How-To Guide
How do you measure social commerce revenue?
After Meta moved Shops to website checkout, the in-app purchase figure is gone. A four-layer way to measure social commerce revenue without double-counting.
Measure it in four layers and keep them separate. Ads Manager gives you Shops-assisted purchases, the platform count that replaced the old in-app purchase figure. Your web analytics gives you sessions and conversion by platform-referred segment, using the parameters Meta appends to the checkout handoff such as utm_source=IGShopping, utm_medium=Social and cart_origin. Your order system gives you booked revenue net of refunds, and it is the authority for money. A post-purchase survey or holdout test gives you a range for whether social caused the purchase. Report assisted purchases as a share of orders, never add them to booked revenue, and reconcile by order ID.
- Shops-assisted purchases is a platform-attributed count, not revenue; the old in-app purchase figure no longer exists.
- Meta appends utm_source=IGShopping, utm_medium=Social, cart_origin and fbclid to the checkout handoff; store them on the order record.
- Booked revenue and refunds come from your order system, and incrementality comes only from a survey or holdout test as a labelled range.
What changed in the measurement
The old report was simple because Meta owned the transaction. A shopper tapped a product inside Instagram, paid inside Instagram, and the platform recorded a purchase it could count. That is no longer the flow. Outfeed's Instagram Shop guide quotes Meta's Business Help Center saying that as of September 2025, Shops on Facebook and Instagram use website checkout, and that payment processing has been discontinued for Shops. Coruzant's commerce report dates the wind-down earlier: Meta began moving US merchants off in-app purchasing on 5 June 2025, and native in-app checkout was fully phased out across US merchants in 2025.
The measurement loss is specific. Coruzant reports that Shops ads lost View Content, Add to Cart and Initiate Checkout as available conversion events, leaving Purchase as the only one, and that the hybrid Website and shop conversion location was retired. The old in-app purchase figure is gone. What remains is a Purchase event on a shop ad and a new platform metric that is not the same number. Any report built on pre-2025 in-app purchase counts is now describing a flow that no longer exists.
The flow that replaced it is a handoff, and the handoff is where measurement now lives. Meta's Shops ads reference says buyers discover products and build their carts on Facebook and Instagram, and the final checkout happens on the seller's own website. That sentence is the whole model. Discovery is still social. The order, the payment, the refund and the margin are yours. You cannot measure social commerce revenue by asking the social platform for the final number, because the platform no longer holds it.
Two dates matter and they are not the same date. June 2025 is when Meta started moving US merchants off in-app purchasing, per Coruzant. September 2025 is the month Meta's own help page gives for Shops switching to website checkout, and Outfeed notes that Meta's page gives the month, not an exact day. When a report is compared year over year, name which month you are using. A report that silently mixes the two will show a step change that is an artefact of the date, not a change in demand.
Original diagram of the measurement change. The removed items follow Coruzant's commerce report; the surviving checkout handoff follows Meta's Shops ads and checkout URL references, all read November 2026.
Make it work for your brand
Find it. Remake it. Post it. Superdirector turns the viral formats in your niche into your brand's carousels, X posts and video scripts, and keeps the calendar full.
The number that replaced in-app purchases
The replacement metric is Shops-assisted purchases. Coruzant reports that a new Shops-assisted purchases metric replaced the old Meta purchases figure in Ads Manager, and is explicit that it is a different number: "Shops-assisted purchases is not the old Meta purchases metric." Its warning is the one to tape to the dashboard: "Anyone still benchmarking against pre-2025 ROAS figures for in-app conversions is comparing against a number that no longer describes anything." The new figure belongs to the handoff era, not the checkout era.
Read it for what it is: a platform-attributed count of purchases that Meta can connect to a shop surface it can see. It answers which ad, shop or product interaction preceded a purchase the platform can observe. It does not answer how much money you booked. It does not know your refunds, your discounts after the handoff, your shipping revenue or your margin. Treat it as a directional count, the top of the funnel, never as the revenue line in a client report.
The Purchase event still exists on the ad side, which is why the metric is useful at all. Coruzant says Purchase is the only conversion event Shops ads have left. For the count to mean anything, the plumbing behind it has to work: products need to be in a Meta catalog before anyone can buy them, per Meta's catalog documentation, and Meta's order management reference says products become purchasable after a seller is onboarded and inventory is uploaded. A missing catalog item is not a demand problem. It is a measurement hole.
There is a second reason the count is partial. Meta's Shops ads reference describes Shops ads as an additive enhancement to website-traffic campaigns, not a replacement for the seller's checkout. That means the platform sees the assisted path it can instrument, while the sale completes in a system it does not run. The gap between the count and your order total is expected. The job is to explain its direction, not to make it disappear.
Trace the handoff with the parameters Meta appends
Because checkout moved to your site, attribution became a URL problem you can inspect directly. Meta's checkout URL reference says the checkout process takes place on your own website, and that the seller must provide a checkout URL that can receive cart information: product IDs, quantities and coupons. The same page gives the tracking behaviour that makes a shop session nameable. Meta automatically appends utm_source=IGShopping and utm_medium=Social to traffic coming from a shop, and appends cart_origin with a value of facebook, instagram or meta_shops depending on the source of the checkout.
Two more parameters finish the picture. Meta's reference says fbclid is a Facebook Click ID added automatically to all outbound links from Facebook and Instagram, and that cart_origin is appended in addition to any parameters you set on the checkout URL and any UTM parameters set at campaign level. The reference also warns that UTM parameters are passed in the URL, so the seller must incorporate its own tracking on the checkout page, and that if it redirects to another URL it must pass those parameters along.
That gives you a chain you can test. A shopper taps a tagged product. The checkout URL arrives with products, quantities, coupon, utm_source, utm_medium, cart_origin and fbclid. Your checkout page reads them and stores them on the order. Later, your analytics and your order system can both name the same session. The parameters are also added on top of any utm_content you set at product level in the catalog, so keep one naming convention or you will double-tag the same visit.
This is the cheapest measurement upgrade available, and it is a build task, not a reporting task. Capture cart_origin and fbclid in the order record, not only in the web analytics tool. Web analytics expires, ad blockers interfere, and a redirect can strip a parameter. The order record survives, and it is the join key that lets you reconcile platform-assisted counts with real orders later.
- Confirm your checkout endpoint accepts products, quantities and coupon.
- Read utm_source, utm_medium, cart_origin and fbclid on the landing page and store them before any redirect.
- Write those values onto the order record, not only into the analytics session.
- Test one shop tap end to end and confirm the same order shows both a platform value and an order ID.
Original diagram of the checkout handoff. Every parameter shown is named in Meta's checkout URL reference, read November 2026; storing the values on the order record is the article's method.
Reconcile the four layers of revenue
No single number is social commerce revenue. There are four layers, and each answers a different question. The clearest published measurement answer comes from Shopify's social commerce trends report, which tells brands to track website traffic from each social channel, conversion rate from social sessions, revenue and sales from each platform-referred segment, and post-purchase survey responses that ask how customers first discovered the brand. That is a good spine. Put a named authority beside each layer so the report says what it knows and what it is inferring.
| Layer | What it answers | What it cannot tell you | Authority |
|---|---|---|---|
| Platform assisted purchases | Which social surface preceded a purchase Meta can see | Final order value, refunds, margin or incrementality | Ads Manager |
| On-site sessions by platform-referred segment | How much traffic and conversion the handoff produced | Whether the order would have happened anyway | Web analytics |
| Order system revenue | Money actually booked, net of refunds | Which social touch caused the order | Your order system |
| Post-purchase survey or holdout test | Whether social caused the purchase | Exact revenue for one post | Survey or test record |
The first layer is the Shops-assisted purchases count. Use it to see direction and relative performance between surfaces. Do not use it as revenue. The second layer is your own analytics, where a shop session can be named from the appended parameters. Shopify makes the practical point that where the purchase completes changes how a brand sets up tracking, attribution and fulfilment, which is exactly why the on-site layer exists.
The third layer is the authority for money. Your order system knows the order value, the discount, the refund and the margin, and no social platform does. The fourth layer answers the question the first three cannot: whether the purchase was caused by social or merely preceded by it. Shopify's fourth item, post-purchase survey responses that ask how customers first discovered the brand, is a simple version of that check. A holdout test is the stronger version when the spend justifies it.
Reconcile by order, not by total. Join a platform-assisted purchase to an order using cart_origin, fbclid and the order ID you stored at checkout. Some orders will join cleanly. Others will carry a platform value but no order ID because the shopper abandoned. A third group will have an order but no platform value because the parameter was stripped or the customer typed the URL. Report all three counts. The size and direction of the unjoined set is the honest measure of how much the platform number and the order number can disagree.
A hypothetical month shows the arithmetic, and every figure in this paragraph is invented for illustration. Suppose the order system books 400 orders and $48,000 net revenue. The on-site report attributes 120 orders and $14,400 to platform-referred segments. Ads Manager reports 150 Shops-assisted purchases. Multiplying 150 by a $120 average order value would suggest $18,000, and that is not money you booked; it must not be added to the $48,000. The useful statements are that 120 of 400 orders, or 30 percent, were platform-referred, and that 30 assisted purchases did not join to an order and need a cause. Those numbers are illustrative only.
Original diagram of the four measurement layers. The four-part structure follows Shopify's social commerce measurement answer; the authority beside each layer and the no-adding rule are the article's own reporting standard.
Report revenue without double-counting
Choose one primary number per question, and never add the layers together. If the question is how much we sold, the order system is the answer. If the question is which social surface drove discovery, the assisted-purchases count is the answer. If the question is how much of that revenue social caused, only the test or survey can answer, and it answers with a range. A report that adds assisted purchases to booked revenue counts the same order twice and flatters the channel.
Report assisted purchases as a share of orders, not as money. On-site revenue belongs by platform-referred segment. Incrementality is a labelled range from a test, with the assumption written beside it. Keep the platform's own definition visible, because Coruzant's point stands: Shops-assisted purchases is not the old Meta purchases metric, and the account's own definition is the only one that describes what you are reading. Do not import a definition from a blog, including this one.
Put the method note in the report, dated. Say that Meta's public page gives September 2025 for the switch while Coruzant dates the US wind-down from 5 June 2025, and name the month the comparison uses. Make clear that the assisted count is platform-attributed and the revenue is order-system booked. Give the survey or test window. A reader who can see those three statements can tell a real change from a change in method.
| Monthly line | Source | How to read it |
|---|---|---|
| Orders and net revenue | Order system | The money line; net of refunds |
| Platform-referred orders | On-site analytics | Orders joined to cart_origin or fbclid |
| Assisted purchases | Ads Manager | Platform-attributed count, not money |
| Assisted share of orders | Derived | Assisted purchases divided by orders, labelled approximate |
| Incrementality range | Survey or holdout test | A labelled range, never a single figure |
| Unjoined platform purchases | Reconciliation | A count with a named cause where possible |
Keep the report stable month to month. Change the method only with a dated note, and re-run the previous month under the new method before comparing. Shopify describes social commerce as becoming part of unified commerce measurement, which is the right ambition: one order record, several attributed touches, one money authority. The platform metrics sit beside the order record, not above it.
Where Superdirector fits, and where it does not
Superdirector is on the publishing side of this problem, not the measurement side. The Growth plan is $199 a month for one AI-run operated account. It posts eligible Instagram carousels and X text posts once posting is turned on for the account, under account-level standing posting permission granted by the account owner. The Starter plan is $29 a month for MCP access with hands-on creation. Neither plan runs a web analytics property, a Meta catalog or an order system, so neither plan can report Shops-assisted purchases or booked revenue.
That division is useful if you state it plainly. Superdirector can keep the publishing cadence that creates the discovery surface, and your analytics and order system remain the authority for what the discovery earned. Do not ask a publisher for a revenue number it cannot see, and do not let a platform-attributed count stand in for your own books. The measurement method in this guide belongs to your stack, whatever tool publishes the posts.
What the sources do not settle
Three gaps are worth naming in the report itself. First, no analysed source publishes the exact definition of Shops-assisted purchases: its attribution window, its treatment of view-through, and whether it counts a purchase once or once per impression. Coruzant names it as the replacement and says it differs from the old figure, but the account's own definition is the only complete one. Second, the effective date is reported differently: Coruzant starts the US wind-down on 5 June 2025, while Outfeed and Shopify say September 2025, and Meta's page gives the month. Third, no source proves incrementality. An assisted purchase is attributed, not proven causal.
The 2020 announcement is in the dossier only to explain what was removed. Meta's 2020 Facebook Shops announcement described ordering without leaving the app when a business had enabled checkout in the US, and added a shop tab to the Instagram navigation bar. Coruzant records that the dedicated Shop tab left Instagram's bottom navigation bar in February 2023 and that Live Shopping was discontinued. Those are history. They are useful for explaining why an old report looks the way it does, and wrong as a description of today.
What to do with an unsettled question: label it. A report can say that the assisted metric is platform-attributed, that its definition is read from the account, that the effective date is stated, and that incrementality is a range from a named test. That is more useful than a confident number built on a definition nobody checked. The honest version of social commerce measurement is a reconciliation with visible gaps, not a single figure that pretends the handoff does not exist.