How Do You Measure the Success of a Cart Recovery Campaign?

To measure the success of reminders sent to shoppers who abandon a cart, you need three numbers: the share of messageable carts that turn into orders, the extra orders the reminder brings in, and the profit those extra orders leave after message fees and discounts. Do not divide by every abandoned cart in your e-commerce platform’s report; the denominator should include only carts whose owner has a phone number and messaging consent on file and who did not complete the order during the waiting period. To see the extra orders a reminder adds, send no message to a small, random share of these carts and compare the two groups; where your tool cannot do that, use the normal weeks before the reminders were switched on as your benchmark.

Fill in the worked example below with your own figures to see how many extra orders the reminders bring in over a period and what each extra order costs you. The same table will also show you when a weak number points to the waiting time, the discount code or an obstacle on your site.

Count only the carts a message can reach

Dividing recovered orders by the abandoned-cart count your e-commerce platform shows is the first calculation that comes to mind. But a reminder can never reach some of those carts. A rate calculated that way says less about how well the message works than about how many carts you can reach. Take three groups of carts out of the reminder’s calculation:

• The cart of a visitor who added products without logging in and never typed a phone number into any form. There is no number to send the message to.
• The cart of a customer whose number you have but who has not given you consent to message them. According to the WhatsApp Business Messaging Policy, you may message someone only after receiving their number from them and their opt-in to your later messages.
• The cart of a customer who completed the order on their own before the reminder was due. This customer has already come back; a reminder would be both pointless and annoying.

Pushouse’s WhatsApp cart recovery module likewise makes sure no “you forgot your cart” reminder reaches a shopper who has already ordered, and it does not write to customers without consent. Call the remaining carts reachable carts: carts whose owner has a number and messaging consent, and that had still not turned into an order when the waiting period ended. In practice, you read this number as the carts that received a reminder plus, if you have one, the carts you held back in a control group. This is the denominator of the recovery rate.

Track the share of reachable carts separately

Also note what share of all abandoned carts are reachable. If that share is low, changing the wording or the timing of the message will make little difference; first look at how you ask for messaging consent in your checkout and sign-up forms. The more customers give consent, the more carts you can remind with the same traffic. If the flow is not set up yet, our explainer on WhatsApp cart reminders covers how it works; the measurement below makes sense once reminders have been running for a few weeks.

How is the cart recovery rate calculated?

Take a home textiles store that has run reminders for four weeks with no campaign on. The figures are hypothetical and chosen only to show the calculation; do not treat them as a target for your own rate. At the end of the period, the table shows:

• Abandoned carts in the platform report: 4,800
• Reachable carts, i.e. carts that received a reminder: 1,240
• Reminders delivered: 1,170
• Recovered orders: 112
• Recovered revenue: 128,800 Turkish lira (TL)

The same 112 orders give three different rates depending on what you divide them by. Divided by reachable carts, the result is 9.0%. Use this one when you talk about the recovery rate, because it accounts for every cart the reminder could reach. Divided by delivered reminders, it is 9.6%; this rate shows what share of the customers who got the message came back. Divided by all abandoned carts, only 2.3% is left. That low figure says less about a weak message than about the fact that you can reach only a quarter of the carts.

Ask for the denominator before comparing shared rates

Before you compare a rate shared by another store, an agency or a tool with your own, ask which number it was divided by. Because there is no way of calculating the cart recovery rate that everyone shares, the same result can look like 2.3% or 9.6% depending on the denominator, as in the example above. That is why Pushouse’s cart recovery calculator states its denominator: it calculates the rate from delivered reminders and shows the estimate as a range rather than a single number.

Revenue per cart and undelivered reminders

Add revenue per reachable cart to the table as well: 128,800 TL divided by 1,240 carts comes to about 104 TL. When basket values change from one period to the next, this figure changes even if the order rate stays the same; it shows the difference in value that the rate alone misses. If your tool also shows read and click counts, these help you see at which step the message stalls; measure success by orders and revenue.

The gap between reminders sent and delivered points to a separate problem. If the owner of the number does not use WhatsApp, or the number was entered incompletely, the message does not arrive; a reminder that hits WhatsApp’s per-user marketing message limit is not delivered either. Our guide to reading message reports explains how to tell apart the causes of undelivered messages. When the delivery rate falls, the rate per reachable cart falls with it, while the rate per delivered reminder may hold steady. Tracking the two side by side tells you whether a drop comes from the message or from delivery.

Take the figures from the same source every period

If you use Pushouse, you can read recovered orders and revenue in your panel. At the end of each period, write these figures into a single table with the start and end dates, and take them from the same source every time; if the source changes, the two periods’ rates cannot be compared. In the sections below, you will add control group, cost and opt-out columns to the same table.

Separate the orders that would have come anyway

Some of the 112 recovered orders would probably have come in even without a reminder. Some customers who abandon a cart come back on their own, for example once they have finished comparing prices, once their salary has been paid or once they have dealt with whatever interrupted them. Looking at a single order, you cannot tell whether it came because of the message or would have been placed anyway. You can only make that distinction at group level, by comparing with similar carts that received no message. The recovered order figure stays as it is; the extra orders the reminder adds go into the table as a separate column.

If your tool allows it, do not send a reminder to a small, randomly selected share of reachable carts. Do not pick the carts that will go without a message yourself, by value, customer or time of day; if the selection is not random, the two groups differ from the start and you cannot put the gap down to the message. Count orders the same way in both groups, for example those placed within the same number of days after the cart was abandoned.

A control group example and what it costs

The example store tries this over the next four normal weeks. It randomly holds back 195 of 1,300 reachable carts, or 15%, and sends a reminder to the remaining 1,105. Of the carts that received a reminder, 104 turn into orders, a rate of 9.4%. Of the 195 carts left without a message, 7 also turn into orders on their own, a rate of 3.6%. The 5.8-point gap corresponds to about 64 extra orders across the 1,105 reminded carts; that is the reminder’s contribution in this period. The other 40 of the 104 recovered orders would probably have come in without a message.

Holding back a control group has a cost too: had the 195 held-back carts received a reminder, about 11 more orders would have come in at the same gap. So rather than keeping the group running permanently, set it up for a few weeks at a few points in the year. The control group in the example has only 7 orders; a swing of two or three orders changes the result noticeably. Do not decide on one period’s figure; read the results of several periods together. We explain a practical way to split groups at random, and how to read results from small groups, in the last section of our customer segmentation guide.

If you cannot hold back a control group

If your tool does not let you leave some carts without a message, the best comparison available to you is the period before you switched the reminders on. For the four normal weeks before the launch, note from your platform’s report what share of abandoned carts later turned into orders. If the report does not provide that, look in your analytics at what share of shoppers who added items to their cart went on to buy. Recalculate the same rate from the same source for the four normal weeks after the launch. The increase gives a rough estimate of the reminder’s contribution.

The estimate is rough because prices, ad budget and the mix of visitors also change between the two periods. If the reminders have been running for months, the seasonal gap between the two periods weakens the estimate further. And because reminders only go to reachable carts, a rate calculated over all carts barely moves. The example store sees around 4,800 abandoned carts in four weeks; among them, 64 extra orders lift the rate by only about 1.3 points. A small change on the site can move the rate that much as well. If you changed prices, shipping fees or the checkout during those weeks, do not credit the difference to the reminder.

Read sale weeks and paydays separately

Whichever method you use, mark the weeks you compare on a calendar. 11.11 and the late-November sales, your own campaign weeks and paydays can change both the number of carts and how quickly customers come back on their own. Do not put these days in the same comparison as normal weeks; keep them on a separate row. A control group largely evens out this effect, because both groups live through the same days; still, read the results of sale weeks separately, as the share of customers returning without a message may also be different then.

Work out the cost per extra order and the profit left

The cost of the reminder has two items: the message fee and the discount you give. In its WhatsApp pricing documentation, Meta states that it charges for a template message only when the message is delivered, and that all marketing templates are charged. Its template categorization documentation lists a message reminding people of items left in their cart among the marketing template examples. For the message item, enter the amount your tool actually bills you for these reminders; if the tool charges per delivered message, that amount is the number of delivered reminders multiplied by the unit price.

The example store includes a 5% discount code in the reminder. In the four weeks with the control group, reminders were delivered to 1,040 of the 1,105 carts they were sent to, and the code was used on 70 of the 104 recovered orders. The average order is 1,150 TL and the gross margin is 38%. The calculation runs in four steps:

1. The gross profit on the 64 extra orders is 64 × 1,150 TL × 0.38 = 27,968 TL.
2. The discount costs 57.50 TL on each of the 70 orders where the code was used, 4,025 TL in total. Calculate this item on all 70 orders, because the code can be used not only by the customers behind the extra orders but also by those who would have come without a message.
3. Before the message fee, 27,968 − 4,025 = 23,943 TL is left.
4. Divide this amount by the number of billed messages, 1,040 delivered messages in the example, and you get about 23.02 TL. If what you pay for one reminder is below that, the flow made a profit in this period.

Discount and message cost per extra order

Calculate the cost per order with two denominators as well. Divide the 4,025 TL discount by the 104 recovered orders and you get 38.70 TL per order; divide it by the 64 extra orders the reminder brought in and the figure rises to about 63 TL. Use the second one when you make decisions, because it shows what an extra order really costs you. Add the message fee too: divide the total you paid for these reminders by 64 to get the message cost per extra order. Compare the sum of the two with the 437 TL of gross profit an order leaves.

Keep the discount here as a single cost item. Decide whether to offer a code at all, and how many times more orders a message with a code must bring in than one without, with the multiplier in our article on the profit math of a discount code in cart messages. If you sell in a category with a high return rate, subtract your average return losses from the gross profit of the extra orders in your own profit calculation.

Break the result down by basket value, customer, device and time

A single recovery rate hides where the reminder works and where it does not. Read the result from the four angles below, and in each breakdown take the numerator and the denominator from the same group: find the rate for carts above 750 TL by dividing only by the reachable carts above 750 TL. You do not need these breakdowns every period; look at them when the overall rate changes unexpectedly or before you change a setting. If your tool’s report has no breakdowns and you can export cart and order records, you can do the same calculation in a spreadsheet.

Basket value

When the overall rate falls, basket value is the first breakdown to check. Take this example. In the first period, 30 of 500 carts under 750 TL (6%) and 84 of 700 carts above it (12%) turn into orders; the overall rate is 9.5%, with 114 orders from 1,200 carts. In the second period, new visitors brought in by ads go for lower-priced products: 60 of 1,000 carts under 750 TL (6%) and 60 of 500 carts above it (12%) come back. Recovered orders rise from 114 to 120 and neither group’s rate moves, yet the overall rate falls to 8%. In a drop like this, leave the waiting time and the wording alone; what pulls the rate down is the growing share of low-value carts.

Returning and new customers

Count a customer who has ordered before as returning, and one who has never ordered as new. A returning customer who knows your store may come back to the cart more often even without a reminder. If you have a control group, calculate the rate of coming back without a message separately for returning and new customers. If the gap between those who got the message and those who did not is small for returning customers, the reminder there mostly accompanies orders that would have come anyway, and any code they receive turns straight into cost. If the gap is large for new customers, that is where the reminder does its real work.

Device

A cart left on a desktop still gets its reminder on the customer’s phone. A customer who opens the link there and finds an empty cart may give up rather than search for the products again. If carts left on a desktop come back noticeably less often than carts left on a phone, run a test: using your own number, add products to the cart on a computer and leave, open the link in the reminder on your phone and check whether the cart opens full. If it does, take the order all the way through on the same phone. Our article on high clicks and low sales in WhatsApp campaigns shows how to check whether the login and saved details carry over to the browser the link opens in.

Time the cart was left

Split carts into four groups by the time they were left: morning, daytime, evening and night. If carts left at night come back noticeably less often than those left during the day, first check what time the reminder reaches the customer. Review the waiting time and sending hours against the criteria in our guide to cart message timing.

Put opt-outs, blocks and replies next to sales

A change can raise recovered orders while also wearing customers out. To see this, write three numbers next to the sales figures every period: customers asking to leave the list, customers who blocked you if your tool shows it, and customers who replied to the reminder. Writing them per 1,000 delivered reminders lets you compare periods with different sending volumes.

In cart reminders, every opt-out request has a cost that carries forward. A customer who opts out will not get reminders for their later carts either, and your pool of reachable carts shrinks. In Pushouse, a customer who does not want the reminders can leave the list with one message. If opt-out requests climb noticeably while recovered orders rise, consider reversing the change you made.

The risk of blocks and the clues in replies

Blocks carry a wider risk. WhatsApp’s business messaging policy points out that users can block or report a business, and states that a business’s messaging may be capped in response to that feedback. Such a cap could affect not only cart reminders but all your WhatsApp messages, including order and shipping notifications.

Replies to the reminder, on the other hand, are often the most direct clue to what stopped the customer. At the end of the period, count replies by topic: shipping fee, delivery time, size or color, payment option and “just looking”. If the same question comes up again and again, answer it on the product or cart page; explaining it to each customer by message leaves the obstacle on the site.

Read the reminder tool’s figure next to your analytics

The number of recovered orders your reminder tool shows will most likely not match the orders your analytics credits to WhatsApp. The two tools count the same order by different rules. Analytics decides for itself which traffic source an order is credited to, and it may never see the session of a visitor who rejected cookies. Accept this gap as normal and do not use one figure in place of the other.

What you should expect is for the two figures to move in the same direction from period to period. If one rises noticeably while the other falls, first check your analytics setup and whether the link in the reminder opens the right page on a phone. When you prepare a revenue-by-channel report, do not add up the two figures; the same order may appear in both. We cover how to match the orders in your platform panel with analytics, and which questions to ask about a tool’s results report before choosing it, in our guide to choosing digital marketing tools. Neither figure on its own tells you how many extra orders the reminder brought in; get that from the control group or from the comparison with the earlier period.

Which number is weak, and what should you change?

When the table holds a few periods of figures, change one setting based on the number that stays weak.

• If the ratio of delivered to sent reminders is falling, rewording the message will not close that loss. Check that the phone field in your checkout and sign-up forms captures the number completely and in the right format.
• If the rate per reachable cart is low and the gap to the control group is also small, the reminder is not bringing customers back. First look at when the message arrives; our article on timing the first reminder walks you through the criteria for choosing the waiting time.
• If the rate is high but held-back carts come back at almost the same rate, the reminder mostly reaches customers who were coming back on their own. If you include a code, the discount goes to these customers as well; work out whether to drop the code or lower it with the multiplier in the discount code article.
• If the cost per extra order is getting close to the gross profit an order leaves, shrink or drop the discount first; measure a period without a code before you switch off the flow itself.
• If replies keep raising the same obstacle, such as the shipping fee, delivery time or forced account creation, fix the site first. Our guide to growing online sales shows how to find and remove obstacles on the cart and checkout pages.
• If opt-outs and blocks are rising, look at the time the reminder arrives and at your other messages going out on the same day.

After each change, measure the next normal period with the same table. Change two settings in the same period and you cannot tell which of them moved the result. If you want to set up reminders and follow recovered orders and revenue in your panel, see how the cart recovery module works.