DEV Community: Mercur

What Is a Marketplace Buy Box and How It Works

Rigby — Fri, 31 Jul 2026 12:03:50 +0000

Three sellers list the same iPhone on your marketplace. A buyer lands on the product page and clicks "add to cart." Which seller gets the sale?

That decision is the buy box, and it is one of the least-understood mechanics in multi-vendor commerce. Most teams planning a marketplace never think about it until several sellers compete on the same product and the platform has to pick a winner.

This guide explains what a marketplace buy box is, how the winning offer gets selected, and why choosing a winner fairly is harder than it looks - hard enough that most platforms leave you to build it yourself.

What a marketplace buy box is

A buy box is the mechanism that selects one winning offer when multiple sellers offer the same product. It is the offer attached to the main "buy now" button, with the other sellers' offers shown as alternatives.

The concept only exists when a single product card carries many offers. That requires the platform to recognise that ten sellers listing the same item are selling one product, not ten separate products - the master-data foundation behind shared master products. Without it, there is no group of competing offers to choose from.

The buy box is where seller competition resolves into a sale. It decides which seller wins the default purchase, so it directly shapes seller behaviour, buyer trust, and your marketplace's economics.

Why the buy box decides who wins the sale

On a marketplace, most buyers take the default. They click buy without comparing every seller's offer, so whichever offer holds the buy box captures the majority of sales for that product.

That makes the buy box the single most valuable position on a product page. Sellers compete for it, and how you award it sets the incentives for your whole seller base - on price, on stock reliability, on service quality.

Get the logic wrong and you reward the wrong behaviour: a seller who is cheapest but always out of stock, or one who wins by a rule buyers can't understand. Get it right and the buy box becomes a lever that pushes your sellers toward the outcomes you want.

How the winning offer gets picked

A buy box resolves a group of competing offers into one winner using a defined order of criteria. A sound default works through them in sequence.

Availability first

An offer that is out of stock cannot win. The buy box filters to sellers who can fulfil the order before it looks at anything else. A cheaper price on an item nobody can ship is worthless.

Price next

Among in-stock offers, the lower price generally wins. Price is the criterion buyers care about most and the one sellers compete on hardest, so it carries the most weight after availability.

Seller quality as a tiebreaker

When offers tie on price, a signal of seller quality breaks the tie - for example a preferred or premium-seller flag. This is where the platform can reward the sellers who fulfil reliably, rather than deciding purely on the lowest number.

A final, stable tiebreaker

If offers still tie, the platform needs one last deterministic rule so the winner is always the same. Without it, the buy box would pick differently on each page load, which brings us to the point most teams miss.

Why a good buy box has to be deterministic

A buy box has to return the same winner every time, given the same offers. If the selection is even slightly random, the winning seller flickers between page loads, and that breaks trust on both sides of the marketplace.

Buyers see a price change when they refresh. Sellers can't tell whether they hold the buy box or why they lost it. A buy box that isn't deterministic isn't a buy box - it's a coin flip that happens to look like a ranking.

Determinism also has to survive change. When a price or stock level updates, the winner should recompute promptly, so the buy box never shows a winner whose price or availability is already stale.

Showing who sells is a legal duty, not a nicety

A buy box doesn't just pick a winner - it has to show which seller stands behind the winning offer. In most markets, telling the buyer who they are purchasing from is a legal requirement, not an optional detail.

That means the winning offer and its alternatives must each carry clear seller identity - who they are and whether they are active - alongside the price. A buy box that hides the seller is not just worse for trust; it can put you out of compliance.

Why most platforms make you build the buy box yourself

Here is the part teams discover late: buy box logic is core marketplace mechanics, yet most marketplace platforms don't ship a real one. Their storefront layers assume a single seller per product, so when you go multi-vendor, selecting and exposing a winning offer becomes custom work you own and maintain.

That is a significant build. It needs a winning-offer selection with a defined, deterministic order, offer and seller data exposed on the storefront, faceted search that works across offers, and the ability to change the ranking logic later as your marketplace matures. Buy box selection sits on the hot path of every product page, so building it as an afterthought is where marketplaces get slow and inconsistent.

Mercur

Mercur is an enterprise-grade Open Core marketplace platform - zero license fees, zero GMV fees, full code ownership. Open Core covers roughly 80% of what it takes to launch; the buy box sits in the Enterprise tier that handles the more complex 20%.

Mercur Enterprise ships a native buy box: deterministic winning-offer selection across competing offers on a shared product, with the seller's identity exposed on the storefront and faceted search across offers. The selection logic is extensible - the ranking is a pluggable strategy, so you can replace the default with your own rather than building the buy box from scratch.

That extensibility is the direction of travel too: per-category rules and ranking driven by seller performance are where a configurable buy box heads next, on a foundation you own rather than a closed platform you rent. See Mercur Enterprise and the marketplace software comparison.

Frequently asked questions

What is a buy box in a marketplace?

A buy box is the mechanism that picks one winning offer when several sellers offer the same product. The winning offer gets the main "buy now" button; the others appear as alternatives. It only exists when one product card carries multiple sellers' offers.

How is the buy box winner chosen?

Through an ordered set of criteria: availability first (out-of-stock offers can't win), then price (lower usually wins), then a seller-quality signal to break ties, and finally a stable deterministic tiebreaker so the winner is the same on every page load.

Why does the buy box winner need to be deterministic?

So the same offers always produce the same winner. If selection is random, the winning seller and displayed price change between page loads, which confuses buyers and makes it impossible for sellers to know whether they hold the buy box or why they lost it.

Do marketplace platforms include a buy box out of the box?

Most do not. Storefront layers usually assume one seller per product, so selecting and exposing a winning offer becomes custom work. Mercur Enterprise is an exception, shipping a native, extensible buy box on a foundation you own.

Why Inventory Risk Is the Silent Killer of Ecommerce Growth?

Rigby — Thu, 30 Jul 2026 13:38:45 +0000

Inventory risk is the financial exposure you take on every time you buy stock before you have a buyer for it. You pay for goods now, against a forecast, and you absorb whatever the difference turns out to be.

That exposure sits quietly on your balance sheet while revenue grows. It rarely appears as a line item called “inventory risk,” which is why most teams meet it as a cash flow problem, a margin problem, or a warehouse problem instead.

This article breaks down:

The 7 types of inventory risk
How to measure your exposure before it reaches your P&L?
What you can do about it inside your current operation?
What changes when the stock stops being yours?

Key takeaways

Inventory risk is the cost of owning stock you have not yet sold, and it grows with catalog breadth rather than with revenue.
It fails in two opposite directions at once: too much inventory locks up cash, while inventory shortages hand sales to competitors.
Getting stock levels wrong costs global retail 6.5% of sales, and the stockout side of that is more than twice the size of the overstock side.
Shrinkage, obsolescence, supplier delays, and markdowns are separate risks with separate fixes, so treating them as one problem produces the wrong response.
Inventory KPIs such as turnover, aging buckets, and inventory accuracy expose the exposure months before your financial statements do.
Better forecasting, safety stock discipline, supplier diversification, and real-time inventory data all reduce the variance around your bet.
None of those tactics change who funds the stock, which is the reason marketplace models keep coming up in this conversation.

What is inventory risk, and why does it grow faster than revenue?

Inventory risk is the probability that stock you own loses money instead of making it. The loss can come from goods that never sell, goods that sell below cost, goods that disappear, or goods that arrive after the demand has passed.

Every purchase order is a forecast with money attached. You commit capital at the point when you know the least about future customer demand, and you find out whether you were right months later. The size of the bet is set by how many SKUs you carry, not by how much you sell.

Revenue can double on the same 500 products, while a catalog that goes from 500 to 5,000 products multiplies the number of independent forecasts you have to get right.

Why does inventory risk scale non-linearly with your catalog?

Forecast error compounds as you add products. A new SKU has no historical sales data, so its first buy is close to a guess, and long tail items have the thinnest demand signal of all.

At the same time, your inventory management processes stay roughly the same. The team that could review 500 stock levels weekly cannot review 5,000, so decisions shift from analysis to reorder points that nobody revisits.

The result is a portfolio in which a growing share of inventory value is tied to items nobody is actively watching. Slow-moving inventory accumulates in the gaps between what the team has time to review and what the system reorders automatically.

This is the mechanism behind the scaling wall that shows up in the comparison of marketplace vs eCommerce growth models. Assortment growth funded from your own balance sheet turns catalog expansion into a cash flow decision.

The two directions of failure: too much inventory and inventory shortages

Significant inventory risk runs in both directions simultaneously, which is what makes it hard to manage with a single lever.

IHL Group put the combined cost of getting this wrong at 6.5% of global retail sales in 2025. Its breakdown of that loss puts out-of-stocks at $1.2 trillion and overstocks at $572 billion. The shortage side of inventory risk is more than twice the size of the excess side, which is the opposite of where most teams focus their attention.

Excess stock ties up cash, fills warehouse space, and eventually gets written off or discounted. Inventory shortages do the opposite kind of damage, since a stockout produces an immediate lost sale and sends the buyer to a competitor who had availability.

Most operations carry both at once. The same business can have twelve months of cover on one product line and repeated stockouts on its best sellers, because the forecasting error is distributed unevenly across the catalog.

Pushing hard in one direction makes the other worse. Cutting stock levels to release cash raises stockout risk, and raising safety stock to protect inventory availability raises carrying costs.

7 types of inventory risk sitting on your balance sheet

Inventory risk is a category. Each type has a different cause, a different signal, and a different fix, so a diagnosis at the category level tends to produce a response that misses.

The seven below cover almost everything that turns owned stock into a loss. Common inventory risks rarely appear alone, since one of them usually triggers the next.

1. Forecast risk

Forecast risk is the gap between predicted and actual demand. It sits upstream of most other inventory risks, because the purchase quantity is set before anyone knows how demand will behave.

Accurate demand forecasting depends on historical sales data, and that data degrades as consumer preferences and market trends move. Demand variability rises in categories exposed to market volatility, seasonality, or a competitor's pricing decisions.

2. Stockout and lost sales risk

Stockout risk is the exposure created by not having enough. The immediate cost is the lost sale, and the second cost is the customer dissatisfaction that follows a buyer who came to your site ready to purchase.

IHL Group attributes $690.9 billion of retail losses to empty shelves alone, ahead of every other cause of inventory distortion it measures. The revenue you lose to a stockout never appears in your reporting, which is why underestimating demand tends to go uncorrected for longer than overstocking.

3. Obsolescence and dead stock risk

Obsolete inventory is stock that still exists physically but no longer has a buyer at a price you would accept. Technology categories reach this point through product cycles, apparel through seasons, and food through stock shelf life.

Dead stock is the end state, where the item stops moving entirely and only generates carrying costs. Our breakdown of how excess inventory builds up and what it costs covers the cash flow side of this in detail.

4. Shrinkage risk

Inventory shrinkage is the difference between the stock your inventory records show and the stock you physically have. It comes from inventory theft, inventory damage, and administrative errors in receiving, picking, and returns.

Shrink cost US retailers $112.1 billion, or 1.6% of sales, in fiscal 2022, which is the last year the National Retail Federation measured it on that basis. In the UK, the British Retail Consortium recorded 5.5 million detected shoplifting incidents costing retailers close to £400 million in its 2026 crime report, and notes the true figure is higher because many incidents go undetected.

Poor inventory control widens the gap further, since inventory loss you have not detected is inventory you are still counting on to fulfill orders.

5. Supplier and lead-time risk

An unreliable supplier converts your inventory plan into a guess about someone else's operations. Late deliveries create stockouts, early or partial deliveries create receiving errors, and quality failures create stock you cannot sell.

Supply chain disruptions extend lead times, and longer lead times force you to order earlier against a weaker demand signal. Concentration makes this worse, because a single sourcing relationship gives you no fallback when it fails.

6. Price and markdown risk

Price risk is the exposure between what you paid and what the market will now pay. Input costs, tariffs, and currency movements change your cost base, while competitor pricing changes your ceiling.

Markdown risk is the same problem discovered late. Once stock ages past its window, discounting becomes the only way to convert it to cash, and each markdown cycle resets the margin you planned for.

7. Carrying cost risk

Carrying costs are what you pay to hold stock: storage, insurance, capital, handling, shrinkage, and obsolescence. APQC benchmarking puts the median inventory carrying cost at 10% of inventory value across more than 6,000 companies.

Increased carrying costs are the compounding element here. Stock that sits for two quarters instead of one delays revenue and consumes a tenth of its own worth again on the way.

How to measure your inventory risk exposure?

You cannot manage inventory risk from a total inventory figure, because a single number hides which of the seven types you are carrying. Measurement has to break the stock down by age, accuracy, and movement.

The metrics below surface exposure while you can still act on it. Financial statements report the same problem later, after the write-off.

The inventory KPIs that expose risk before your P&L does

Four measures do most of the diagnostic work. Each one points at a different type of risk, so tracking them together tells you which lever to pull.

1. Inventory turnover by product line

Turnover shows how many times you sell and replace stock in a period. Falling turnover on a line with stable market demand means inventory levels have moved ahead of real customer demand.

Measure it per line rather than for the whole business. A blended figure hides the slow-moving inventory inside an otherwise healthy average.

2. Days inventory outstanding against a benchmark

Days inventory outstanding tells you how long stock sits before it sells. The Hackett Group found European companies pushed DIO to 68.9 days in 2025, the highest in a decade, as buffers went up to absorb supply chain uncertainty.

For a wider reference point, the US Census Bureau publishes a monthly retail inventories-to-sales ratio you can compare your own cover against. A ratio drifting up while your sales stay flat is inventory risk accumulating in plain sight.

3. Inventory aging buckets

Aging buckets group stock by how long it has been held: under 30 days, 30 to 90, 90 to 180, and over 180. The tail past 180 days is your obsolescence exposure expressed in inventory value.

Watch the direction of travel rather than the absolute number. A tail that grows quarter over quarter means your purchasing is running ahead of your selling on a specific set of SKUs.

4. Inventory accuracy rate

Inventory accuracy compares your inventory records against a physical count of the same items. Anything below the high nineties means your availability promises to customers rest on numbers that are partly wrong.

This metric isolates shrinkage and process failure. A falling accuracy rate tells you the loss is happening inside your operation rather than in the market.

Why does your inventory valuation method change the number you see?

Inventory valuation methods determine the cost you carry each unit at, which changes the exposure your reporting shows even when the physical stock is identical.

FIFO assigns the oldest costs to goods sold and leaves recent costs on the balance sheet. In a rising cost environment that reports a higher inventory value, so the same overstock looks like a larger asset and a larger write-off risk.

Weighted average cost smooths the effect and makes period comparisons cleaner, at the cost of hiding how much of your holding was bought at old prices. Either way, the valuation method sets the size of the loss you recognize when stock is finally written down.

Pick one and hold it, then track exposure in units and days alongside currency. Unit and age data does not move when accounting policy does.

How to reduce inventory risk?

Everything in this section is available to you without changing how your business works. These are the standard inventory risk management moves, and they are worth doing.

Each one narrows the range of outcomes around your inventory bet. What none of them does is change the fact that you placed the bet, which matters for the section that follows.

Sharpen demand forecasting and safety stock

Forecasting and safety stock are the two settings that decide how much of your forecast error turns into a loss. Both work better when they operate per SKU rather than per category.

Forecast on current demand signals

Accurate demand forecasting needs inputs that move at the speed of the market: recent sell-through, search and add-to-cart behaviour, return rates, and competitor availability. Historical sales data alone describes a market that has already changed.

McKinsey put the gain from AI-driven supply chain forecasting at a 20% to 50% reduction in forecasting error, and up to 65% less lost sales and product unavailability.

Gartner expects 70% of large organizations to adopt AI-based demand forecasting by 2030, which sets the baseline your competitors will be forecasting against.

AI-driven forecasting tools help most on the long tail, where human review capacity runs out first. Their contribution is consistency across thousands of SKUs rather than accuracy on any single one.

Set safety stock and reorder points per SKU

Safety stock is insurance against demand variability and lead-time variability, and its correct size differs by product. A blanket policy overprotects predictable items and underprotects volatile ones.

Reorder points should reflect each supplier's actual lead time, including the variance. Maintaining optimal safety stock on your top sellers protects against unexpected demand spikes without funding cover across the whole catalog.

Kill SKUs on a schedule

Assortment discipline is a forecasting tool. Every product you stop carrying removes a forecast you were going to get wrong.

Set a review cadence with a defined exit rule, such as no sales in two quarters. Optimal inventory levels are easier to hit on a catalog you have deliberately pruned.

De-risk the supply side

Supplier behaviour sets the lead times and reliability your entire inventory plan is built on. Three moves reduce how much of that you are exposed to.

Diversify suppliers instead of optimizing one

A second qualified supplier for your highest volume items costs you some unit economics and removes a single point of failure. Supplier diversification builds resilience into the supply chain rather than into your warehouse.

The alternative is paying for the same protection in safety stock, which is more expensive and depreciates. Avoiding reliance on a single supplier is the cheaper version of the same insurance.

Put delivery performance in the contract

Supplier contracts should specify delivery dates, quality standards, and what happens when either is missed. Accountability measures turn supplier reliability from a hope into a term you can enforce.

Run regular audits against those terms. Monitoring delivery performance and quality gives you the evidence to renegotiate or to move volume before a delay becomes a stockout.

Run scenario planning on your top exposures

Scenario planning prepares the response before you need it. Work through what you would do if your largest supplier missed a shipment by six weeks, and decide the answer while nobody is under pressure.

Concentrate the exercise on the SKUs carrying the most inventory value. Preparing for supply chain disruptions on your top twenty items covers most of the exposure.

Close the visibility and shrinkage gap

Shrinkage and inventory inaccuracy are the risks most often left to an annual count. Closing the gap is largely an operational and tooling question.

Move to real-time inventory data

Cloud inventory management software gives you real-time stock visibility across locations, which is the input every other decision here depends on. Inventory management software can automate reorder points and flag exceptions instead of waiting for a person to notice.

Inventory tracking technologies such as barcode scanning and RFID tags reduce the manual steps where errors enter. IHL Group expects RFID deployment to grow 291% over the next two years, while fewer than a quarter of retailers have rolled out AI and machine learning in the areas where inventory distortion actually happens.

Automated systems remove most human error from receiving and picking, which is where inventory records usually diverge from reality.

Cycle count instead of waiting for the annual count

Cycle counting checks a subset of items continuously rather than everything once a year. It helps identify theft and process discrepancies sooner, while the trail is still traceable.

Weight the count frequency toward high-value and high-movement items. Those are where inventory shrinkage costs you most per incident.

Add access controls where stock is handled

Security measures reduce inventory theft, and most of the reduction comes from basic controls rather than sophisticated ones. Access controls on stock areas, surveillance systems at handling points, and named accountability for counts cover the majority of internal loss.

Pair the controls with the cycle count data. A discrepancy that repeats in one location tells you where to look.

Make quality control a scheduled process

Regular quality checks prevent damaged stock from becoming unsellable stock you discover at picking. Stringent quality control procedures at receiving catch supplier quality failures while you still have recourse.

Inspect on receipt and again on any item held past its aging threshold. Stock quality degrades in storage, and finding out at the point of shipment costs you the customer as well as the unit.

These measures reduce how often you are wrong and how much it costs when you are. What stays constant is that your growth is funded by stock you bought in advance, and that is the exposure the next section addresses.

How do marketplace models move inventory risk off your balance sheet?

The tactics above manage inventory risk. Changing who owns the stock removes most of it from your books entirely.

In a marketplace model, third-party sellers list and fulfill their own inventory. Your catalog expands without a purchase order, so assortment growth stops consuming working capital.

What changes when sellers hold the stock?

Forecast risk, obsolescence, shrinkage, and carrying costs follow ownership. The seller who holds the goods carries all four, which is a considerable share of the seven types listed earlier.

The amount of cash at stake is not marginal. The Hackett Group calculated that $1.7 trillion sits trapped in excess working capital across publicly traded non-financial companies, or 35% of their gross working capital.

Your exposure shifts to a different set of problems: seller quality, availability accuracy, fulfillment performance, and catalog data. Those are operational risks with operational fixes, and they do not consume cash the way owned inventory does.

The economics change with them. Testing a new category costs you onboarding effort rather than a buy, so the cost of being wrong about a category drops to close to zero.

Adding third-party supply does not remove your own inventory operation, and for most retailers it should not. The point is to stop funding every increment of catalog growth with your own cash.

How Mercur keeps stock records on the seller's side

Everything above depends on one condition: your platform has to record the stock as belonging to the seller. A system that treats all inventory as a single pool you own will hand you the operational load of a marketplace without the balance sheet relief.

That makes the location of stock records the first thing to examine in a marketplace platform, ahead of vendor dashboards, commission rules, and payout timing. It is also where the architecture either supports the model or quietly undermines it.

Mercur is an open-source marketplace platform under an MIT license, built on the modern Medusa commerce framework. It runs alongside an existing commerce stack instead of replacing it.

Ownership is built into the data model. Every inventory item, stock location, shipping profile, and fulfillment set is linked to a specific seller, so stock records live with the party holding the goods.

Sellers publish offers against a shared master variant, each with their own SKU, price, and stock across their own stock locations. Availability is checked when a buyer adds the item to a cart, and inventory is reserved when the order is placed.

A single customer cart splits into one order per seller, each with its own shipping method, with commission calculated per line. Which offer a buyer sees is a merchandising decision Mercur leaves to your storefront rather than deciding for you.

The code is available on GitHub, so you can review how the seller scoping works before committing to it.

There is also a live demo running the complete stack, with a storefront, multiple sellers, and real products, so you can click through exactly what you would ship. Try it yourself at demo.mercurjs.com.

If you would rather start from your own numbers, talk to a marketplace expert about which part of your assortment should move to third-party supply first.

Summary: inventory risk is a model decision, not a warehouse problem

Inventory risk is not a reporting gap you can close with better software. It is the consequence of a growth model that funds assortment from your own capital, and it grows every time you add products.

Better forecasting, tighter safety stock, supplier diversification, and real-time inventory tracking are all worth doing, and they will reduce your losses. They will not change who absorbs the forecast error.

If your cash is tight while revenue grows, if your aging tail keeps extending, and if every category test needs a buy, the constraint is the model rather than the execution. Adding third-party supply alongside your own inventory lets you keep growing the catalog without growing the bet.

The related question of whether your business is ready for that shift is covered in our five signs an eCommerce is ready to become a marketplace.

If you want to work through what a first version would look like in your category, talk to a marketplace expert.

FAQ on inventory risk management

What is inventory risk?

Inventory risk is the financial exposure created by owning stock you have not yet sold. It covers goods that never sell, goods that sell below cost, goods lost to theft or damage, and sales missed because stock ran out. The exposure exists because you commit capital against a forecast, before you know what customer demand will be.

What are the types of inventory risk?

There are seven distinct types: forecast risk, stockout and lost sales risk, obsolescence and dead stock risk, shrinkage risk, supplier and lead-time risk, price and markdown risk, and carrying cost risk. Each has a different cause and a different fix. Forecast risk sits upstream of most of the others, since the purchase quantity is set before demand is known.

How do you calculate inventory risk exposure?

Start with inventory aging buckets to size your obsolescence exposure, then add your inventory accuracy rate to quantify shrinkage and process loss. Inventory turnover by product line shows where stock levels have moved ahead of demand, and days inventory outstanding shows how long cash stays locked in stock. Track all of it in units and days as well as currency, because inventory valuation methods change the reported figure without changing the physical risk.

How can you mitigate inventory risk?

Mitigating inventory risk works on two levels: reducing the variance around your bet, and reducing the size of the bet itself. The first level covers demand forecasting on current signals, per-SKU safety stock and reorder points, supplier diversification, cycle counting, and real-time inventory data. The second means shifting part of your assortment to third-party supply, so sellers fund and hold the stock instead of you.

What causes inventory risk in the supply chain?

Long and variable lead times are the main driver, because they force you to order earlier against a weaker demand signal. Supplier concentration adds to it, since a single sourcing relationship leaves no fallback when deliveries slip. Supply chain disruptions then compound the problem, as teams raise safety stock during a disruption and often do not reduce it once supply stabilizes.

Sources

Mercur - Open source marketplace platform to build and iterate fast

Rigby — Tue, 18 Feb 2025 15:13:14 +0000

In 2024, we built Mercur - an open source marketplace platform, with Medusa as a foundation, designed for projects that want full control over their ecosystem.

Originally Mercur was a simple backend starter, but we are currently developing it into a fully-fledged platform with a complete Next.js frontend, vendor panel, key marketplace functionalities and integrations like Stripe Connect, Resend, Easypost, and much more. All available for everyone - developers, agencies, and businesses - under the MIT license.

New site. New repo. New space to build.

• Explore Mercur → https://www.mercurjs.com
• Star the repo → https://github.com/mercurjs