Method · 7 min read

Scenario planning: putting a number on assumptions you can state

Published by Variance Works - Updated 8 September 2026

"What if we raised prices three percent on the National brand?" is a question anyone can ask in a meeting. Answering it usually means a spreadsheet nobody can audit a week later. It should mean an assumption, a real period, and a number that adds up.

The question it answers

Scenario planning here starts from something that actually happened: one real period of your own export, read as it is. You then state changes - a price move, a discount, a cost, a volume, a target mix - and read the bridge from what happened to what you assumed.

The result is not a single number. It is a bridge with two readings: what each assumption moved, one line per assumption, and how the total splits into price, volume and mix. Those two do not overlap, and the screen does not pretend otherwise - a price rise on one group produces a price effect and a mix effect, and splitting the second between assumptions would need an attribution convention nobody has agreed to.

Why this is not a forecast

This is the most important paragraph on the page, so it is near the top rather than in a footnote.

A forecast is an opinion about the future. This is arithmetic on assumptions you state, applied to a period that already exists. It holds no view on what will happen, assigns no probability, and carries no date. Change nothing, and the result is the period itself.

The distinction matters when the number leaves the room. A figure labelled "forecast" gets planned against; a figure labelled "what a three percent price rise would have been worth on March" gets challenged on its assumption, which is exactly where the conversation belongs. The wording follows the calculation everywhere - on the screen, in the review pack, and in the printed document.

The levers you can state

Five levers, and one target mix:

• Net price - what you actually charge, as a percentage change.
• Gross price - list price, before discount.
• Discount rate - in points, not percent. A rate moves in points; treating it as a percentage of itself is a common and expensive mistake.
• Unit cost - the margin side of the same question.
• Volume - units, at unchanged unit prices.
• A target mix - a share you want each group to reach on one chosen dimension. Total volume is preserved; only its split changes.

Each lever applies to the whole portfolio, to a group, or to a single product. And a lever whose data is missing from your file is not offered at all - a tool that proposes something and then refuses it is the worst way to inform someone.

Why the order is shown

Assumptions are applied in a fixed order: gross price, discount rate, net price, unit cost, volume, then target mix. That order is written on the screen.

It is shown for two reasons. The first: it is the attribution convention behind the audit lines. For independent multipliers such as a portfolio price and a portfolio volume, it decides how the gap is split without changing the gap itself. Start from 100, raise price by 10% and volume by 20%: you land on 132 either way, because 100 x 1.10 x 1.20 and 100 x 1.20 x 1.10 are the same number. No assumption moves another.

What the order decides is how that +32 is split between the two lines:

• price first: price +10, then volume +22;
• volume first: volume +20, then price +12.

Both add up to +32. The product always applies the first split, so that two runs of the same scenario read the same way and a line can be checked twice. You have a right to know which one you are reading.

The second reason is the target mix, and it is why the order belongs to the definition of the scenario itself. A target mix does not multiply: it reallocates quantities between groups at constant total volume, and it is applied last, after every assumption. Applied after a targeted volume rise, it redistributes a total that assumption has already reshaped; applied before, it redistributes the original one. Those two do not commute.

On the sample file: raising Kitchen volumes by 20% and then targeting a 60/40 brand mix gives 4,123.64 of revenue. Applying that same mix first, then the same rise, gives 4,200. Neither is wrong - they are two different scenarios. The product applies the mix last, always, so that "this scenario" names one thing.

Each assumption also gets its own audit line: how many products it touched, and what it moved in units, revenue and cost. Each line measures the move from the state the previous assumptions left, and the lines add up to the simulated gap exactly. The bridge below answers a different question: it splits the same total into price, volume and mix following appendix A, and those effects are not attributed to individual assumptions.

What it will not do

No elasticity. A price assumption never moves volumes, and a volume assumption never moves unit prices. Estimating how a market would react needs a model that neither the app nor your export contains, and inventing one would produce a confident number resting on nothing.

No contradictory assumptions. Fixing a product's net price and, in the same scenario, its gross price or its discount, describes the same number twice. Nothing says which should win, so the tool refuses instead of choosing silently.

No impossible result. A discount assumption that would push a product past a hundred percent off is refused while you type, naming the product, its rate today, the rate it would reach, and by how much the assumption must come down. Nothing is capped or corrected on your behalf.

Questions

Can I keep several scenarios?

Yes, in a monthly review. Each one is named, keeps its own assumptions, and can be added to the review pack as its own section. Two scenarios on the same month sit side by side rather than overwriting each other.

Can I re-run one later?

Yes. A saved scenario keeps its assumptions in full, so it re-runs exactly as it was stated - not as the editor happens to look at that moment.

What does the review pack say about a scenario?

It carries the assumptions in the section itself, plus the sentence that it is not a forecast. A simulated figure without its assumptions is not verifiable, and a document read in a committee is exactly where that matters.

Does my file leave the browser?

No. Everything runs in the page, on your machine. How to verify that in two minutes.