Skip to content

BUILT FOR SHOPIFY STORES

Which discountsactually createa sale?

Some customers need the offer. Others were going to buy anyway. Invisible Offer measures the difference.

Stop paying for sales you already had.

Install on Shopify See the money question
  1. Connect Shopify
  2. Run a controlled experiment
  3. See what your discount actually changes
  • First experiment free
  • Nothing changes until you press Start
  • Never blocks checkout
  • One-click kill switch

Illustrative demo · not live store data

01 THE QUESTION

Every discounted orderlooks like a win.

The receipt shows the $18 you gave. It can't show whether that shopper would have paid $180 anyway.

SHOPIFY ORDER #2841PAID
Classic Sneaker
$180.00
Discount · WELCOME10
−$18.00
Customer paid
$162.00
Looks like the discount made a sale.

02 THE MONEY LEAK

Discount spend you can see, and the part nobody can explain

1,000discounted orders
$15average discount
$15,000discount spend

How much of it actually changed a purchase?

UNKNOWN

Shopify shows every dollar you gave away. Not one of those receipts shows what it bought.

Illustrative example

$ YOUR NUMBERS

Now run it withyour store's numbers.

Four numbers from your Shopify admin. The spend is easy to see — what it bought is what the test answers.

$
%
%

ESTIMATED DISCOUNT SPEND

$17,640/mo

$211,680 a year

Created salesSplit unknown until tested Given away
Install on Shopify Illustrative estimate from your inputs — not measured waste or savings.

03 THE EXPERIMENT

Randomly compare your offeragainst no offer.

Four steps, nothing technical on your side. Methodology stays in the fine print — and in the result.

  1. 01CONNECT

    Connect Shopify. Nothing changes yet.

    YouApprove the app and switch on the app embed.

    We learn your store's normal buying baseline. Your discounts stay exactly as they are.

    Shopify Web Pixel · no theme code edits

    Invisible Offer · OverviewOBSERVING

    OBSERVED BASELINE · NO CHANGES TO SHOPPERS

    Conversion3.1%
    Eligible shoppers / day412
    Orders observed1,284
    ✓ Web Pixel connected✓ App embed on Collecting baseline

    Baseline window · day 9 of 14

  2. 02APPROVE

    Approve the test.

    YouChoose how long a purchase counts.

    We size the test from your own traffic and lock the plan before any result exists.

    Pre-registered sample size, outcome window and refund maturity

    Invisible Offer · Measurement planLOCKED
    What shoppers see50% keep 10% off · 50% no offer
    Count a purchase within14 days
    Let refunds mature for30 days
    Estimated shoppers needed25,684
    PLAN APPROVED & LOCKED · v1

  3. 03START

    Press Start.

    YouOne click to start. One click to stop.

    Half your eligible shoppers keep the 10% offer, half see no offer — assigned at random, before they buy.

    Discount delivery verified first · sticky assignment · never blocks checkout

    Invisible Offer · ExperimentRUNNING
    10% OFF6,421shoppers
    NO OFFER6,418shoppers
    Kill switchStops the test immediately. Observation continues.

  4. 04SEE THE ANSWER

    See what your discount changed.

    YouRead the result when it's ready.

    Conversion, revenue and discount spend per shopper, after refunds.

    Unlocks at the planned horizon · insufficient data → measuring → evidence available

    Invisible Offer · ResultsEVIDENCE READY
    1. INSUFFICIENT DATA
    2. MEASURING
    3. EVIDENCE AVAILABLE
    Revenue / shopper, after refunds+$0.80
    Conversion lift+1.1 pts
    Discount spend$12,431

    Illustrative example · not merchant data

04 THE EVIDENCE

Conversion, revenue and discount spend.Not just coupon usage.

  • WHAT MOST REPORTS COUNTCoupon redemptionsTells you the code was used. Not whether it changed anything.
  • DID MORE SHOPPERS BUY?Conversion per shopper10% OFF vs NO OFFER, on randomized groups.
  • DID YOU EARN MORE?Revenue per shopperAfter discounts and refunds — lost sales are never hidden.
  • WHAT DID IT COST?Discount spend per shopperThe price of the offer, spread across everyone it was shown to.

05 THE MONEY ANSWER

What did the discountactually buy?

Same store, same weeks, shoppers split at random. Put the discount spend next to what changed — then decide whether it earns its place.

10% OFF
5.4%conversion
$10,395discount spend
NO OFFER
5.3%conversion
$0discount spend

What did $10,395 buy?

+0.1pp

observed difference in conversion

Invisible Offer measures whether that difference is real enough to justify the discount.

Illustrative example · 12,842 randomized shoppers per group · similar observed rates don't prove equivalence

→ THREE POSSIBLE ANSWERS

EXPERIMENT REPORT · 10% OFF vs NO OFFERWhat did $10,395 buy? About +0.1pp of conversion.
QUESTION THE OFFER
MEASURE10% OFFNO OFFER
Eligible shoppers12,84212,842
Conversion5.4%5.3%
Revenue / shopper, after refunds$7.57$8.26
Refund rate4.0%3.8%
Discount spend$10,395$0
Per eligible shopper+$0.69more revenue with no offer95% interval: +$0.17 to +$1.21

Assignment, delivery and attribution checks passed

Conversion barely differed, while every offer-group order carried the discount. In this example no offer earned more per shopper — a real verdict also needs the planned sample and clean delivery and attribution.

Illustrative example · not merchant performance

06 WHAT INVISIBLE OFFER DOES NEXT

COMING NEXT · AFTER YOUR FIRST EXPERIMENT

Don't just ask whether 10% works.Find out how much incentive the sale actually needs.

Your first experiment answers one question: does your existing 10% offer create sales? That evidence is the start of a bigger one.

Understand shopper context. Measure actual response. Learn where incentives matter.

Today: your existing 10% offer vs no offer. Other incentives are not part of the first release.

  1. NO OFFER
  2. 5% OFF
  3. 10% OFFtested today
  4. 15% OFF
INCREMENTAL RESPONSE
ECONOMICALLY JUSTIFIED INCENTIVE
  • Some shoppers may need 10%.
  • Some may need 5%.
  • Some may need nothing.
  • Measure the difference.

07 QUESTIONS

Before you changean offer.

What the test can answer, what it can't, and what happens before anything shopper-facing changes.

Read the discount field guide
01What does Invisible Offer actually measure?

It compares eligible shoppers who keep the normal 10% offer with shoppers who receive no experiment offer. Purchase and refund outcomes—not a prediction about one shopper—support an estimate of whether the discount creates incremental behavior. Read the guide to discount incrementality.

02Why can't I tell from a discounted order alone?

The receipt shows the discount and the purchase, but not what that same shopper would have done at full price. Random assignment lets you compare groups instead of guessing about an individual order. See discount spend versus incremental revenue.

03Will connecting Shopify change my storefront discounts?

Connection sets up measurement and requires Shopify permissions, including pixel and discount capabilities. It does not itself start a shopper-facing experiment. A controlled test is a separate, deliberate step, with readiness checks before it starts.

04Does this depend on my theme's Add to Cart button?

Primary behavior measurement uses Shopify standard customer events through a Web Pixel, not button text or theme-specific selectors. Storefront intervention still has setup and readiness requirements that should be checked on your store before a test.

05What do shoppers in the no-offer group see?

Your store, at its normal prices, without the experiment offer. Nothing else about the storefront changes. Which group a shopper is in is decided at random before they buy, and it stays consistent for that shopper's browser during the test.

06Does AI decide who gets a discount?

No. Experimental assignment is randomized before the purchase outcome. Merchant constraints and deterministic policy govern the permitted action; semantic signals are not proof that a discount caused a sale.

07How soon will I know whether the discount works?

There is no universal time or order count. Both groups need enough eligible shoppers, valid outcome attribution, and a pre-defined evaluation window. Early differences are labeled as measuring, not a win. See how a Shopify holdout test works.

08Does a no-offer purchase mean I saved 10%?

No. A single purchase cannot establish what would have happened with the offer. Avoided discount spend is not automatically realized profit: conversion, revenue per eligible shopper, refunds, and the full experiment result must be considered together. Learn about refund-adjusted outcomes.

IO INVISIBLE OFFER

Make every discountearn its place.

Measure which incentives actually change purchase behavior — and which simply give away margin.

Install on Shopify Estimate your discount spend
  1. 01Connect Shopify
  2. 02Run a controlled experiment
  3. 03See what your discount actually changes

Founding pilots: your first incrementality experiment is free.

  • First experiment free
  • Nothing changes until you press Start
  • Never blocks checkout
  • One-click kill switch

Shopify permissions include orders, products, customer events, pixels and discounts. Installation is not read-only — the discount permission is what lets a test run once you start one.