Get back to your decision.

It compares campaign contribution with a specified no-promotion baseline and explores the implications of your pricing and demand assumptions. Free tools cover common promotion and pricing calculations.

No. Use the sample, the in-product instructions and these explanations to work through the calculation yourself.

It is the sales you expect over the campaign dates without the promotion. Use comparable dates and account for seasonality and availability. The preceding period is not automatically a credible baseline.

No. Contribution subtracts the costs included in the model from revenue. Net accounting profit may also include overhead, financing, tax and other expenses. Keep the cost definition visible when sharing results.

The discount reduces contribution on every discounted unit, including sales that would have happened anyway. Additional volume must recover that reduction and the cost of running the offer.

No. A positive contribution difference is conditional on the supplied baseline and assumptions. A causal claim needs suitable controls, an experiment or another defensible identification strategy.

Scenario tools calculate what follows from assumed volume, price or elasticity. Their outputs are not automatically demand forecasts. Compare several plausible cases.

Include COGS and the other variable costs relevant to the decision, plus additional fixed campaign spending. Treat per-unit, percentage-of-revenue and fixed costs consistently; do not count the same cost twice.

Use a consistent basis. For contribution analysis, prices and costs are usually entered net of tax collected for a tax authority. Do not compare a tax-inclusive selling price with a tax-exclusive cost without adjusting it.

Enter support earned under the supplier's eligibility rules. If the same allowance has already reduced the unit cost or is already included in contribution, do not add it again.

Yes. A discounted selling price can be below variable cost, or campaign spending can exceed product contribution. Negative values should stay visible rather than being clipped to zero.

Some ratios have no valid finite value. A zero-cost product has undefined markup, and a discount with zero or negative unit contribution cannot recover a positive full-price baseline by selling more units.

You incur the cost of all delivered units even when a customer pays for only some of them. Keep bundle counts separate from unit counts.

Keep full precision during calculation. For a whole-unit or whole-bundle sales target, round the required quantity up. A displayed percentage rounded to two decimals may not reproduce the exact threshold.

The signed convention describes a downward-sloping demand scenario. A commonly reported elasticity magnitude of 1.5 becomes -1.5 in this calculator.

It is an explicit modeling assumption, not a universal law. Large changes can extrapolate beyond the data and imply implausible volume. Inspect several elasticities and check stock and capacity constraints.

Use the contribution lost on displaced sales, not the full revenue. Avoid counting the same lost units in both peer-product displacement and later-period payback.

No. Seasonality, availability, joint promotions and other factors can produce that pattern. The calculator aggregates effects you supply; it does not establish their causes.

For units already owned, that acquisition cost is common to both strategies and cancels from their difference. The markdown tool compares future recoverable values over the same terminal horizon.

Yes. Enter a negative net terminal value when unsold units incur a disposal cost. Use the same convention in both strategies.

The free calculators process values in your browser. They do not upload them or save them in browser storage. A CSV is created only when you choose Download calculation CSV.

No. The EUR, USD and GBP selector changes the display label. Enter all values in the same currency before calculating.

Complete every required numeric field and fix any highlighted range errors. For an inventory comparison, expected sales cannot exceed stock. Very large inputs can exceed the numeric range.

Scientific notation keeps extreme values readable. For example, 1.000e+14 means 100 trillion. Check that an input was not entered at the wrong scale.

Choose Reset example in a free calculator. This restores its hypothetical sample inputs and sets the currency label to EUR.

Use Campaign review to reconcile a promotion against a supplied baseline, Experiment studio to plan or analyze a controlled comparison, and Portfolio planner to choose among candidate offers under constraints. The same business question can use all three, but their input tables describe different objects.

No. Campaign review uses product–campaign rows, the experiment tools use assignment-unit or matched-pair outcomes, and the planner uses product–offer candidates. Download the example for the selected workspace and map the source data to that grain.

Use the unit that was independently assigned, such as a store. Transactions or repeated days within one assigned store are not additional independent assignments. Aggregate the chosen outcome to the analysis unit before importing unit outcomes or calculating group summaries.

Choose strata when assignment should be balanced within pre-specified groups. Choose paired allocation when each pre-matched pair contains exactly two units and one should receive treatment. Supply stratum or pair_id before outcomes, and analyze in a way that respects that design.

No. The seed and unit list reproduce the allocation, but they do not provide allocation concealment, tamper-proof history or a public registration. Fix the list, seed, outcome, window and exclusion rules before outcomes; do not keep trying seeds for a preferred result.

Use group,n,mean,sd with one control and one treatment row. n is the count of independent assignment units, and sd is the sample SD across their outcomes. Group means alone cannot recover paired covariance or the unit changes needed for a before/after design.

No. pair_id,control,treatment describes an independent matched treatment/control pair in each row. Comparing one treated unit with itself before and after has no contemporaneous control and does not by itself identify a promotion effect. Use the before/after-for-both-arms structure when appropriate.

Supply unit_id,group,before,after once per independent unit. The tool compares treatment and control distributions of within-unit changes. A causal interpretation needs defended parallel untreated trends and the other design assumptions; one pre-period cannot establish those trends. Staggered adoption and repeated cross-sections are outside this tool.

The descriptive contrast can exist without enough independent observations or estimable variation for the stated t interval. The tool withholds uncertainty rather than displaying a misleading zero-width interval. Check aggregation, rounding, sample sizes and the supplied data structure.

No. Compare the effect interval with a pre-specified commercially worthwhile increase as well as zero. A p-value is not the probability that the null is true, and repeated peeking or selecting outcomes changes error rates. A narrow interval also cannot verify a causal design.

With one outcome per randomized store or other unit, use the SD across those unit outcomes and cluster size 1. Use observation-level SD only when explicitly applying the equal-cluster approximation with observations per unit and ICC. Do not inflate an already cluster-level SD a second time.

No. The studio creates a local assignment and plan draft. You implement the protocol in your own operational systems, prepare the measured outcomes and import them afterward. MarginTrial does not connect to a commerce API, change prices or retrieve live orders.

At most one candidate offer, or the product's regular-price no-promotion policy. Each candidate row is a mutually exclusive alternative. Shared product inputs must match across its offers, and unpromoted products remain part of total capacity rather than disappearing from the portfolio.

The planner's cash constraint counts selected campaign spending. Committed supplier funding changes modeled contribution but does not offset that spending constraint. This keeps contribution support separate from the cash available to fund the plan; contingent rebates should not be entered as unconditional support.

Manual mode uses the expected_units you supply. Elasticity mode calculates a constant-elasticity scenario from a supplied non-positive elasticity. Neither mode estimates a demand curve from sales history. Stock, capacity, substitution and future response still need defensible inputs and an appropriate model boundary.

It bounds the additional contribution a better feasible combination could still achieve under the supplied model. A proved optimum has no remaining improvement gap. A stopped search can return a feasible best-found plan with a positive gap; that gap is not sampling uncertainty, a confidence interval or a profit guarantee.

The planner keeps the constraint failure visible instead of silently clipping demand to a limit. A candidate may be ineligible, and the overall regular-price portfolio can also be infeasible. Review the constraints and response assumptions; a different offer may be feasible, but the tool does not invent a stock-rationing demand model.

The hosted previews support 20 campaign rows, 100 assignment or unit/pair outcome rows, and a planner with 8 products and 32 offers. Offline Personal includes campaign review and Experiment studio up to 25,000 relevant rows. Agency adds the offline planner up to 120 products and 2,000 candidate offers, plus client use and report branding.

Export and reopen the JSON format belonging to the workspace: campaign project, experiment workspace or planner project. The purchase-recovery JSON contains a private license credential and restores download access, not analysis data. Keep those files distinct; none is an automatic cloud backup.

The campaign data dictionary and public template gallery describe campaign inputs and fixed synthetic cases. Experiment studio offers a separate example for each assignment/outcome structure; the planner has its own candidate CSV template and guide. Public examples are fictional, and their links contain fixed example identifiers rather than private input data.