The short answer

You find margin leaks by SKU by allocating the real production costs (changeover, rework, scrap, freight) down to each product, not by trusting a blended gross margin. Do that and the pattern is almost always the same: a small share of SKUs earns nearly all the profit, and a long tail quietly destroys it. Across most manufacturers, cost of poor quality alone runs 5 to 30 percent of sales, and when you rank products by true net profitability, the bottom cohort can erase 50 to 100 percent of the profit the winners generate. The leak is invisible because standard costing spreads those production costs evenly across the catalogue, so the losers hide behind the winners on every report you look at. Here is how a CFO finds the leak in a manufacturing catalogue, and why two SKUs with the same gross margin can have very different real economics.

Why gross margin lies at the SKU level

Gross margin uses standard cost, which is material plus a standard labour and overhead rate. That rate is an average. It assumes a short-run specialty grade absorbs the same setup, scrap, and handling as a high-volume workhorse SKU, which is never true. A product that needs a two-hour changeover every time you run it, throws off more rework, and ships in small pallet-inefficient loads carries far more real cost per unit than its standard cost admits. The overhead pool papers over all of it. So on the P&L, both products post the same healthy gross margin, and one of them is losing money on every unit sold.

The reason this survives audit and month-end close is that standard costing is not wrong on the total. Add up all the units and the plant-level number reconciles to the general ledger. What breaks is the split. Overhead is pooled and then reapportioned per machine hour or per unit, so the pool is correct in aggregate but arbitrary at the line-item level. Every SKU that consumes more than its average share is being subsidised by every SKU that consumes less, and the P&L is structurally incapable of showing you which is which. A CFO looking only at the finished income statement has no way to see the transfer happening inside the overhead pool.

The three production costs that hide inside overhead

Three cost buckets do most of the damage. Each one is real, each one is driven by product choices, and each one is smeared flat by the standard rate.

  • Changeover and lost capacity. Every switch between products consumes downtime, quality-adjustment waste, and setup labour. A line that loses even one changeover hour a day gives up over 15 days of productive capacity a year; an eight-hour line losing two hours to changeovers surrenders 25 percent of its productive time. That cost belongs to the short-run, high-variety SKUs that force the switches, but standard costing charges it to everyone. The workhorse that runs for three shifts without a stop is paying for the specialty that triggers a setup every hundred units.
  • Rework and scrap. Scrap and rework range from under 1 percent for top performers to over 2 percent of revenue for the bottom cohort, and the true figure is typically three to five times the visible scrap number once you count re-inspection, expedite, and lost capacity. It is never spread evenly across SKUs. A few finicky grades generate most of it. The specialty product with the tight tolerance that fails first-pass inspection twice as often is the one draining the quality budget, and yet the scrap cost lands on the whole catalogue at the standard rate.
  • Freight and cost to serve. Pallet inefficiency, small-drop deliveries, and expedited shipments load different costs onto different products. A low-density SKU that fills a truck by volume before weight costs more to move per unit than a dense one, regardless of what the freight-per-tonne average says. Add customer behaviour to the mix, because the same SKU sold in full-truckload quantities and in single-carton drops carries wildly different logistics cost, and standard costing sees none of it.

There is a fourth bucket worth naming, because it is the one CFOs most often forget: working capital. A slow-moving specialty grade that sits in finished-goods inventory for four months ties up cash and warehouse space that the fast mover would have turned six times. That carrying cost rarely appears in any product margin at all, yet it is a genuine drag on return on the money invested in that SKU.

Two SKUs, same gross margin, one loses money

Here is the archetype for a mid-market Indian auto-components maker. Both SKUs carry material at 60 percent of price and post the same 40 percent standard gross margin. The difference only appears when you allocate the real production costs down to the unit.

Line item (per unit) SKU A (workhorse) SKU B (specialty)
Selling price Rs 100 Rs 100
Material cost Rs 60 Rs 60
Standard gross margin Rs 40 (40%) Rs 40 (40%)
Changeover cost allocated Rs 2 Rs 14
Rework and scrap Rs 1 Rs 9
Freight and cost to serve Rs 3 Rs 11
True (pocket) margin Rs 34 (34%) Rs 6 (6%)

On paper they are identical. In reality SKU A funds the plant and SKU B barely covers its own production cost, and if volumes shift toward B, the blended margin falls even as the top line grows. This is cross-subsidy: the workhorses are paying for the specialties, and no report on your CFO's desk shows it.

The dangerous part is what the standard view tells the commercial team to do. Because SKU B looks like a healthy 40 percent margin product, the sales team is rewarded for winning more of it, the plant is told to make room for it, and marketing prices the next tender off the same false floor. The business grows revenue and shrinks profit at the same time, and everyone is following the incentives the numbers gave them. Getting the pocket margin on the table is not an accounting nicety. It is the difference between a growth plan that compounds and one that quietly eats the base.

Ready to fix your data foundation?

Book a free diagnostic call and find out where your stack stands.

Book a Call

How to find your number without a new ERP

You do not need a new ERP to see this. You need three data sources reconciled: the production log (run times, changeovers, scrap by SKU), the costing sheet (standard cost build-up), and the dispatch and freight records (loads, drops, expedite flags). Almost every plant already generates all three; they simply live in different systems and never get joined. Here is the model we build, step by step.

  1. Anchor on the SKU. Pick the product code that appears in all three systems and make it the join key. If the production log uses a machine code and the costing sheet uses a finished-goods code, build a one-time mapping table. This unglamorous step is where most internal attempts stall, so it is worth doing carefully.
  2. Pull the causal drivers, not the averages. For each SKU over a representative period, count the number of changeovers it triggered, the setup minutes each one consumed, its first-pass yield and rework hours, and its dispatch profile (average drop size, pallet fill, expedite frequency). These are the drivers that make one product cost more than another.
  3. Rate each driver. Convert a changeover minute into rupees using the fully loaded line cost per minute, including lost contribution during the stop. Convert a scrap unit into rupees at material plus the labour already sunk into it. Convert a truck slot into rupees per pallet position. Now every driver has a price.
  4. Allocate by cause, not by average. Multiply each SKU's driver count by its rate and assign the result to that SKU alone. A product that caused 40 changeovers carries 40 changeovers of cost; a product that ran clean carries none. This is the single move standard costing refuses to make.
  5. Compute pocket margin and rank. Pocket margin is selling price minus material minus the production costs actually caused by that product. Rank the SKUs from most to least profitable and you get your own version of the whale curve, where cumulative profit rises steeply on the winners, peaks, and then falls as the loss-making tail drags it back down.

The industry figures here are ranges, not promises. Your losers could be a handful of grades or a third of the catalogue. The only way to know is to allocate your own costs. That reconciliation is the first thing we build. Our AI Stack Audit x-rays your production, costing, and dispatch data, allocates the hidden costs down to each SKU, and shows you exactly which products are being cross-subsidised, as a specific rupee-per-unit figure. For how we work with plants and manufacturers, see our manufacturing practice page.

What to do once the losers are visible

Finding the leak is the easy half. Acting on it is a commercial decision, and once the pocket margin is on the table the options are clear. Reprice the loss-makers to a floor that reflects their true cost, and let the customers who only wanted them at the subsidised price walk. Run them in longer batches so the changeover cost per unit falls, which often converts a marginal SKU into a profitable one without touching the price. Raise the minimum order quantity so small, freight-heavy drops stop bleeding the logistics budget. Or rationalise the ones that will never pay, freeing plant capacity, working capital, and management attention for the workhorses that actually fund the business.

None of these moves is possible while the loss is buried in a pooled overhead rate. The value of the pocket-margin model is not the spreadsheet. It is that it turns an invisible structural transfer into a short list of specific, defensible commercial decisions, each with a rupee figure attached, that a CFO can take to the board.

Key takeaways

  • Gross margin at the SKU level lies because standard costing spreads changeover, rework, and freight evenly across the catalogue instead of to the products that cause them.
  • Two SKUs with identical gross margin can have very different true margins once you allocate real production costs; the workhorses quietly cross-subsidise the specialties.
  • Cost of poor quality alone runs 5 to 30 percent of sales, and the bottom cohort of products can erase 50 to 100 percent of the profit the winners earn.
  • You find the leak by reconciling three sources you already have (production log, costing sheet, dispatch records) into a pocket-margin-per-SKU ranking. No new ERP required.
  • Once the losers are visible, the fix is a pricing, batch-size, minimum-order, or rationalisation decision, each backed by a specific rupee-per-unit number.

Frequently asked questions

How do I find where we are losing margin by SKU?

Allocate the real production costs (changeover, rework, scrap, freight) down to each product instead of using a blended overhead rate, then rank SKUs by pocket margin (price minus material minus the costs that product actually causes). The loss-makers surface immediately. It needs three data sources reconciled: the production log, the costing sheet, and dispatch records.

Two products have the same gross margin, so why is one unprofitable?

Because gross margin uses an averaged standard cost that ignores what each product actually consumes. A specialty SKU that forces long changeovers, throws off more rework, and ships in inefficient loads carries far higher real cost than its standard cost admits, so its true margin can be a fraction of an identical-looking workhorse SKU.

What is true product cost or pocket margin?

Pocket margin is the selling price minus the material cost minus the production costs genuinely caused by that specific product: its share of changeover, rework and scrap, and freight and cost to serve. It is the number that tells you whether a SKU makes or loses money, which standard gross margin cannot.

How do I stop cross-subsidising loss-making SKUs?

First make the subsidy visible with a pocket-margin-per-SKU ranking, because you cannot act on what you cannot see. Then the loss-makers become a pricing, minimum-order-quantity, or SKU-rationalisation decision. Reprice them, run them in longer batches to cut changeover cost, or drop the ones that will never pay. The decision is easy once the number is on the table.

Do we need a new ERP or costing system to do this?

No. The three inputs (production log, costing sheet, dispatch records) almost always already exist in systems you own. The work is joining them on a common SKU key and rating the causal drivers, not buying software. We build the reconciliation on top of your current data and hand back a ranking you can act on.