Profit Margin Optimization in FMCG: The Hidden Profit Between Your SKUs

Search for profit margin optimization and the questions Google lists under “people also ask” are almost all arithmetic.

• Is 20% margin the same as 25% markup ?
• Is a 50% margin good?
• What is the formula in Excel? 

Fair questions, and for a business selling one product at one price, the arithmetic is most of the job.

An FMCG manufacturer with forty SKUs across three pack formats and four retail customers has a different problem. Move the 500ml single up in price and some of its volume walks to your own 1L bottle, some walks to a competitor, and some leaves the category. The margin on the SKU you touched improves. The margin on the portfolio can still fall.

That gap, between margin as a calculation and margin as a portfolio decision, is what this article is about. The calculations come first, briefly, because they get asked constantly and take five minutes to settle. Everything after that is the part the calculations cannot see: what actually moves when you change one price, why spreadsheets stop working at exactly that point, and what optimization means when a vendor finally says the word in a meeting.

What profit margin optimization means, and where the usual definition stops

Profit margin optimization is the process of setting prices and managing costs so that a business earns the most profit its sales can produce. In practice, that means adjusting prices, costs, and product mix together, then checking the effect on both the margin percentage and the absolute profit behind it.

As definitions go, that one is fine. It is also built on a quiet assumption: one product, one cost, one price. Change the price, margin moves, done.

FMCG breaks the assumption three ways. First, no product exists alone. It sits in a range where packs and sizes compete with each other for the same shopper. Second, the shelf is shared with competitors and private labels, so a price move is also a move in relative price, which is the price shoppers actually respond to. Third, the manufacturer does not set the shelf price. The retailer does, and every list change passes through negotiations, trade terms, and promotion plans before it reaches a shopper.

So in FMCG the honest definition is longer: profit margin optimization is finding the set of prices, across the whole portfolio, that delivers the most profit while respecting the commercial rules you cannot break. That version has consequences, and the rest of this article works through them.

The arithmetic, settled

The basics get asked every day and take a few minutes, so here they are in one place.

Margin and markup are not the same thing

Margin is profit as a share of the price. Markup is profit as a share of the cost. Same money, different denominator, so the percentages never match.

Markup on cost	Equivalent margin on price
12%	10.7%
15%	13.0%
20%	16.7%
25%	20.0%
33%	24.8%
42.9%	30.0%
50%	33.3%
100%	50.0%

So no, a 20% margin is not the same as a 25% markup, but they do describe the same transaction. A product bought at 1.00 and sold at 1.25 carries a 25% markup and a 20% margin.

The formulas

• Unit margin = (price - cost) / price
• Markup = (price - cost) / cost
• Gross margin = (revenue - cost of goods sold) / revenue
• Operating margin = operating profit / revenue
• Net margin = net profit / revenue

In Excel, unit margin is =(price-cost)/price formatted as a percentage. If a model divides by cost instead, it is calculating markup and reporting it as margin, which is one of the most common quiet errors in commercial workbooks.

What counts as a good margin

There is no honest single number. A margin that would be healthy in premium spirits is a crisis in bottled water. The often repeated rule of thumb that 10% net is average, 20% is strong, and 5% is weak describes companies in general, not FMCG categories, where gross margins vary widely by category, channel, and pack while net margins carry the weight of trade spend. The useful comparison is not a universal benchmark. It is your own category, your own channel, and your own pack against its direct substitutes.

That settles the arithmetic. It is also where most public advice on this topic stops, and where the actual work begins.

One price change is never one number

Price is the most powerful profit lever a manufacturer holds. The classic McKinsey analysis by Marn and Rosiello, published in Harvard Business Review, found that a 1% improvement in price, with no loss of volume, lifts operating profit for the average company, three to four times the effect of a proportionate gain in volume. That leverage is exactly why getting a price move wrong across a portfolio is expensive.

Here is what actually moves when one SKU's price changes:

• Volume on that SKU moves. Own-price elasticity: raise the price and some buyers stop buying it.
• Volume on your other SKUs moves. Cross-price elasticity: some of those buyers switch to your other packs and sizes, so part of the "loss" is really internal migration.
• Competitor volume moves. Whatever does not stay inside your range goes across the aisle, and competitors may reprice in response, which changes the relative price you thought you had set.
• The mix shifts. This is the quiet one, and the most dangerous.

The mix effect deserves numbers, because it is genuinely counterintuitive: blended margin can fall while every individual SKU's margin holds or improves.

Two SKUs. A premium 500ml selling 100 units at 2.00 with a 1.00 cost, a 50% margin. A value 1L selling 100 units at 3.00 with a 2.10 cost, a 30% margin. Together: 500 in revenue, 190 in profit, a blended margin of 38%.

Now raise the 500ml to 2.20. Its unit margin improves to 54.5%. But its volume drops to 60 units, and of the 40 lost, 25 switch to the 1L while 15 leave your range entirely.

	Befoe	After
500ml margin	50.0%	54.5%
1L margin	30.0%	30.0%
Revenue	500	507
Profit	190	184.5
Blended margin	38.0%	36.4%

Every line a category manager checks looks fine or better. The premium SKU's margin went up. The value SKU's margin held. Revenue even grew. And the business made less money, because volume migrated from a unit earning 1.00 to a unit earning 0.90. Scale that pattern across forty SKUs and it will not announce itself in any single row of the report.

This is what revenue growth management (RGM) teams mean by winners and losers: a price change produces both, inside your own range and outside it. The question that matters is never what happens to this SKU. It is what happens to the whole board.

Where the spreadsheet stops working

None of this is an argument against Excel. Most good RGM analysis of the last twenty years has lived in workbooks, and an experienced analyst with a well built model beats an expensive platform used badly. The argument is narrower. There is a specific job this shape of tool cannot do, and portfolio pricing is that job.

The scenario space is too large to enumerate

Take a modest exercise: ten SKUs, five candidate price points each. That is 5^10 possible combinations, or 9,765,625 distinct price scenarios. Twenty SKUs at the same five points is roughly 95 trillion. A spreadsheet evaluates the scenarios someone builds by hand, so in practice a team tests five or ten combinations that seem sensible and picks the least uncomfortable one. The other 9.7 million go unexamined, and the best answer is almost certainly among them. Run the exponent yourself before quoting it in a meeting. The point survives any reasonable assumption.

The substitution matrix can be stored but not estimated

Cross-SKU effects live in a matrix: how each SKU's volume responds to each other SKU's price. Forty SKUs means 40 x 39 = 1,560 cross-relationships. Excel will hold that grid without complaint. What it cannot do is fill it in. Estimating elasticities takes statistical modeling on sell-out and panel data, and no analyst holds 1,560 relationships in their head while deciding a price.

This is worth stating plainly because most RGM teams already own the raw material: EPOS sell-out, shipment history, pricing and promotion records, syndicated data from Nielsen or Circana, customer and margin data. The gap is almost never data collection. It is the estimation layer between the data and the decision.

Constraints can be written down but not solved

A workbook happily records the rules. Keep at least a 15% gap between the 500ml and the 1L. Do not touch the returnable format this cycle. Stay inside the agreed promo depth with the key account. What it cannot do is search for the most profitable set of prices that respects all of those rules at once. Excel's Solver handles small, well behaved problems, and on realistic portfolio problems it tends to return a local answer that looks precise without being best.

Nobody can say which version produced the number

The last problem is quieter but familiar to anyone who has inherited a pricing model called v14_FINAL_rev3.xlsx. Field audits of operational spreadsheets, summarized by researcher Raymond Panko, found errors in 94% of the workbooks examined, with roughly one formula cell in twenty containing a mistake. In a pricing model, one broken cell reference does not crash anything. It just changes a recommendation that someone then presents to a commercial director.

This is the gap that simulation platforms and profit margin enhancement software exist to fill. Not a faster spreadsheet, but a different structure: estimated elasticities instead of assumed ones, scenarios evaluated in bulk instead of by hand, constraints solved rather than merely recorded, and one governed model instead of a folder of personal ones.

Four levels of margin decision making

A useful way to place your own team:

Level	The question it answers	Typical tooling
1. Descriptive	What was our margin last quarter?	Reporting, BI dashboards
2. Diagnostic	Why did it move?	Analyst plus workbook
3. Predictive	What happens if we change this price?	Elasticity models, simulation
4. Prescriptive	Which prices best hit our targets within our rules?	Constrained optimization

Most FMCG teams sit between levels one and two and describe themselves as data driven, which is true in the sense that data gets assembled, reconciled, and reported in large quantities. It is also where the time goes. Anaconda's 2020 State of Data Science survey found data professionals spending about 45% of their working time on data preparation alone, and RGM analysts describe the same imbalance in their own words: too many hours assembling last quarter's numbers, too few spent on the decision those numbers were supposed to inform.

Teams making the jump tend to describe the goal identically, moving from descriptive reporting to predictive and prescriptive decisions. The jump is not a bigger spreadsheet. Levels three and four need the two things a workbook cannot supply, an estimated substitution matrix and a solver that respects commercial constraints, which is why teams usually cross this line at the same moment they start evaluating dedicated tools.

What constrained optimization actually means

The word optimization gets used loosely, mostly as a synonym for improvement. In a pricing tool it means something specific, and knowing the specific meaning is the fastest way to tell vendors apart.

An optimization has three parts:

1. An objective. The number to maximize or protect. Portfolio profit, a margin target, share within a segment.
2. Decision variables. The things you are allowed to change, usually list prices on a defined set of SKUs.
3. Constraints. The rules the answer must respect.

The constraints are where FMCG reality lives:

• Price ladder integrity across the price pack architecture: the 1L stays cheaper per liter than the 500ml, the multipack stays cheaper per can than singles
• Pack rules: some formats are off the table this cycle
• Retailer agreements and promotion calendars already committed
• A minimum gap to private label
• Margin floors by brand or category

Here is the part worth internalizing: an unconstrained optimization is commercially useless. Tell a model to maximize margin and nothing else, and it will cheerfully suggest raising the entry pack until it dies, because on paper the portfolio looks more profitable without its least profitable member. Every experienced commercial lead knows why that answer is wrong. The entry pack defends shelf space, feeds the brand, and blocks a competitor from owning the price point. Constraints are how that knowledge gets into the mathematics. An output you can actually take into a retailer meeting is one that already respects the rules the meeting will be conducted under.

Simulation and optimization are related but different, and the difference matters when comparing tools. Simulation answers what happens if we do X. Optimization answers which X we should do. A team needs both: simulation to test the scenarios people propose, optimization to surface the scenarios nobody thought to propose.

Questions to ask before you buy anything

Two honest points before the checklist.

First, software is not the only way to close this gap. Consultancies model pricing too, often well. The difference is the shape of the outcome. A consulting engagement answers this year's pricing question. A platform builds your team's capacity to keep asking. Budget, team size, and how often your prices move should decide which you need, and for some teams the answer is genuinely both: a consultant to build the first model, a platform to keep it alive.

Second, vendor pages blur together. Every dynamic pricing optimization software site promises simulation, AI, and margin uplift in roughly the same words. The questions below separate a portfolio tool from a repricing widget, and a vendor worth talking to will enjoy answering them.

1. Does it model cross-SKU substitution, or only own-price elasticity?
   
   If cannibalization is not in the model, the tool will systematically overstate the benefit of every price increase.
   
2. Where do the elasticities come from, and how often are they re-estimated?
   
   Elasticities estimated once and frozen go stale with every season and every cost shock.
   
3. Can it optimize under constraints, or only simulate scenarios you define?
   
   Ask to see a constraint being set. This question alone eliminates a surprising share of tools.
   
4. Does it model competitor response, or hold competitor prices fixed?
   
   A simulation that assumes the competition stands still flatters every scenario you run.
   
5. What happens when data is patchy?
   
   Off-trade data always has gaps: missing banners, delayed periods, new SKUs with no history. Ask how the model behaves at the edges, not just in the demo.
   
6. Who operates it day to day?
   
   A tool only the vendor's consultants can drive is a consulting engagement with a login screen.

Five mistakes that cost margin

1. Optimizing SKU by SKU and calling it portfolio strategy.
   
   Each decision looks right in isolation. The mix effect from earlier in this article is the bill that arrives later.
   
2. Treating elasticity as a constant.
   
   An elasticity estimated three years ago describes the market of three years ago. Costs, competitors, and shopper stress have all moved since.
   
3. Making the margin percentage the only objective.
   
   Margin is a ratio, and ratios can be improved by shrinking the business. A margin target without a volume or share guardrail invites exactly that.
   
4. Confusing a price increase with a margin increase.
   
   A 4% list increase can arrive as flat net revenue after the retailer negotiates half of it back, promo depth deepens to defend volume, and input costs move. What matters is net revenue realization, not the announced list change.
   
5. Judging promotions on uplift alone.
   
   Uplift counts pantry loading and cannibalized volume as wins and ignores the dip that follows. Mckinsey’s  multi-market trade promotion analysis found 72% of promotions globally do not break even, and its deep dive into US events put the share at roughly two-thirds. Post-promo evaluation against a proper baseline is where those losses become visible, and most teams skip it because the data preparation is painful. Which is the previous section's point wearing different clothes.

Frequently asked questions

What changes when the model can answer back

The visible change is speed. The real change is the shape of the pricing meeting. Instead of an argument about whose workbook is right, the room compares three modeled options, each with its volume, share, and margin consequences on the table and its constraints already respected. Disagreement moves to where it belongs, the assumptions, and stops hiding in the arithmetic.

Smart Value is an RGM decision platform built for FMCG manufacturers. Its pricing module simulates SKU-level price changes across a portfolio, models substitution between packs and brands, and optimizes against margin targets under the commercial constraints teams actually face, from price ladder rules to formats that cannot move this cycle. If your team sits somewhere between level two and level three, a feature list will not tell you much.

See a pricing scenario run on your own SKUs. Book a working session with the Smart Value team and bring one real pricing question.