Cost of Goods Sold (COGS) for Product KPI

Cost of Goods Sold (COGS) is a critical KPI that directly impacts gross margin and overall financial health.

By accurately measuring COGS, organizations can enhance operational efficiency and make informed pricing decisions.

This metric influences business outcomes such as profitability and cash flow management.

A lower COGS typically indicates better cost control and resource allocation, while a higher COGS can signal inefficiencies or rising material costs.

Companies leveraging COGS insights can improve forecasting accuracy and strategic alignment, ultimately driving ROI.

Regular analysis of this KPI is essential for maintaining a robust KPI framework and ensuring long-term sustainability.

How Cost of Goods Sold (COGS) for Product Connects to Your Strategy

Cost of Goods Sold (COGS) for Product appears in one of KPI Depot's KPI groups, Product Management, a set of sixty-six metrics, and it ranks fifty-third by priority. That is low, and the honest reading is that this group treats it as a supporting metric rather than one of the numbers a product organization is judged on week to week.

Look at what the group ranks above it. The lead metrics are Customer Satisfaction Score (CSAT), Net Promoter Score (NPS), Customer Lifetime Value (CLTV), Churn Rate, Customer Acquisition Cost (CAC), Revenue Growth, Monthly Recurring Revenue (MRR), and Average Revenue Per User (ARPU). Every one of those is either a customer-perception measure or a revenue-side financial measure. Not one of them is a cost of delivery. The composition says something about the reader this KPI group was built for: a product organization whose working model of profitability runs through acquisition cost and revenue per user, with the cost of producing and shipping the thing sitting further down the list. Its ranking reflects how often product teams open this number, not how much it decides.

Its balanced scorecard placement is the financial perspective, and it is lagging in a stricter sense than most lagging metrics are. This one is not measured, it is closed. It arrives after the period ends, after inventory is counted or rolled forward, after variances are disposed of. By the time the figure is final, the decisions that produced it were taken quarters earlier: the sourcing contract, the production run size, the discount that shifted the product mix. Nothing done in the month it is published will move it. That is worth saying in a KPI group whose lead metrics can all be influenced inside a single sprint.

The first tension is with ARPU and MRR, ranked eighth and seventh. Those are the revenue half of gross margin and this KPI is the cost half, and in practice a single decision moves both. Adding a premium tier, a faster shipping option, a heavier support commitment, or a component the customer can see raises revenue per user and raises unit cost of delivery in the same motion. So the interesting question is never whether COGS rose. It is whether it rose by less than ARPU did. A cost target set without its revenue pairing quietly argues for the cheapest available version of the product, which is a different business than the one the group's lead metrics describe.

The second tension is with CAC, ranked fifth, and it is a classification tension rather than an operational one. CAC and COGS sit on opposite sides of the gross-margin line by convention, not by nature. A large band of real spending can defensibly go either way: onboarding labour, implementation work, free-trial infrastructure. Treat that spending as the cost of winning a customer and CAC rises while COGS falls. Treat it as the cost of delivering the product and the reverse happens. Organizations genuinely differ here, and some change their minds. Two metrics this KPI group ranks separately can be pushed against each other by a policy memo, with nothing changing in the plant or the data centre.

The third tension is with CSAT and NPS, the two metrics this KPI group ranks first and second, and behind them Churn Rate at fourth. The usual paths to a lower cost of goods are a cheaper component, a thinner support roster, a slower freight tier, a contract manufacturer with a lower quoted price and a longer lead time. Each lands on satisfaction first and on churn much later. The asymmetry is in the timing: the saving books in the period it is taken, and the consequence arrives across the following year in a metric owned by a different team. A cost program that beats its target while pushing CSAT and NPS down has lost on this KPI group's own terms, and the income statement will not say so for several quarters.

CLTV, ranked third, is the metric that settles those arguments, but only in the version that nets out cost of delivery. Plenty of teams compute lifetime value on revenue, and on that version every trade-off above is invisible, because a customer who buys more at a worse margin simply looks better. Computed on gross profit, CLTV contains this KPI, and what it costs to serve a customer becomes part of what that customer is worth. Which version a team uses decides whether COGS is already embedded inside the group's third-ranked metric or floating on its own, far down the ranking, where nobody is looking at it.

Measuring Cost of Goods Sold (COGS) for Product in Practice

The canonical formula is a sum of direct costs for goods produced and sold, and every word in it is a policy choice before it is a measurement. This is an accounting figure first and a KPI second, which makes it unusual among product metrics: you do not instrument it, you inherit it. The controller's decisions determine what the number is, and a product team that wants to manage it has to understand those decisions before it can read the series.

The Direct Cost Boundary. The word direct is doing enormous work. Raw materials and the labour that touches the product are uncontroversial. After that it gets political. Quality inspection, production supervision, maintenance, plant utilities, warehouse handling before shipment, the engineer who supports a line part time, the cloud infrastructure a digital product runs on, the third-party licence embedded in what is sold. Each of those is direct in some organizations and overhead in others, and both treatments can be defensible. The line is documented in the costing standard rather than in the ledger, and it sometimes differs between product lines inside one company, because two businesses were acquired with two conventions and never reconciled. Find the document that defines the boundary before comparing anything. If nobody can produce it, that is the first finding.

Absorption Versus Variable Costing. Under absorption costing, fixed manufacturing overhead attaches to units and travels with them into inventory, so it hits the income statement only when the unit is sold. Under variable costing, only variable production costs attach and fixed overhead is expensed in the period. The same company, the same quarter, the same physical activity, two materially different figures. Neither is wrong. External reporting generally requires absorption, while plenty of internal management reporting runs on variable costing because it answers the decision question better. A product team may therefore be reading a different number than the one in the filed accounts, with the same name on it.

Inventory Valuation Method. First in first out, last in first out, and weighted average cost assign different acquisition costs to the units that happen to be sold. When input prices are stable they converge and nobody argues. The gap opens when prices move, which is the moment somebody asks for a benchmark comparison. Under rising prices, last in first out pushes the newest and most expensive purchases into cost of goods and leaves older cheaper cost in inventory, while first in first out does the reverse. The difference can be large enough to swamp any real operational improvement. Two companies with identical supply chains, identical volumes, and different valuation methods report different costs, and the comparison is measuring the policy.

Standard Cost Versus Actual Cost. Most manufacturers run on standard costs: a pre-set cost per unit, rolled periodically, with the difference from actual captured as variances. Where those variances land decides what the line means. Fully absorbed, it approximates actual cost. Held in a variance account and released selectively, it is closer to a plan than to reality, and the gap piles up somewhere. Ask when the standard was last rolled, because one set before a supplier renegotiation or a tariff change is stale, and every unit costed against it carries an error that surfaces as a variance rather than as a product cost. Ask also whether a roll landed mid-period, since that puts a step in the series nobody in operations caused.

Freight In, Duties, and Inbound Logistics. Inbound freight, customs duties, tariffs, and port handling are part of the landed cost of the material. Whether they reach the product cost depends on whether purchasing pushes landed cost down to the item or books it centrally as a period expense. The second is common because it is easier, and it makes imported products look cheaper than they are while a lump of unallocated freight sits in overhead. A company that books duty centrally sees no product cost movement when tariffs change, which is not a sign that nothing happened.

Scrap, Rework, Yield Loss, and Idle Capacity. Scrapped material, reworked units, and yield that never materializes are real costs of producing the good units. Normal levels usually belong in the product cost. Abnormal levels, and the cost of capacity that sat idle, are generally expensed in the period rather than capitalized into inventory. That word normal is a judgement, and it is one of the easier places for a cost problem to hide: a quietly raised normal-scrap assumption moves cost out of the period and into inventory without anyone editing a number that has a name. When the cost line improves while operations reports no yield improvement, look here first.

Allocated Labour and Overhead. Whatever overhead is absorbed gets allocated by a base that somebody chose: labour hours, machine hours, material value, units, floor space. The base decides which products look expensive. Allocate on labour hours in a plant that has automated heavily and the remaining labour-intensive product carries overhead it did not cause. Nothing about the product changed, the base became obsolete. Treat a product-level cost built on an unexamined base as an estimate rather than a fact.

Cost of Goods Sold Versus Cost of Goods Produced. These are different quantities and they agree only when inventory is flat. Produce more than you sell and the excess production cost capitalizes into inventory, lowering the cost recognized this period. Sell from stock and the reverse happens. That is the most consequential thing to know before setting a target here, because a team can improve the reported figure by building inventory, which burns cash and creates obsolescence risk. The improvement is real in the accounts and harmful in the business. Read the cost line beside inventory movement, or read cost of goods produced instead when the question is production efficiency.

Timing. The cost recognized is the cost of the units sold, not the units made this period. A sourcing improvement reaches the income statement only as the cheaper units work through inventory and out the door, a lag of a quarter or more for a business holding several months of stock. Teams regularly declare a savings program failed because the line did not move on schedule, when the units carrying the saving had not been sold yet.

Transfer Pricing and Intercompany Margin. In a group with several legal entities, the manufacturing entity sells to the distributing entity at a transfer price set for tax and regulatory reasons. At the entity level, the buyer's cost of goods includes the seller's profit margin, and the consolidated accounts eliminate it. The same product carries different costs depending on which set of books you read, and an entity-level product cost is partly a tax position. Confirm whether a figure is consolidated before comparing it outward.

Currency Translation. Imported inputs are bought in one currency and reported in another, and the rate applied may be the transaction rate, an average rate, or a standard rate fixed at budget time. Hedging results may sit inside the cost line or outside it. For a product with a heavily imported bill of materials, currency can move reported cost more in a quarter than a sourcing program will in a year, and none of it is operational.

Why the Product-Level Figure Is the Weaker One. A company-level cost of goods falls out of the accounts. A product-level cost needs the whole apparatus above: a boundary definition, an allocation base, a landed-cost push, a scrap assumption, a standard roll. Each is an internal decision, which is why a product-level figure is always less reliable than the company-level one, and why it belongs in relative comparisons inside one costing regime rather than in comparisons pointed outward.

The data sits in several systems. The general ledger holds the line as reported. The item master and bill of materials hold the recipe, and the standard cost roll holds the assumed cost per unit. The inventory subledger holds the movements that reconcile production to sales. Purchasing holds landed cost, including freight and duty that may or may not have been pushed down. Payroll holds whatever labour was treated as direct, and the sales system holds the units sold. The join that breaks most often is between the selling hierarchy and the costing hierarchy, because the product a commercial team reports on is usually a family of items with separate cost records and a hand-maintained link. Test that link, then tie the sum of product-level costs back to the ledger line. If it does not tie, the gap is where the unallocated cost went.

Segmentation that changes the answer: by item, since an average across a product family hides a loss-making member; by plant or contract manufacturer, because two sites making the same item rarely cost the same; by channel, since direct, distributor, and retail orders carry different packaging, freight, and handling; and by make versus buy, because a purchased finished good has a clean cost while a manufactured one carries every allocation described above.

A short list of the traps that most often distort the line without anyone noticing:

• Supplier rebates and volume discounts booked centrally at year end and never pushed back to the items, so unit costs are overstated all year and corrected in one month.
• Returns processed with the revenue reversed and the cost not reversed, or reversed at a different cost than the original sale carried.
• Inventory write-downs and obsolescence provisions landing in one period, producing a spike that reads as a cost problem when it is a forecasting problem.
• Freight and duty accruals estimated monthly and trued up at quarter end, which makes the closing month of every quarter look different from the others.
• Intercompany transactions not fully eliminated in a management report, leaving a phantom margin inside the product cost.

The practical conclusion: this metric is interpretable only alongside the policy that produced it. Hold the policy stable, document it, and read movements against volume and inventory rather than on their own. When the policy changes, restate the history or annotate the break, because an unannotated policy change is indistinguishable in the series from a real operational result, and it is usually larger.

Common Pitfalls

Many organizations overlook the nuances of COGS, leading to misinterpretations that can distort financial reporting and decision-making.

• Failing to include all direct costs in COGS calculations can lead to understated expenses. This oversight creates a misleading picture of profitability and can affect strategic decisions.
• Neglecting to regularly review supplier contracts may result in outdated pricing assumptions. As material costs fluctuate, organizations must adjust their COGS calculations to reflect current market conditions.
• Not accounting for inventory write-offs can inflate COGS figures. Companies should regularly assess inventory levels and adjust COGS to avoid misrepresenting financial health.
• Overlooking the impact of production inefficiencies can lead to inflated COGS. Continuous process improvement initiatives are essential for identifying and mitigating waste in production.

Improvement Levers

Enhancing COGS requires a multifaceted approach focused on cost reduction and efficiency improvements.

• Implement just-in-time inventory systems to minimize holding costs. This approach reduces waste and ensures that materials are available only when needed, improving cash flow.
• Negotiate better terms with suppliers to lower material costs. Stronger relationships and bulk purchasing can lead to significant savings, positively impacting COGS.
• Invest in automation technologies to streamline production processes. Automation can reduce labor costs and improve consistency, ultimately lowering COGS.
• Conduct regular variance analysis to identify discrepancies in expected versus actual costs. This practice helps organizations pinpoint areas for improvement and adjust strategies accordingly.

KPI Depot is trusted by consulting, strategy, finance, and analytics teams at leading organizations worldwide, including those listed below.

Cost of Goods Sold (COGS) for Product Benchmarks

We have 9 relevant benchmarks in our benchmarks database.

Browse the Top Benchmarked KPIs in Product Management

Reading the Benchmarks for Cost of Goods Sold (COGS) for Product

Nine source records are tracked for this page, which is the largest set on any page in this batch. Before a customer reads anything into that, one fact has to land: all nine rows come from a single organization, New York University. There is no second reading here. Nine rows look like corroboration and are not, because rows that share an organization share a compilation method, a data source, a classification scheme, and a set of judgement calls about what belongs in the population. They can only agree with each other. What the set actually holds is one reading, cut nine ways.

The cuts are by industry. Seven of the rows carry a sector label, covering electronics, household products, telecommunications services, computers and peripherals, pharmaceuticals, food wholesalers, and apparel. The remaining two are whole-market rows, one for the total market and one for the total market with financial companies excluded. Those two are not independent of the other seven, they contain them. Anyone noticing that the sector rows sit either side of the market rows has noticed arithmetic, not evidence. The distinction between the two market rows is a definitional statement in disguise: financial companies mostly do not have a cost of goods in any recognizable sense, so whether they sit in the population changes a market-wide figure without anything in the real economy having changed.

Every row records the same statement type, an average, and every row records the same time period, a single month at the start of a recent year. Two consequences follow. First, there is no distributional read anywhere in this set. No median, no quartile, no best-in-class figure. For a quantity as skewed as cost of goods, where a handful of very large companies in any sector can carry most of the aggregate, a mean is the statistic most sensitive to the largest members of the population, and there is nothing here to check it against. Second, there is no trend. One compilation date means one point in time. That matters more for this metric than for most, because the thing that moves cost of goods hardest is input prices, and input prices move over exactly the horizon a single snapshot cannot show.

The averages themselves need a qualification that the record does not supply. There is more than one way to average a cost figure across companies. You can take each company's figure and average the figures, which gives every company equal weight regardless of size. Or you can sum the numerators across the population and sum the denominators and divide once, which lets the largest companies dominate. Those two produce different answers from identical underlying data, sometimes very different ones in sectors with a few giants and a long tail. The metric type field says average and stops there, so a customer cannot tell which was done.

Sample sizes vary enormously across the nine rows. The pharmaceutical row rests on hundreds of companies. The food wholesalers row rests on a number small enough that a couple of unusual companies would visibly set the result, and the telecommunications services row is not much better. The whole-market rows rest on thousands. This is a case where two rows presented in the same table, in the same format, with the same label, carry entirely different weight, and nothing in the presentation flags it. A thin sector row is not a benchmark, it is a description of a handful of specific companies that happen to share a classification code.

Two dimensions are blank on all nine rows, and a blank dimension is information. The company size field is empty throughout, so there is no way to tell whether this population is dominated by large listed companies or spread across sizes. The population field says firms and the industry labels are the sector classifications used to sort listed companies, which strongly suggests the population is public filers and therefore excludes the private companies that make up most manufacturers in most economies. If a customer runs a private mid-market business, the question of whether they belong in this population is not answered by the record, and the honest answer is probably not.

The formula text is the second blank, and it is the consequential one. Not one of the nine rows records what was divided by what, or whether a ratio was reported at all. The canonical formula for this KPI is a sum of direct costs for goods produced and sold, which is an absolute amount of money. Most published figures that travel under a related name are not absolute amounts. They are cost of goods expressed as a share of revenue, or they are a gross margin percentage from which a reader is expected to infer the cost share by subtraction. Those are three different quantities. An absolute cost scales with the size of the company and means nothing without a second number beside it. A cost-to-revenue ratio is comparable across companies but moves when selling prices move, so a company that raised prices without touching its supply chain improves on it. A gross margin complement inherits whatever the reporting company chose to put above the gross profit line, which is not standardized. With the formula field empty, a customer cannot tell which of the three a row contains, and cannot reconcile it to the canonical formula without going back to the source and reading the source's own methodology. That reconciliation is what source-attributed data is for, and it is the work that disappears when a figure is copied into a slide.

Geography is recorded as the United States on all nine rows, consistently, which is a strength in one way and a hard limit in another. It is consistent, so the rows are at least comparable to each other on that dimension. But it means there is no international read available here at all, and for this metric the border matters more than usual. United States reporting permits the last in, first out inventory method, which much of the rest of the world does not. When input prices are moving, that single difference changes the reported cost of goods for two otherwise identical companies. A customer benchmarking a European or Asian operation against this set is comparing across an accounting regime boundary, and nothing in the row will warn them.

There is one further gap that no field can express. This KPI is scoped to a product. Every one of these rows is scoped to a company. A company-level cost of goods falls out of the accounts with no allocation required. A product-level figure requires somebody to decide how shared labour, shared plant, shared overhead, and shared logistics are split across products, and that decision is made inside each company by people applying their own conventions. So the two are not one measurement at different zoom levels. The product-level figure carries a layer of judgement the company-level figure does not. These rows can describe the neighbourhood a company-wide number should sit in. They cannot validate a product-level number, and a customer who uses them that way will conclude that a product is expensive or cheap when what they have actually measured is their own allocation policy.

The short version for anyone about to trust an external figure for this metric. Establish whether it is an absolute cost or a ratio, and if a ratio, what sat in the denominator. Establish where the source drew the line between direct cost and overhead. Establish the inventory method and whether prices were moving during the period. Establish whether the average weights companies equally or by size. A figure that answers none of those is not comparable to yours, and nine figures that all come from one organization and answer none of those are not nine figures.

OKRs That Use Cost of Goods Sold (COGS) for Product

No key result in the Product Management KPI group's OKR material names this KPI, which is consistent with where the group ranks it. Two of the group's worked objectives still need it, and in both cases it belongs as a guardrail rather than as a headline.

Drive sustainable revenue growth through focused product expansion and monetization is the obvious host. That objective runs on Revenue Growth, Monthly Recurring Revenue (MRR), Average Revenue Per User (ARPU), and Customer Acquisition Cost (CAC), and every one of those is a revenue or acquisition measure. Nothing in the set notices what the expansion costs to deliver. A product portfolio can hit all four while shifting its mix toward higher-priced items that carry disproportionately more cost, and the objective's own rationale, which talks about revenue quality and monetization depth, is not satisfied by that outcome. The key result that fixes it is directional: hold or reduce unit cost of delivery on the product lines the expansion actually runs through, measured on a stable costing policy, while ARPU rises. Stated that way it cannot be met by mix shift alone. The group's own guidance about pairing acquisition cost with long-term value points the same direction, since lifetime value computed on gross profit is the version that contains this metric and the version computed on revenue is not.

Create exceptional product experiences that boost user retention and satisfaction uses it differently, as a constraint. That objective pairs Churn Rate, Customer Satisfaction Score (CSAT), Product Adoption Rate, and Retention Cost. The presence of Retention Cost is the interesting part, because retention cost and cost of delivery are adjacent and the line between them is drawn by policy rather than by nature. Support labour, replacement units, goodwill credits, and expedited shipping can sit in either bucket. A team that improves Retention Cost by reclassifying spending into cost of goods has improved nothing. So the useful key result is a joint one: reduce the combined cost of serving and retaining a customer rather than either line on its own, with the classification frozen for the duration of the cycle.

Two cautions apply to any target set on this metric, and both come from how the number is produced rather than from the objective it serves. First, it improves when a company produces more than it sells, because the excess cost capitalizes into inventory. That makes it possible to hit a cost target by building stock, which costs cash and creates write-down risk later. Pair it with an inventory movement measure or the target is gameable by accident. Second, it is a closed accounting output rather than an instrumented one, so it responds a quarter or more after the decision that moved it. Set the cycle accordingly, and judge the sourcing or design work on its own evidence in the meantime.

See OKR Examples for Product Management

Free Whitepaper

Want to achieve performance excellence in Product Management? Download our in-depth whitepaper: Definitive Guide to Product Management KPIs.

Download the Free Guide

KPI Categories

This KPI is associated with the following categories and industries in our KPI database: