Dynamic Pricing Models and Price Optimization: Types, Methods, and How to Choose

Choosing the wrong dynamic pricing model means a rule that undercuts your margin or a price that never moves when demand does. Most sellers hit one of the two before they work out which signal should be driving their prices.

The phrase covers a lot of ground. A competitor-matching rule, an inventory trigger, and a demand-forecasting engine are all called dynamic pricing, and they fail in different ways. Picking among them is a decision about what your catalog needs from price: protect margin, win the Buy Box, clear stock, or find the point where demand and profit balance. This guide separates the models, explains how price optimization works on top of them, and gives a way to choose. For the broader framework on rolling pricing out across a business, see our guide to implementing a dynamic pricing strategy; for scenario-level walk-throughs, see dynamic pricing examples.

Quick answer: match the pricing signal to the job, then set the limits

• The distinction: a dynamic pricing model is the signal that moves a price (cost, competitor prices, demand, or inventory). Price optimization is the method that decides how far to move it, usually by maximizing profit or revenue inside a minimum and a maximum.
• What each model buys you: cost-plus protects margin, competitor-based pricing defends the Buy Box, demand-based pricing captures willingness to pay, and inventory-based pricing paces sell-through.
• Most catalogs need more than one, run by rules, by an algorithm, or both, with a hard price floor under everything.
• How to choose: by margin structure, competitive pressure, inventory risk, and how much of the logic you want to see and control.
• The upside and the risk: margin and sell-through a static price leaves on the table, against a floor set wrong, which turns automated pricing into an automated loss.

A pricing model picks the signal, and price optimization picks the size of the move

The two terms get used interchangeably in search results, and that is why the topic feels muddy. Keeping them separate makes every later decision easier.

The model answers "what should move my price?"

The model is the input. Your landed cost, a competitor's offer, your sales velocity, and your days of inventory cover are all candidate inputs. Each one produces a different behavior, and none of them is correct in general. It depends on what the product is doing in your catalog.

Optimization answers "how far, and toward what goal?"

Optimization is the decision layer. Given a signal, it picks a price that best serves an objective such as profit per unit, total contribution margin, or sales velocity, while respecting constraints such as a floor, a ceiling, and a maximum change per day. Dynamic price optimization is the practice of doing this repeatedly as the signals change, not once per quarter.

Four signal-based models cover most marketplace pricing

Cost-plus pricing protects margin and ignores the market

Cost-plus adds a fixed markup or margin target to cost. On marketplaces, cost has to include fulfillment fees, referral fees, and storage, or the margin you think you have is not the margin you earn. The upside is predictability. The downside is that cost-plus has no view of what buyers or competitors are doing, so it leaves money on the table when demand is strong and stalls when a competitor undercuts you. It works best as a floor calculator, not as the whole model.

Competitor-based pricing defends the Buy Box and needs a floor

Competitor-based pricing moves your price in response to other sellers' offers: match the lowest, stay a set amount above or below, or price relative to the Buy Box winner. On Amazon, where the Buy Box largely determines which seller gets the order, this is the most common model. Its failure mode is a race to the bottom. Two sellers running match-the-lowest rules with no floor will drive each other below cost. Every competitor-based rule needs a minimum price tied to your margin, and a rule for what to do when the competitor is out of stock or is a marketplace seller with different fulfillment.

Demand-based pricing follows velocity, timing, and willingness to pay

Demand-based models raise price when sales velocity or traffic climbs and lower it when demand softens. Time-based variants schedule changes around known peaks, such as seasonal periods or promotional events. The strength is that it captures value cost-plus misses. The weakness is signal quality: on a marketplace, velocity is distorted by ad spend, promotions, and Buy Box ownership, so a sales spike may reflect a campaign, not real pricing power. Read demand signals alongside what else changed that week.

Inventory-based pricing paces sell-through against cover

Inventory-based models tie price to days of cover. Thin cover before a restock calls for a higher price to slow sell-through; heavy cover calls for a lower one to free cash and storage. This model is the one most sellers underuse, because the trigger is easy to define and the cost of missing it, a stockout on a top SKU or aged inventory fees, is concrete.

Rule-based and algorithmic pricing differ in who writes the logic

Any of the four models can run as rules or as an algorithm. The difference is whether a person writes the condition or a system searches for the price.

Rule-based pricing executes conditions you wrote

A rule reads like a sentence: if the competitor's price drops, match it down to my floor; if days of cover fall below a threshold, raise price by a set step. The operator can read every rule and predict its behavior. The cost is maintenance. Rules encode today's assumptions, and they conflict with each other unless someone owns the order of precedence.

Algorithmic pricing searches for a price against an objective

An algorithmic model takes an objective, such as maximizing margin dollars, and searches within constraints for the price that best serves it, updating as data arrives. It can find moves a person would not write as a rule. The tradeoff is inspectability: if you cannot see why a price changed, you cannot defend it to a finance lead or fix it when inputs go bad. Constraints are what keep it safe.

In practice the split is not either/or. Operators tend to run rules for the constraints that must never be violated and let an optimizing model work inside them.

Price optimization methods trade a goal against constraints

Elasticity-based optimization finds where profit peaks, not revenue

Price elasticity is how much unit sales change when price changes. If sales barely move when price rises, the price was too low; if they collapse, it was near the limit. Revenue and profit peak at different prices, which is why the objective matters. Hypothetical illustration: a product costs $12 all-in and sells 100 units a week at $20, or 92 units at $22. Revenue moves from $2,000 to $2,024, a small gain, while profit moves from $800 to $920. The number is invented to show the mechanics, and your own elasticity will differ by product and channel.

Constrained optimization keeps the objective inside a floor and ceiling

Constrained optimization states the goal and the guardrails together: maximize contribution margin, never price below X, never above Y, never move more than Z percent per day. The constraints do the safety work, and they are also where an operator's judgment lives. Set them from margin after fees, brand positioning, and any minimum advertised price obligations, not from what the system suggests.

Test-and-measure loops verify that a change worked

Every optimization method assumes its estimate of demand is right. A test-and-measure loop checks that: change price on a defined set of SKUs, hold everything else you can, compare against a control group over a fixed window, and record what else changed (ads, promotions, Buy Box share, competitor stock). On a marketplace, clean attribution is hard, so treat results as evidence and not proof.

The five models side by side

Choose by margin, competition, inventory risk, and how much control you want

Work through these questions in order. The first two set the floor; the rest select the model.

1. What is your margin after fees? This sets the minimum price. Nothing else matters if the floor is wrong.
2. Who else is on the listing? Shared listings with many sellers push you toward competitor-based rules. A private-label product with a distinct listing has room for demand-based pricing.
3. How exposed is inventory? Long lead times or aging stock make an inventory trigger the first rule to add.
4. How clean is your data? Optimization needs reliable cost, fee, and sales history. If those are messy, start with rules.
5. How much do you need to explain? If finance or a client asks why a price moved, favor logic with a visible change log.

What the usual half-measures keep and give up

Most operators pass through the same four steps. A spreadsheet with manual edits keeps full control and cannot react faster than the person maintaining it. Marketplace-native repricing rules react automatically but cover a limited set of conditions, so check whether they account for margin after fees. A standalone repricer handles competitor matching well and often works without inventory or advertising context. A pricing platform that combines rules, inventory triggers, and margin visibility handles more, at the cost of having to define your own guardrails up front. For how repricing differs from dynamic pricing as a discipline, see dynamic pricing vs repricing.

Where Trellis Dynamic Pricing fits: operator-set guardrails with automatic execution

Trellis Dynamic Pricing is the part of the Trellis platform that automates price changes for marketplace sellers, alongside its ads automation. In pricing terms, it lets an operator define the conditions and limits, then executes price changes automatically inside them. Two of the models above map directly: Buy Box competitive repricing for the competitor-based case, and Days on Hand rules for the inventory-based case, where price adjusts when cover crosses a threshold you set. Margin visibility after fees lets you set the cost-plus floor from real numbers instead of estimates.

The honest limits are worth stating. Trellis Dynamic Pricing executes the logic you configure; it does not choose your pricing strategy or repair a cost structure that leaves no room to move. Ads and pricing run as parallel mechanisms with shared visibility, so you can see how a price change lines up with advertising performance, but they are not algorithmically coordinated with each other. For the Amazon-specific mechanics, see our Amazon dynamic pricing guide and the overview of dynamic pricing software, and for other retail channels see ecommerce dynamic pricing.

Conclusion: pick the model per job, then put a floor under all of them

There is no single best dynamic pricing model. Cost-plus sets the floor, competitor-based rules defend the Buy Box, demand-based pricing captures value where you have pricing power, and inventory triggers pace sell-through. Rules make the logic legible; optimization finds prices you would not write by hand; the strongest setups combine them with hard limits. Start with the floor and one inventory or competitor rule, watch the change history for a few weeks, then add complexity where the data supports it. If you are weighing how to run this across a larger catalog, book a walkthrough and we will show how the guardrail model works in Trellis.