A useful return loop remembers intent, waits for a real trigger, and takes the customer back to the next step
A shopper finds the right sneakers, selects size 43, and leaves to think. The next day, the store sends a discount notification. He taps it and lands on the homepage. Now he has to remember the model, find the product again, and select the size again.
The store knew enough to send a message. It forgot everything that would have made the message useful.
Duolingo handles the return differently. A reminder takes the learner back to the next lesson, with the streak and progress intact. The useful idea for e-commerce sits underneath the owl, the badges, and the cheerful pressure: preserve the customer’s context, wait for a relevant reason to return, and reduce the next action to one clear step.
The owl didn’t create 23% growth on its own
Duolingo reported 58.7 million average daily active users in Q2 2026, up from 47.7 million a year earlier. That’s 23% year-on-year growth. The company attributed the acceleration to three factors: product changes, marketing, and a one-time Streak Revival campaign.
The attribution matters. A retailer can’t copy one reminder and expect the same result. The repeatable part is the operating system behind it: a reason to return, stored progress, a low-friction next step, and experiments that separate product effect from seasonality.
Duolingo’s leading retention metric makes that logic explicit. Current User Retention Rate, or CURR, measures the share of current daily users who return the next day. A “current” user has also been active at least once during the previous seven days. In Q2 2026, CURR reached a record 84%.
For a store, the equivalent metric depends on the buying cycle. Grocery and pet food may run on days or weeks. Furniture may run on years. A single daily-activity target would flatten those very different rhythms into one number.
One lesson beat the full daily plan
Duolingo once required learners to complete their full daily goal to extend a streak. The team later separated the two mechanics and let a single lesson keep the streak alive.
The experiment produced three relative changes: Day 14 retention rose 3.3%, overall daily active users rose 1%, and the share of daily users with a streak rose 10.5% during the first 20 days. Daily-goal completion fell. That trade-off is the useful part of the story. A lower threshold improved the return habit even though a nearby engagement metric moved in the wrong direction.
An e-commerce return loop needs the same discipline. The smallest useful action might be checking whether size 43 is back, reopening a saved comparison, repeating the last order, or seeing the current delivery status. Sending someone through the whole catalogue again raises the threshold at exactly the wrong moment.
A sofa has no use for a daily streak
Languages can be practiced every day. Sofas are bought less often. The right return frequency comes from the category and the customer’s unfinished job.
Consumables create predictable windows for replenishment. Fashion and electronics often create a shorter consideration period: people revisit a product, wait for a price change, compare specifications, or watch for a size to return. After purchase, delivery, installation, supplies, warranty, and repair create a different set of reasons to come back.
This is why “hasn’t visited recently” is weak targeting. It knows a date. A useful trigger knows what changed for the customer.
Return the value the customer already created
Duolingo’s Streak Revival campaign restored a learner’s longest streak after opt-in and three completed lessons. In June 2026, 15.4 million learners revived a streak; nearly 8 million had no active streak when the campaign began.
The campaign worked with value that users had already earned. Retail products hold similar customer-created value: a saved list, a configured product, a basket, a comparison history, loyalty status, or a watched price.
A five-item basket is already a detailed brief. Replacing it with a generic bestseller page forces the customer to repeat work the product has already observed.
A notification should resume the task
The notification is only the door. The destination determines whether the customer continues or starts over.
A useful deep link opens the exact product, size, color, saved filter set, comparison, or basket state. To make that possible, the product must connect customer identity with scenario state: product ID, SKU, selected variant, price at the time of interest, availability, and notification consent.
Events such as product_viewed, favorite_added, cart_updated, and order_completed show where the customer stopped. A new signal, such as stock returning or a watched price changing, can then trigger the message. The push notification sits at the end of that chain, not at the beginning.
Five layers sit between a product page and a useful push
Take the sneaker example. The return loop needs five layers:
- Events: the shopper viewed a specific model and selected size 43.
- Profile: the system connects the shopper with the model, color, size, permitted channel, and consent.
- Decision: size 43 returns to stock, creating a concrete reason to send a message.
- Delivery: the message opens that model with size 43 already selected, while frequency rules prevent notification fatigue.
- Experiment: only part of the eligible audience sees the new flow; the control group keeps the previous experience.
The control group should exist before launch. Orders may rise because of a notification, a promotion, payday, weather, or season. Showing the flow to everyone removes the reference point and leaves the team with a result it can celebrate but can’t explain.
Start with an abandoned task, not an inactive user
The first candidate should combine visible intent with a new, useful signal: a size is available again, a watched price has changed, or a replenishment window has arrived.
Next, define the smallest action that resumes the task. For apparel, that may be a deep link to one SKU and size. For recurring goods, it may be a prefilled repeat order. For complex products, it may be the saved comparison rather than the catalogue.
Then write one hypothesis and one business outcome before building the flow. For example: a back-in-stock message for the selected size will increase completed orders among shoppers who previously viewed that SKU. That statement is narrow enough to test and broad enough to matter.
More DAU can still mean a worse product
Duolingo once tested an offline promotion for its paid plan. The experiment increased subscription purchases and produced a slight revenue lift. It also reduced retention and daily active users, so the team shut it down.
Retail teams face the same trap. A push can increase visits. A discount can increase orders. Together they can also erode margin and exhaust the audience.
Open rate is therefore a delivery metric, not a business verdict. The measurement chain should cover:
- whether the customer returned to the intended scenario;
- whether the customer completed the target action or order;
- what happened to revenue and margin;
- whether unsubscribes, notification opt-outs, returns, or support requests increased.
DAU won’t fit every store either. A closer analogue to CURR may be the share of customers who return to a saved product, basket, or repeat order within the category’s normal time window. For infrequent purchases, saved-list conversion and post-purchase support may tell the team more than daily visits ever will.
The rule: match the loop to the buying rhythm
Duolingo’s loop is simple: remember where the learner stopped, wait for a relevant reason to return, take the learner to the next action, and test the effect against a control group.
E-commerce can use the same order of operations. The frequency changes by category. Pet food may create a weekly or monthly signal. Electronics may create a signal when the price moves. Apparel may create one when the right size returns.
The product should continue the customer’s existing task. A generic campaign can wait.
Qtim designs and develops e-commerce products around customer journeys, integrations, and measurable business outcomes. Start with one interrupted scenario, connect its events to stored context, and test whether the new route improves completed orders without damaging margin or customer tolerance.
Top comments (0)