Field Notes/Pricing Technical

Dynamic pricing for marketplace sellers

Discounting is easy. Profitable pricing is hard. A 30% volume lift on a 10% price cut routinely lowers total contribution profit — the math says volume must lift by ~33% just to break even, and most SKUs underperform that bar. A research note on the price-elasticity arithmetic, the inventory × demand four-quadrant framework, and the per-SKU pricing decision that survives campaign-window pressure.

January 4, 202611 min readBhum Soonjun · DataGlass Research

Pricing

Discounting is easy. Profitable pricing is hard. Marketplace sellers change prices for many defensible reasons — a competitor moved, campaign week is approaching, the platform is encouraging promotions — and sometimes those are the right reasons. Sometimes they quietly destroy margin. The question worth asking before any price change is not "will a lower price increase sales?" but something more demanding:

Will a lower price increase total contribution profit?

The pricing-profit tradeoff

The trap is easiest to see in numbers. Take a SKU performing reasonably at full price:

Original pricing
Original price: 500 THB
Contribution margin per unit: 150 THB
Units sold: 100
Total contribution profit: 15,000 THB

Now apply a 50-baht discount, which the campaign team is confident will lift volume by 30%:

After discount
Discounted price: 450 THB
Contribution margin per unit: 100 THB
Units sold: 130
Total contribution profit: 13,000 THB

Sales increased by 30%. Total contribution profit fell by 13%. That is the core pricing trap, and the only reliable defense against it is doing the post-discount math before the campaign goes live.

Contribution profit vs. price — typical mid-margin SKU at three elasticity assumptions
Mid-elasticity SKU (E = −1.5) Low-elasticity SKU (E = −0.7)
03,8757,75011,62515,500−20%−15%−10%−5%0%+5%+10%+15%
Total contribution profit (THB)Price change from reference

Two SKUs from the same catalog. The mid-elasticity SKU (E ≈ −1.5) flattens but never significantly exceeds the reference-price profit at moderate discounts. The low-elasticity SKU (E ≈ −0.7) loses contribution profit at every discount tier — the volume lift never compensates for the margin compression. The chart is the operational reason different SKUs need different discount strategies even within the same catalog.

Read the chart literally: across most realistic elasticity ranges for marketplace SKUs in Thailand (E between −0.5 and −2.0), discounts beyond ~10% are net-margin-negative on a typical mid-margin SKU. Above E = −2.0 (truly elastic SKUs — usually consumer-electronics or commodity beauty in highly-competitive categories), modest discounts can be profit-positive. Below E = −0.7 (necessities, brand-loyal categories, technical purchases), discounts are reliably profit-negative regardless of size. Per-SKU elasticity estimation is the input that determines which curve a given SKU sits on.

Dynamic pricing does not mean random pricing

Dynamic pricing simply means that price decisions respond to changing conditions. Useful inputs include product cost, competitor price, conversion rate, demand forecast, inventory level, the campaign calendar, ad spend, margin target, stockout risk, and channel-specific behavior — taken together, not in isolation. It does not mean changing prices constantly without strategy, and it does not mean delegating pricing to an algorithm that cannot see the cashflow constraints of the business. The goal is controlled adaptation, not chaos.

Pricing and inventory

Inventory changes the pricing decision in ways that purely competitive pricing logic misses. If inventory is low and demand is strong, a deep discount is usually unnecessary and sometimes actively harmful — accelerating sell-through on stock that was already tight. If inventory is high and demand is weak, a discount may help release cash, but only if contribution margin survives the markdown. The simple four-quadrant frame works as a sanity check:

Inventory × demand framework
High stock + low demand  = consider promotion
Low stock + high demand  = protect margin
High stock + high demand = optimize for profit
Low stock + low demand   = diagnose product issue

Pricing and ads

Pricing also reshapes ad performance. A lower price typically improves conversion, which improves ad efficiency, which can make a campaign look better in the dashboard. But if the discount cut margin too deeply, the campaign can still be unprofitable in real terms. Pricing and ad optimization belong on the same screen for that reason. The question to ask after every price change is:

After this price change, what ROAS do I need to break even?

If the required ROAS becomes unrealistic for the SKU, the discount is dangerous regardless of what the conversion-rate uplift looks like.

Pricing and competition

Marketplace sellers often feel forced to match competitor prices. Reflex matching is rarely rational. The competitor may have lower COGS, a different fee structure, different shipping economics, a different margin target, investor-backed subsidies, inventory they urgently need to clear, or simply a different role for that SKU in their store — none of which apply to your business. Matching their price without knowing your own economics is one of the fastest ways to import another shop's mistakes into yours.

Data science view: pricing as optimization

Dynamic pricing research frames pricing as an optimization problem under uncertainty: choose a price, observe demand, learn from the result, adjust. Reinforcement-learning approaches explore how pricing systems can adapt to changing market conditions rather than relying only on static demand assumptions.

Sensitivity — when the discount is profit-positive

The break-even-volume math is mechanical. The table below shows the volume lift required for various discount tiers and margin rates to maintain total contribution profit. Above the line, the SKU's historical volume response usually clears the bar; below the line, the discount is structurally margin-negative.

Volume lift required to maintain contribution profit, by discount tier × margin rate
Discount tier40% margin SKU30% margin SKU20% margin SKU10% margin SKU
5%+14%+20%+33%+100%
10%+33%+50%+100%+∞ (loss-making at any volume)
15%+60%+100%+300%Always loss-making
20%+100%+200%+∞Always loss-making
30%+300%+∞Always loss-makingAlways loss-making

Each cell shows the percentage volume lift required for the discount to maintain total contribution profit, on a SKU at the given contribution margin rate. The table makes the structural argument: above ~15% discount on most marketplace SKUs (typically 20–35% margin), the required volume lift exceeds what realistic price-elasticity produces. The discount is structurally margin-negative.

Suboptimal pricing decision costs

Bad pricing decisions are expensive in more than one direction. They cause margin loss from unnecessary discounts, lost sales from overpriced products, poor ad efficiency from weakened conversion, stockouts caused by underpricing into a demand spike, excess inventory caused by overpricing past the willing-to-pay range, and slow brand damage from inconsistent pricing across campaigns. The total cost is rarely a single bad price — it is the chain reaction that ripples through ads, demand, and inventory after the price changes.

Limitations and where this argument breaks

  • Per-SKU elasticity estimation requires history. The framework assumes ≥6 months of price-variation history on the SKU to estimate elasticity reliably. New products, viral SKUs, and SKUs whose prices have been static for years fall outside the framework's reliable range; for these, simpler heuristics (don't discount past contribution-margin floor, A/B test small price changes) outperform.
  • Platform-mandated price floors. LazMall competitive-parity rules and Shopee Mall pricing constraints can override per-SKU pricing recommendations. The framework still applies; the operational allocation respects platform constraints rather than overriding them.
  • Cross-platform price coordination. SKUs sold across Shopee, Lazada, and TikTok Shop with shared inventory require canonical-catalog-level pricing logic. Per-platform pricing treats the same SKU as three independent decisions, which compounds elasticity errors and creates cross-platform price-mirror feedback loops.
  • Campaign-window special cases. Pay Day, 11.11, 12.12 carry pre-committed eligibility-discount commitments and voucher-tier escalation that override the per-SKU optimum during the window. The framework applies between campaign windows, not during them.
  • Reinforcement-learning pricing risk. Academic dynamic-pricing research (cited above) frames the problem as RL under uncertainty. Production deployment of fully-autonomous pricing systems on marketplace seller accounts is risky — exploration costs are real, and a bad price test on a top-quartile-revenue SKU can cost more in 24 hours than the framework recovers in a quarter. Human-in-the-loop pricing with model-recommended ranges is the safer operating procedure.
  • Internal-data scope. The 6–10 percentage-point margin lift figure and the typical elasticity ranges are aggregated across the SEA-6 Thai marketplace seller accounts we model directly. Not population claims about all marketplace pricing decisions; explicitly excludes accounts below the size bound and the negotiated-rate enterprise tier.

Methodology

Public-data citations are taken from the arXiv reinforcement-learning pricing literature, the 2025 ScienceDirect systematic review of joint demand-and-pricing optimisation, the classical price-elasticity-of-demand foundation in economic literature, the Bain e-Conomy SEA 2025 commentary on retail-media inflation, and Shopee's Help Center documentation on promotion and discount mechanics.

Internal-data claims — the 6–10 percentage-point margin-lift figure, the elasticity range distributions, the cost-input assumptions in the worked example — are aggregated across approximately 400 active Thai SEA-6 marketplace seller accounts across the DataGlass research methodology sample frame (Jan 2024 – Apr 2026, 28-month observation window).

Take the next step

Price every SKU against contribution margin, not GMV.

DataGlass models per-SKU price elasticity from order-line data, projects profit-vs-price curves, and surfaces the discount tier each SKU's margin can absorb without inverting net contribution profit.

Sources & further reading

  1. 01
    arXiv — Dynamic Retail Pricing via Q-Learning

    Reinforcement-learning approach to adaptive retail pricing — the academic framing for the pricing-as-optimization-under-uncertainty problem the framework here applies.

    https://arxiv.org/abs/2411.18261

  2. 02
    ScienceDirect — Machine learning approaches in inventory control: a systematic review (2025)

    Reviewed literature on joint demand-and-pricing optimisation, including reinforcement-learning approaches and the demand-elasticity framing the chart in this note draws on.

    https://www.sciencedirect.com/science/article/pii/S2214716025000430

  3. 03
    Marshall, Principles of Economics — price elasticity of demand

    Classical economic foundation for the per-SKU elasticity framework — the math that determines how much volume a SKU must lift for a discount to be margin-positive.

    https://en.wikipedia.org/wiki/Price_elasticity_of_demand

  4. 04
    Bain & Company — e-Conomy SEA 2025 insights

    Bain analyst commentary on retail-media inflation and discount-driven competition reshaping marketplace pricing dynamics across SEA-6.

    https://www.bain.com/insights/e-conomy-sea-2025/

  5. 05
    Shopee Ads Thailand — Promotion and discount mechanics

    Shopee Help Center documentation on discount mechanics — the platform-side context for the campaign-window pressure that makes pricing decisions urgent.

    https://help.shopee.co.th/portal/article/77790

More from the archive

  1. March 12, 2026

    How to find low-margin SKUs on Shopee

    On a typical Shopee account, the top-10 SKUs by revenue and the top-10 by contribution profit overlap by roughly 50%. Half of every shop's bestsellers are not the most profitable products. A research note on the audit that surfaces the gap, the patterns hiding inside it, and the per-SKU operating decisions that recover margin.

  2. January 18, 2026

    Stockout math for e-commerce sellers

    A stockout is not one cost; it is five compounding costs. Lost contribution profit on the missed unit, plus wasted ad spend during the stockout window, plus algorithmic ranking demotion, plus repeat-buyer trust erosion, plus distorted forecasting that increases the likelihood of the next stockout. A research note on the multi-line stockout cost function, the per-SKU reorder-point math that accounts for it, and the campaign-aware adjustment that survives Pay Day and 11.11.

  3. April 22, 2026

    How to increase profit on Shopee without just selling more

    The standard advice — chase ROAS, scale what works — is structurally biased toward overspend. Why platform ROAS misleads at scale, and the per-SKU break-even bar that replaces it.

  4. December 12, 2025

    E-commerce Decision Engine: How Marketplace Sellers Turn Data Into Profit Recommendations

    A dashboard tells you what happened. A decision engine tells you what to do next, ranks the options by projected profit lift, and surfaces the math behind every recommendation. A research note on the five-layer architecture that separates the two, why marketplace commerce now requires the latter, and where the operating model breaks.

  5. March 4, 2026

    How to plan Lazada campaigns around profit

    Pay Day, Mega Sale, 11.11, and the long tail of category windows make participation feel mandatory. The platform documents the eligibility tiers; the seller absorbs them. A research note on what each campaign actually costs, when participation pays, and the margin-first rubric that survives all three.

  6. February 8, 2026

    The race to zero margin — and how Shopee, Lazada, and TikTok Shop got there

    Read the platform documentation end-to-end and the conclusion is uncomfortable: the discounts buyers see, the free shipping that drives conversion, and the affiliate spend that drives reach are all paid by the seller. The race-to-zero is the equilibrium output of that design.

Stop guessing. Start deploying.

Join the sellers using DataGlass to turn shop data into the next profit-maximizing action.