Recovered Revenue Is Not the Same as Revenue Your Messages Created

A customer leaves a basket unfinished.

The retailer sends an email two hours later. The customer returns the next morning and buys.

The dashboard calls the order recovered revenue.

But did the email create the sale?

Perhaps the customer needed the reminder. Perhaps they were interrupted, planned to return after payday, opened a new tab to compare delivery, or simply remembered the purchase without seeing the message.

The order happened after the email. That sequence is useful to record, but it does not prove causation.

A current retailer discussion asked a more practical question:

"Ecommerce owners: what did you actually do with your abandoned carts last week?"

Email, WhatsApp, SMS, a phone call or nothing at all may each be part of the answer. The harder question is which action changed the outcome enough to deserve more time, budget or discount.

Source discussion:

https://www.reddit.com/r/shopify_growth/comments/1vwztm4/ecommerce_owners_what_did_you_actually_do_with/

Start with the decision, not the dashboard

Before measuring a recovery flow, write down the decision the result must support.

For example:

• Should we keep sending the first reminder?
• Does SMS add enough value after email to justify its cost and consent burden?
• Does a discount create extra orders or subsidise customers who were returning anyway?
• Should staff call only high-value abandoned checkouts?
• Is checkout friction a bigger problem than follow-up messaging?

This prevents the report from becoming a collection of attractive numbers with no operational consequence.

Current NIQ guidance on incrementality uses the same order of thinking: work backwards from the decision, then make sure the data, infrastructure and method can answer it.

Reference:

https://nielseniq.com/global/en/insights/analysis/2026/retail-media-incrementality-measurement/

Define what "abandoned" means in your store

An item added to a basket is not the same as a checkout started, a payment failed or a shopper who asked staff to hold a product.

Choose one event for the test.

Possible definitions include:

• a known customer started checkout but did not complete within 60 minutes
• a payment attempt failed and no successful order followed within 30 minutes
• a registered shopper left a basket containing at least one available item
• a staff-assisted quote or reserved basket expired without purchase

Record the event time, customer or session key, products, basket value, channel and location where applicable.

Then define exclusions before looking at results:

• the item became unavailable
• the customer completed another linked order
• the basket was a duplicate
• the customer had not consented to the intended channel
• the payment was still processing
• the transaction was a test or staff order

If the denominator changes every week, the recovery rate cannot be trusted.

Separate four different measurements

Recovery programmes often mix four questions into one number.

1. Delivery and attention

Was the message delivered, opened or clicked?

These measures help diagnose execution. They do not prove that an order was created.

2. Attributed orders

Did a customer purchase within the rule used by the platform, such as seven days after a message or three days after a click?

This connects activity with an outcome. The rule may still credit customers who would have returned unaided.

3. Incremental orders

How many more eligible customers purchased because they received the recovery flow than would have purchased without it?

This is the causal question.

4. Commercial contribution

After message costs, discounts, returns, payment charges and staff time, did the extra orders improve profit enough to continue?

A programme can perform well at the first two levels and disappoint at the fourth.

Build one recovery evidence table

You do not need an elaborate attribution platform to improve the first version of this measurement. You do need one joined record.

For every eligible abandonment, capture:

• abandonment ID and time
• customer or privacy-safe session key
• basket value and product group
• new or returning customer
• recovery group: treatment or holdout
• every message channel and send time
• click or response where available
• purchase time and order ID
• discount and incentive cost
• return or cancellation
• gross margin or a defensible margin estimate

Keep the raw events. Do not overwrite the first email when an SMS is sent later.

That history exposes overlap. Otherwise, the email tool, SMS provider, advert platform and staff call log may all claim the same order.

Use a holdout to estimate what would have happened anyway

Where volume and tooling permit, randomly assign a share of eligible abandonments to a holdout that receives no recovery treatment being tested.

The treatment group receives the normal flow. Both groups remain subject to the same product availability, prices, website experience and normal trading conditions.

The basic calculation is:

`Observed incremental lift = treatment purchase rate - holdout purchase rate`

Suppose 300 eligible customers receive the recovery flow and 36 purchase. The treatment purchase rate is 12%.

Another 100 eligible customers are held out and 8 purchase without the flow. The holdout purchase rate is 8%.

The observed lift is four percentage points. Applied to 300 treated customers, that represents 12 observed incremental orders, not all 36 treatment orders.

This example explains the method. It is not automatically a statistically reliable conclusion. A small sample can move sharply because of a handful of purchases, product mix, promotions or chance.

BCG's current measurement guidance makes the wider principle clear: last-touch attribution tells you what happened after an action, while structured experiments help answer whether the action changed behaviour.

Reference:

https://www.bcg.com/publications/2026/measuring-incrementality-in-next-best-action-programs

Small retailers should be honest about sample size

A store with thousands of eligible carts can reach a useful comparison faster than a specialist retailer with 20 abandoned checkouts a month.

Low volume does not make measurement pointless. It changes the level of confidence.

Practical options include:

• run the test for longer
• test one major change at a time
• keep the eligibility rule stable
• review absolute orders as well as percentages
• group only genuinely comparable products and customers
• record promotions, stockouts and website changes that may distort the period

If random assignment is not practical, a retailer can compare matched cohorts, alternating periods or a before-and-after baseline. Those are weaker designs because seasonality, product mix and external events can explain the difference.

Label the conclusion accordingly. "Purchases increased after the change" is not the same as "the change caused the increase."

Prevent contamination and double counting

A useful holdout can be ruined quietly.

Check whether holdout customers still receive:

• a generic newsletter containing the same products
• paid retargeting for the abandoned basket
• an SMS after being withheld from email
• an automated platform reminder from a second app
• a staff call triggered by the same event

If the test asks whether the whole recovery programme works, the holdout must be withheld from the whole programme being measured.

If the test asks whether SMS adds value after email, both groups may receive email while only the treatment group receives SMS.

The design must match the decision.

Also suppress messages immediately after a verified purchase. A reminder sent after checkout creates a poor customer experience and corrupts the reporting.

Do not let a discount take credit for the customer

A discount can increase conversion and still weaken the business.

It may:

• remove margin from a customer who only needed a reminder
• encourage shoppers to abandon deliberately
• shift a purchase forward rather than create a new one
• attract orders with higher return rates
• make full-price recovery look less effective than it is

Test the lightest useful intervention first:

1. A clear route back to the basket.

2. Delivery, returns or product reassurance.

3. Help with a genuine payment or checkout problem.

4. An incentive only where evidence justifies the margin cost.

Then calculate contribution, not just revenue:

`Incremental contribution = incremental net sales margin - discounts - message cost - staff cost - expected returns`

The exact formula depends on the business, but the costs should not disappear from the decision.

Fix preventable abandonment before improving the reminder

Recovery is not a substitute for a usable checkout.

If customers repeatedly leave because delivery charges appear late, payment fails, mobile forms are difficult or product information is unclear, sending more messages treats the symptom.

Review abandonment reasons alongside the recovery report:

• payment failure
• unavailable stock
• unexpected delivery cost
• forced account creation
• unclear returns
• slow or confusing checkout
• customer distraction or delayed intent

Some customers need a reminder. Others need the original obstacle removed.

Our guide to silent checkout loss explains how to investigate friction before adding more follow-up:

https://ezycarto.com/blog/the-customers-you-lose-at-checkout-never-complain-first

Set the decision rule before the result arrives

Write the action threshold in advance.

For example:

• continue the first reminder if observed incremental contribution remains positive across two review periods
• add SMS only if it creates additional net orders beyond email and complaint rates remain acceptable
• stop a discount when incremental margin is below its incentive and delivery cost
• move high-value failures to staff review when the expected contribution justifies the time
• fix checkout first when the same preventable reason dominates abandonment

An agreed rule makes it harder to move the goalposts because the dashboard looks impressive.

Where EzyCarto fits

EzyCarto is designed to connect supported customer, sales and operational context rather than leave each retail workflow in a separate tool.

Its current public scope includes unified customer profiles, purchase history and supported customer journeys in CRM. EzyCarto Analytics can provide sales, shopper-engagement and timeframe views, while Back Office includes plan-supported communications and activity records.

That connected context can support the evidence table: who the customer is, what was in the journey, which purchase followed and what the wider sales record shows.

It does not make EzyCarto the judge of causation.

EzyCarto does not currently claim automatic abandoned-cart detection across every external channel, automatic holdout assignment, omnichannel recovery orchestration or causal attribution. Retailers still need to define the event, comparison, consent, costs and decision.

Explore EzyCarto CRM:

https://ezycarto.com/crm

Review EzyCarto Analytics:

https://ezycarto.com/analytics

If the customer record itself is fragmented, use the CRM migration context test before designing a recovery programme:

https://ezycarto.com/blog/the-customer-context-test-every-crm-migration-should-pass

Run this one-month recovery audit

Before buying another recovery app or increasing message volume, take one month of eligible abandoned carts and answer:

1. What exact event placed each cart in the recovery group?

2. Which customers were excluded, and why?

3. Which messages, calls, ads or incentives did each customer receive?

4. How many bought without any recovery treatment?

5. How many orders were claimed by more than one channel?

6. What was the net contribution after discounts, costs and returns?

7. Which abandonment reason should be fixed at checkout instead?

8. What decision will the next test change?

The purpose is not to make recovered revenue smaller.

It is to make the number honest enough to guide the next action.