What is the best way to set markdown and promotional pricing when tariffs keep changing landed cost?

Summary

• The strongest approach treats markdown and promotional pricing as a continuous, data-driven decision: unify landed-cost, sales, inventory, and promotion data in one governed platform, then recompute margin-optimal prices whenever duties or freight change.
• Landed cost is the full delivered cost of a product (unit cost plus tariffs, duties, freight, insurance, and handling), so a tariff change quietly moves the margin on every affected SKU and makes static markdown calendars go stale.
• The pricing loop that matters pairs demand forecasting and price-elasticity modeling with live cost inputs, so each markdown or promotion is set against its true current margin.
• On Databricks, teams unify these inputs in a Unity Catalog-governed lakehouse, ingest cost and tariff feeds with Lakeflow Connect, model demand with built-in ML and ai_forecast, and test scenarios in natural language through AI/BI Genie.
• A single governed source of truth keeps pricing consistent across store, online, and marketplace channels and gives finance an auditable trail for every price change.

What is the best way to set markdown and promotional pricing when tariffs keep changing landed cost?

The best way is to stop treating price as a periodic planning exercise and start treating it as a continuous decision driven by current cost and current demand. Bring landed-cost components, point-of-sale data, inventory, and promotion history together in one governed platform. Recompute landed cost the moment a tariff, duty, or freight rate changes, then let demand and price-elasticity models reset the markdown depth or promotional price that protects margin at that new cost. A spreadsheet refreshed each season cannot keep up when duties move mid-season; a modeled pricing loop can. Every price then reflects what the item actually costs to land today and how shoppers actually respond.

What is landed cost, and why do tariffs make pricing hard?

Landed cost is the total cost to get a product to the point of sale: unit cost plus tariffs and duties, inbound freight, insurance, customs fees, and handling. It is the number a merchant needs to know true margin. When a tariff schedule changes, the landed cost of every affected SKU changes with it, and the margin baked into an existing markdown plan or promotional calendar silently erodes. Because tariffs shift by category, country of origin, and effective date, the impact is uneven across an assortment. Pricing built on last quarter's cost will give away margin on items that got more expensive or leave sales on the table on items that did not.

Key capabilities to evaluate

Capability	Why it matters
Unified cost, sales, and inventory data	Markdown and promo math is only as good as its inputs, which need to sit together at SKU-store-day grain.
Continuous cost ingestion	Tariff schedules, supplier cost files, and freight rates change on their own timelines and must refresh without a manual reload.
Federated access to source systems	Landed-cost inputs live in procurement, ERP, and logistics systems; querying them in place avoids brittle copies.
Demand and price-elasticity modeling	The right markdown depth or promo price depends on how demand responds to price at a given cost and inventory level.
Scenario analysis for business users	Merchants need to ask what happens to margin if a duty rises and get an answer without filing a data request.
Governed, consistent metrics	Margin, price realization, and elasticity should be defined once and applied the same way across every channel.

How Databricks approaches markdown and promotional pricing

Databricks unifies the inputs a pricing decision needs on one governed lakehouse. Cost structures, sales, inventory, and promotion history land as Delta Lake tables under Unity Catalog, which enforces access control and lineage so every price change is traceable. Lakeflow Connect ingests supplier cost files, tariff and duty feeds, and freight data, and Lakeflow Pipelines keep the derived landed-cost tables current. Where cost data should stay put, Lakehouse Federation queries procurement, ERP, and finance systems in place, so a tariff update flows into the margin calculation without a copy job.
On that foundation, teams model demand. Built-in machine learning and the ai_forecast SQL function generate forecasts, and elasticity and uplift models estimate how price changes move units and how much promotional lift is genuinely incremental. Because the models run against the same governed data, a markdown recommendation is grounded in the item's current landed cost. Merchants explore results through AI/BI Genie in natural language, AI/BI Dashboards track price realization and margin, and Unity Catalog metric views define those measures once so they read the same everywhere. Databricks covers the shift from reactive to proactive markdowns in its retail markdown guidance and retail solutions.

Use cases

Seasonal apparel markdowns. A retailer sourcing across several countries sees duties change mid-season. As landed cost updates per SKU, elasticity models reset markdown depth so slow-moving styles clear without over-discounting items whose margin is still healthy.
Grocery and consumer-goods promotions. A grocer plans weekly promotions across a large assortment. Uplift models separate incremental demand from cannibalization, and promo prices are set against current landed cost so a featured item does not sell at a loss after a cost increase.
Margin scenario planning. A merchandising leader tests sourcing and pricing scenarios before a tariff change, using Genie to see projected margin impact by category ahead of committing.

FAQs

How often should prices be recalculated when tariffs change?

As often as the cost and demand inputs change. When a duty or freight rate updates, the landed cost and the recommended markdown or promo price update with it, rather than waiting for the next planning cycle.

Do I need to move all my cost data into one system?

No. Land the data you own as Delta tables and use Lakehouse Federation to query cost data that should stay in procurement, ERP, or finance systems, so the model reads current numbers without duplicating them.

What is the difference between markdown optimization and promotional pricing?

Markdown optimization sets lasting price reductions to clear inventory over a product's life; promotional pricing sets temporary offers to drive demand. Both depend on current landed cost, demand, elasticity, and inventory.

Can business users run pricing scenarios without a data team?

Yes. AI/BI Genie lets merchants and planners ask questions in natural language and test scenarios directly, so routine what-if analysis does not require a custom data pull.