Best Practices for Managing Inventory in E-commerce

11 October 2026

Inventory is the largest cash commitment most e-commerce businesses make. It is also the one that receives the least analytical attention. Marketing gets dashboards. Inventory gets a spreadsheet that someone updates on Fridays. That imbalance quietly destroys margins. A business can grow revenue 40 percent year over year and still run out of money because too much of that revenue is sitting in a warehouse in the wrong SKUs.

This article is about doing inventory management properly. Not the textbook version, but the version that survives contact with suppliers who ship late, demand that spikes without warning, and customers who expect two-day delivery on everything. The goal is to give you a framework you can actually operate, along with the reasoning behind each decision so you can adapt it to your own situation.

Why Inventory Management Is a Cash Flow Problem Before It Is a Logistics Problem

Most operators think about inventory in terms of availability. Do we have it? Can we ship it? That framing is incomplete. Inventory is cash that has been converted into physical form. Every unit sitting on a shelf is money you cannot spend on advertising, hiring, or product development.

This matters because the cost of carrying inventory is far higher than most people estimate. When you add up the purchase cost, freight, duties, warehousing, insurance, shrinkage, and the opportunity cost of capital tied up for months, the true annual carrying cost typically lands somewhere between 20 and 30 percent of inventory value. A business holding $500,000 in stock is effectively paying $100,000 to $150,000 per year just to own it. That number rarely appears on any report, which is exactly why it gets ignored.

The practical implication is that inventory decisions are financial decisions. Ordering an extra 5,000 units to hit a supplier's volume discount is not a purchasing decision. It is a capital allocation decision with a measurable return, and it should be evaluated like one.

Start With Demand Forecasting You Actually Trust

Every inventory system rests on a forecast. If the forecast is wrong, everything downstream is wrong. The mistake most e-commerce businesses make is treating forecasting as a single number when it should be a range with a confidence level.

Segment Your SKUs Before You Forecast Them

Not every product deserves the same forecasting effort. A useful starting point is to classify SKUs by revenue contribution and demand variability. High-volume, stable products can be forecast with simple moving averages and managed with tight reorder points. Low-volume, erratic products need either higher safety stock or a made-to-order model.

A common approach is to sort products into four buckets:

- High revenue, low variability. These are your workhorses. Forecast them carefully, negotiate hard on cost, and never run out.
- High revenue, high variability. These are the dangerous ones. They generate cash but also generate dead stock when a trend reverses.
- Low revenue, low variability. Automate replenishment and stop thinking about them.
- Low revenue, high variability. These are candidates for discontinuation or dropshipping.

The value of this segmentation is that it tells you where to spend your attention. Most businesses spread analytical effort evenly across all SKUs, which means the products that matter get the same treatment as the ones that do not.

Use Multiple Forecast Horizons

A single forecast number is fragile. A better approach is to maintain three views: a short-term forecast for the next 30 days used for replenishment, a medium-term forecast for 90 days used for purchase orders, and a long-term view for 6 to 12 months used for supplier negotiations and capacity planning.

These horizons will disagree, and that disagreement is useful. If your 30-day forecast says sell 400 units and your 90-day forecast implies 200 per month, something in the data needs explaining. Seasonality, a promotion, or a data error. Catching that discrepancy early is worth more than any forecasting algorithm.

When Simple Beats Sophisticated

There is a strong temptation to buy forecasting software with machine learning models. Sometimes that is right. Often it is not. If you have fewer than 200 SKUs and stable demand, a spreadsheet with a 12-week moving average and a seasonality multiplier will outperform a complex model because it is transparent and you will actually maintain it.

Sophisticated forecasting pays off when you have thousands of SKUs, multiple sales channels, and enough historical data to train on. Below that threshold, the overhead of maintaining the model usually exceeds the accuracy gain.

Safety Stock: The Most Misunderstood Number in Inventory

Safety stock is not a buffer against laziness. It is a calculated response to uncertainty in both demand and lead time. Getting it wrong in either direction is expensive.

If safety stock is too low, you stock out, lose sales, and damage search rankings and customer trust. If it is too high, you tie up cash and risk obsolescence. The right level depends on two variables: how variable your demand is, and how variable your supplier's lead time is.

A practical formula many operators use is:

Safety stock = Z x sqrt((average lead time x demand standard deviation squared) + (average demand squared x lead time standard deviation squared))

The Z value reflects your target service level. A 95 percent service level corresponds to roughly 1.65. A 99 percent service level corresponds to roughly 2.33. The jump from 95 to 99 percent is expensive, and many businesses should not make it. Ask yourself what a stockout actually costs. If the customer simply buys from a competitor and never returns, the cost is high. If they backorder and wait, the cost is low.

Lead Time Variability Is Usually the Bigger Problem

Most businesses focus on demand variability and ignore lead time variability. In practice, lead time is often the larger source of stockouts, especially for imported goods. A supplier who quotes 45 days but delivers anywhere between 35 and 75 days creates enormous uncertainty that no amount of demand forecasting can fix.

The fix is operational, not analytical. Track actual lead times for every supplier. Build a distribution, not an average. Then either hold more safety stock for unreliable suppliers or replace them. A supplier who is 10 percent cheaper but 40 percent less reliable is usually more expensive once you account for the safety stock required to compensate.

Reorder Points and Order Quantities: Two Decisions, Not One

A reorder point answers when to order. An order quantity answers how much. Conflating them is a common error.

The reorder point is the inventory level that triggers a new purchase order. It equals average demand during lead time plus safety stock. If you sell 20 units per day and your lead time is 30 days, you need 600 units just to cover the lead time. Add safety stock of 150 units and your reorder point is 750 units.

The order quantity is a separate optimization. The classic approach is the economic order quantity, which balances ordering costs against holding costs. In e-commerce, the calculation is complicated by volume discounts, container economics, and cash constraints.

The Container Trap

Many importers order in full container quantities because the per-unit freight cost drops significantly. This is often correct. But it can also push you into ordering 6 months of supply when you only need 2, which converts a freight saving into a carrying cost problem.

Run the numbers both ways. If a full container saves $3,000 in freight but adds $8,000 in carrying costs and obsolescence risk, the full container is a bad deal. The right answer depends on your product's shelf life, demand stability, and cash position.

Cash Constraints Change the Math

If cash is tight, you cannot simply optimize for lowest total cost. You have to optimize for return on the cash you have. This sometimes means ordering smaller quantities at higher per-unit cost to keep capital moving. It is a deliberate trade-off, not a mistake, and it should be made consciously rather than by accident.

Multi-Channel Inventory: Where Most Systems Break

Selling on your own site, Amazon, and a wholesale channel simultaneously creates a synchronization problem. If inventory is not shared in real time, you will oversell. If it is shared too aggressively, you will under-sell on channels where you could have captured demand.

There are three common approaches:

Pooled inventory. All channels draw from one pool. Simple and safe against overselling, but it means a slow channel can block a fast one.

Allocated inventory. Each channel gets a dedicated quantity. This protects channel-specific service levels but requires more forecasting and can leave stock stranded.

Hybrid with buffers. Most inventory is pooled, but each channel holds a small buffer to absorb demand spikes.

The right choice depends on your channel mix. If one channel drives 80 percent of revenue, allocate generously to it and treat the others as opportunistic. If channels are balanced, pooling with a small buffer per channel usually works best.

The Amazon-Specific Complication

Selling on Amazon introduces constraints that do not exist elsewhere. FBA inventory has its own lead times, storage limits, and long-term storage fees. Restock limits can force you to choose which SKUs to send, which effectively means Amazon is making your assortment decisions for you.

The practical response is to treat FBA as a separate inventory node with its own reorder logic, not as an extension of your warehouse. Calculate FBA-specific reorder points that account for Amazon's receiving delays, and monitor long-term storage fees as a signal that a SKU is not performing.

Common Mistakes That Quietly Destroy Margin

Ordering on Gut Feel After a Good Month

A strong month is not a signal to double your next order. It is a signal to check whether the increase is seasonal, promotional, or structural. Many businesses have gone bankrupt chasing a trend that reversed in 60 days.

Ignoring the Cost of Stockouts

Stockouts are often treated as a minor inconvenience. In reality, they cost you the sale, the customer's future purchases, and in some cases your search ranking. Amazon, for example, penalizes listings that go out of stock by reducing their visibility. That penalty persists after restocking.

Treating All Suppliers the Same

A domestic supplier with a 5-day lead time and a 30-day payment term is fundamentally different from an overseas supplier with a 60-day lead time and prepayment. They deserve different inventory policies, different safety stock levels, and different payment terms in your cash flow model.

Never Auditing Dead Stock

Every catalog accumulates SKUs that no longer sell. These tie up cash, take up space, and distort your reporting. A quarterly review that flags any SKU with no sales in 90 days is a simple discipline that pays for itself. The decision to liquidate, discount, or bundle is separate, but the review must happen.

Confusing Revenue Growth With Inventory Health

A business can grow revenue while inventory turnover deteriorates. If inventory is growing faster than sales, you are accumulating risk. Track inventory turnover and days of inventory alongside revenue. If those metrics are moving in the wrong direction, growth is masking a problem.

Metrics That Actually Matter

You do not need 20 inventory KPIs. You need a handful that tell you whether the system is working.

Inventory turnover. Cost of goods sold divided by average inventory. Higher is generally better, but too high means you are risking stockouts. Compare against your category, not against an arbitrary target.

Days of inventory. The inverse of turnover, expressed in days. Useful for cash planning.

Sell-through rate. Units sold divided by units received, measured over a defined window. This is the single best indicator of whether a buying decision was correct.

Stockout rate. Percentage of time a SKU is unavailable. Track this by SKU, not just in aggregate, because averages hide the products that matter.

Carrying cost as a percentage of revenue. This is the number most businesses never calculate and should.

Technology: When to Invest and What to Avoid

Spreadsheets work until they do not. The breaking point usually arrives somewhere between 500 and 2,000 SKUs, or when you add a second sales channel. Before that, a well-built spreadsheet is faster, more flexible, and cheaper than any software.

When you do invest, prioritize integration over features. An inventory system that does not sync automatically with your sales channels creates manual work and errors. The most common failure mode is buying a powerful tool that nobody maintains because the data entry is tedious.

For most small and mid-sized e-commerce businesses, the practical stack is an ERP or inventory platform connected to your storefront, your 3PL, and your accounting system. The specific vendor matters less than the integration quality.

Bringing It Together

Inventory management is not a single task. It is a set of connected decisions about forecasting, safety stock, reorder timing, order quantity, channel allocation, and supplier relationships. Each decision has a cost, and the goal is not to eliminate cost but to allocate it deliberately.

The businesses that do this well share a few traits. They forecast at multiple horizons. They track supplier reliability, not just supplier price. They know their carrying cost. They review dead stock on a schedule. And they treat inventory as a financial asset rather than an operational afterthought.

None of this is glamorous. It is the unglamorous work that determines whether a growing e-commerce business becomes profitable or becomes a cautionary tale.