidate floor evaluated, with its modelled revenue and fill. The chosen value can be compared against what was rejected.
Expected effect Predicted change in clearing price, fill, and revenue, recorded before the write and scored against what actually happened.
Prior state The floor being replaced, kept for one-step reversion without recomputing the decision.
Policy check The bounds the value was tested against, and whether it executed automatically or queued for review.
Operating model

Zero added headcount.

Not a better dashboard for an analyst. The manual reprice loop goes away; the operator keeps control of the bounds it runs inside.

DimensionModelDetail
Reprice cadence Real time versus a periodic or ad-hoc manual review
Decision granularity Per segment geo, device, placement, and contextual dimensions
Analyst time required None bounds are configured once; the loop runs unattended
Reversibility Full prior floor retained on every write for one-step rollback

Intended operating characteristics. Realized cadence and granularity depend on SSP write limits and the segment dimensions available in your inventory.

Where it sits

One input to the wider monetization stack.

Floor pricing reads the same joined data and writes through the same execution path as everything else. No parallel system with its own version of the truth.

Reads from The joined data set built by AI-Native Data Intelligence. A pricing decision and a revenue forecast work from the same numbers.
Prices against Segments defined in AI-Native Data Intelligence. A high-engagement audience carries its own floor instead of being averaged into run-of-site.
Writes to Connected SSPs through their native APIs, with per-platform rate limits and write semantics handled by the connector.
Shares Guardrails, audit trail, and kill switch shared with every other automated action. One place to set bounds, one place to review what ran.
Contact us

Tell us about your floors.

Send a work email. We will get in touch about one segment: what it is priced at now, and what the engine would price it at.

Dynamic Floor Pricing

Start with one inventory group.

Most publishers start in observe-only mode on one segment, comparing what the engine would have priced against what the manual floor did.