NewWeather Demand Modelling is live. Forecast demand before it arrivesWeather Demand Modelling is live

Every promo as one contribution-margin line, with a verdict.Every discount read in profit, not revenue.

Margin OS is a promo ledger in true contribution margin: each code, automatic and manual discount as a single line showing what you gave away versus the realized CM1 the promoted orders produced, with a Healthy, Margin-thin or Underwater verdict on every one. The give is split by mechanic into product give and shipping subsidy, scored on incrementality, and run through a restructure simulator. Then it hands you who to wean onto full price and who to suppress. We tell you the move; we do not run your store.

Discounting · promo ledgerDemonstrative data
Every promo · discount given vs realized CM1, with the margin verdict
PromoGivenCM1RateVerdict
Sitewide 25% code$31.4k$18.2k24.1%Margin-thin
Free shipping auto$19.7k$26.0k33.8%Healthy
Clearance 50% code$14.2k-$3.1k-9.4%Underwater
Welcome 10% auto$9.6k$11.8k36.2%Healthy
Bundle BOGO manual$8.9kno CM data-No CM data
Where the give goes · product give vs shipping subsidy
Product give $53.6k hits CM1Shipping subsidy $30.2k gated CM2
You gave away $83.8k across 5 promos; the orders they touched produced $52.9k of realized CM1 after the product give. 1 promo ran underwater and one is margin-thin against the 31% house CM1 rate. Each line opens to the redeemers and the give-by-mechanic split.
Discount given
$83.8k
product + shipping
Realized CM1
$52.9k
after the give
Promo CM1 rate
27.4%
vs 31% house
Underwater promos
1
realized CM1 below zero
Subsidy leak
$22k
to full-price loyals
Most efficient lever
$ off
$1.31 give per $1 CM1
Margin intelligence from the award-winning team behind 100+ brands
What it answers

The questions behind every promo decision.

Not "how much revenue did the sale do." What it cost in margin, who it actually won, and who to stop subsidising.

01
Which promos make money, and which run thin or underwater?
The promo ledger reads every code, automatic and manual discount as one CM1 line: discount given versus the realized contribution the promoted orders produced, each carrying a Healthy, Margin-thin or Underwater verdict. A promo that looks like growth while destroying margin has nowhere to hide.
02
Where does the give actually go: product margin or shipping?
The give-by-mechanic waterfall splits every promo dollar into product give, which hits CM1, and shipping subsidy, held in gated CM2, so a free-shipping cost never falsely sinks the product verdict and you see which mechanic the leak sits in.
03
Which discount lever earns its keep, and is the lift real?
Mechanic effectiveness scores each lever (% off, $ off, free shipping, BOGO) on realized-CM1 efficiency, cannibalization and pull-forward, then measures genuine incrementality against a holdout, so you keep the lever that drives net-new orders and retire the one that just discounts demand you already had.
04
Who would have bought anyway, and who is genuinely discount-driven?
Full-price propensity scores the whole base into dependency bands, from full-price loyal to discount-only, surfacing the margin leaking to customers who pay full price anyway and the valuable customers genuinely hooked on a code.
05
What would the next promo give away, and to whom, before we run it?
The promo-restructure simulator lets you configure depth, eligibility and reach and reads the give, the audience and the net CM1 in true contribution, plus a portfolio restructure that re-depths each dependency band, so you tune the promo before a dollar leaves the door.
06
Who should we wean to full price, and who should we suppress?
Two ranked, exportable audiences: suppress the full-price-loyal buyers being needlessly discounted, and wean the valuable discount-dependent ones onto full price, each with the discount at stake, handed to your email tool to action.
Mechanic effectiveness & incrementality

Which lever earns its keep, and is the lift real?

A promo verdict tells you about one code; the mechanic tells you about the lever. Margin OS scores each discount mechanic on realized-CM1 efficiency, then asks the harder question: does it cannibalise full-price sales, just pull orders forward, or drive genuinely incremental volume? Incrementality is measured against a holdout, so you keep the lever that wins net-new orders and retire the one that discounts demand you already had.

  • Give per $1 CM1. Every lever ranked on the discount it hands out per dollar of realized margin
  • Cannibalization & pull-forward. Give flowing to full-price buyers and orders that came early
  • Measured incrementality. The causal lift each mechanic drives, read against a holdout
See your mechanic leaderboard
Discounting · mechanic effectivenessDemonstrative data
Give per $1 of realized CM1 · by mechanic, lower is leaner
$0$1$2$3$4$ off$1.31Free shipping$1.76% off$2.40BOGO / bundle$3.85
Measured incrementality · CM1 lift vs holdout, last campaign
$ off +18% liftFree shipping +11% lift% off +4% lift
$ off is the leanest lever at $1.31 given per $1 of CM1 and drives +18% measured incremental margin against the holdout; BOGO hands out $3.85 and runs underwater. Cannibalization and pull-forward read on the full leaderboard, with the holdout closing the loop on what was truly incremental.
Promo-restructure simulator

Tune the promo before a dollar leaves the door.

Before you run a promo, see what it gives away and who it targets, in true CM1. Configure depth, eligibility by dependency band and audience reach, and the give, the audience and the net contribution recompute live in a CM1 cascade. A portfolio restructure re-depths every band at once, so you can hold the full-price loyals out of the next sitewide sale and watch the blended discount rate and the margin recovered move together.

  • Single-promo designer. Depth, eligibility and reach, with the give and audience live
  • CM1 cascade. Baseline margin, the give, the incremental lift and the net, end to end
  • Portfolio restructure. Re-depth each dependency band and read the margin recovered
Run a promo what-if
Discounting · promo simulatorDemonstrative data
EligibilityExclude full-price-loyal
Discount depth20% off
Audience reach80% of eligible
Projected CM1 cascade · baseline → give → lift → net
$0$46.0kBaseline CM1-$15.4kDiscount given+$7.8kIncremental CM1$38.4kNet CM1
This promo would give away $15.4k to 3,420 eligible customers and, after a measured +$7.8k of incremental margin, net $38.4k in true CM1. Excluding the full-price loyals cut the give without losing the orders. Open the cascade to the audience list, or re-depth every band in the portfolio what-if.
Discount dependency & full-price propensity

Know who to wean, and who you're just subsidising.

Stopping discounts blind loses customers; discounting blind burns margin. Margin OS scores every customer on full-price propensity and sorts the whole base into dependency bands, from full-price loyal to discount-only, in true CM1. Then it turns the edges into two exportable audiences: the full-price-loyal buyers you are needlessly subsidising, and the valuable discount-dependent ones worth weaning onto full price.

  • Dependency bands. The whole base scored on full-price propensity
  • Subsidy leak surfaced. Loyal customers being discounted for no reason
  • Suppress & Wean audiences. The two lists this engine produces, exported to your email tool
See your dependency bands
Discounting · dependency & propensityDemonstrative data
Discount-dependency bands · share of base
Full-price loyal31% · 4,210 custMixed51% · 6,980 custDiscount-dependent14% · 1,870 custDiscount-only4% · 540 cust
Action lists · the two audiences this engine produces
1
Suppress 980 full-price-loyal buyers still being discounted
subsidy leakKlaviyo · Email
$22kdiscount at stake
2
Wean 420 valuable but discount-dependent buyers onto full price
CM1 at stakeKlaviyo · Email
$48klifetime CM1
Propensity is a transparent composite of full-price order and spend share over each customer's history, in true CM1. The size of the prize is the factual discount currently flowing to each list; recovery depends on how many convert. Each audience exports to your email tool, where you approve and send. We never touch your store.
Where the numbers come from

Built on your store, reconciled to true profit.

Discounting is read straight off your orders and costs. No ad connection is required to see the margin cost of a promo.

Orders & discounts

Every order, refund and discount from your store, so the discount given is net and real and tied to the customer who used it.

ShopifyWooCommerceBigCommerce

True cost

COGS, landed cost, shipping and payment fees, so the discount is measured against the margin it actually erodes, not the headline price.

Cost feeds3PL / freightPayment fees

Audiences & holdouts

The suppress and wean audiences sync to your email tool, and the same connector runs the holdout that measures whether a mechanic drove genuinely incremental margin. The discounting decision becomes a send your team controls.

KlaviyoEmail / SMS
Every figure is reconcilable to its source, the discount cost a CFO signs off.
The team behind it

Built by an award-winning analytics team.

Margin OS comes from Blufire, trusted by 100+ mid-market and enterprise brands and recognised across the APAC and Global Search Awards. The same people now model your margin.

100+
Brands served
$5M-$1B
Turnover served
4
Industry awards
100 Fast StartersAPAC Search Awards 2025 WinnerGlobal Search Awards 2025 FinalistGlobal Agency Awards 2025 Finalist
Questions

The things buyers ask.

Every promo, every code, automatic and manual discount, read as one contribution-margin line: the discount you gave away versus the realized CM1 the promoted orders produced. Each line carries a Healthy, Margin-thin or Underwater verdict against your house CM1 rate, and opens to the redeemers and the give-by-mechanic split. It is the kill-list and the keep-list in one table.
Shopify shows the revenue you discounted. We show the contribution-margin cost of it, split the give into product margin and shipping subsidy, score each mechanic on efficiency and measured incrementality, and hand you the customers to suppress or wean. You optimise for profit, not a revenue line that hides the margin you gave away.
Yes. The mechanic view scores each lever on realized-CM1 efficiency, cannibalization and pull-forward, and measures genuine incrementality against a holdout run through the activation connector, so you keep the lever that wins net-new orders and retire the one that just discounts demand you already had.
Before you run a promo, you configure depth, eligibility by dependency band and audience reach, and it reads the give, the audience and the net CM1 in a live cascade. A portfolio restructure re-depths every band at once, so you see the blended discount rate and the margin recovered move together before a dollar leaves the door.
Two exportable audiences, synced to your email tool: suppress the full-price-loyal customers you are needlessly discounting, and wean the valuable discount-dependent ones onto full price. Propensity is a transparent composite of full-price order and spend share over each customer's history, in true CM1. You approve and send; we never run your store.

See what your discounting really costs.

Connect your store and your costs. We read every discount in true contribution margin, show you the customers it won, and hand you the audiences to act on.

01

Connect your store

Orders, refunds and discounts. No data team.

02

Load your costs

COGS, shipping and fees, mapped with us.

03

Read the promo ledger

Every promo as a CM1 line, with a verdict.

04

Judge the mechanics

Efficiency, incrementality, and a what-if.

05

Get the audiences

Suppress and wean lists, exported to send.

Book a demo Talk to an analyst