E-commerce brands grow repeat orders and basket size by timing retention to each cohort's rhythm and matching offers to what each cohort buys.
E-commerce brands increase repeat orders and basket size by timing retention to each cohort’s buying rhythm and matching offers to what each cohort already buys, in the channels shoppers use. Whilter’s Snype did this for bigbasket from one system: repeat orders rose 31%, basket size 26% and funnel completion 42%.
Most retention programs treat the three levers as separate projects. One team owns the abandoned-cart flow, another the loyalty program, a third the cross-sell emails. Each has its own calendar and its own creative, so a gain on one rarely carries the others.
Platforms that personalize at the cohort level and act on timing, not only on segments. On Whilter’s e-commerce page, the stack is three engines. Snype handles high-SKU personalized imagery and cohort retention, Konne CX handles conversational support and re-orders, and ElevateOS handles performance acquisition.
The published result is bigbasket’s. As a high-SKU grocery marketplace, bigbasket had been sending generic broadcasts, chasing funnel completion, repeat orders and basket size one campaign at a time. Snype ran personalization and retention from a single run across channels. Funnel completion rose 42%, repeat orders 31% and basket size 26%, measured against bigbasket’s own baseline for the same audience and product line over the engagement window.
The right message at the moment they are ready to reorder. That moment differs by cohort: a household restocking staples every week, a shopper buying gifts twice a year, a customer who orders only during sales. A broadcast on a fixed day reaches all of them at the wrong time except one.
bigbasket’s case page credits timing directly. Retention tuned to each cohort’s rhythm reached shoppers when they were ready to reorder, which is what moved repeat orders 31%, rather than a louder broadcast. In practice that means three things:
The cohort with the longest silence needs its own message. Snype lists activating dormant cohorts among its use cases, and a lapsed shopper needs a different reason to return than a weekly one: a reminder of what they used to buy, not the newest arrivals.
Offers matched to what a cohort is already likely to add. A discount on the whole range buys basket size with margin. A relevant suggestion earns it. At bigbasket, matching the offer to the cohort surfaced items a shopper was already likely to add, so baskets grew 26% without discounting the whole range.
The imagery carries the offer. When the products shown match what that cohort buys, the suggestion reads as useful rather than as an ad. Snype generates that imagery per shopper from your catalogue and first-party data, with product recommendations and offers merged into one brand-safe template. The same imagery helps earlier in the funnel too: when the products shown matched what a cohort buys, bigbasket’s shoppers hesitated less mid-funnel, which is the case page’s read on the 42% lift in completion.
Because they share the same signal. The cohort data that times the repeat order also picks the basket offer and the imagery that keeps a shopper moving through checkout. Split across three tools, that signal is copied, delayed and diluted. Run from one system, the same run that completes a funnel also earns the next order and grows the basket, which is how bigbasket’s case page describes the compounding.
It also makes the results readable. When three campaigns run at once from three teams, nobody can say which one moved repeat orders. One system with one test plan can.
At the moment a shopper has a question, because an unanswered question can turn into an abandoned basket. Konne CX runs WhatsApp, Instagram, web chat and voice as one thread with a shared context, and answers in under 150 milliseconds. Its product page shows the pattern: a shopper asks on Instagram whether an item comes in their size, the reply offers to hold one, and the order is confirmed the next morning on WhatsApp, booked to their usual address, with a payment link in the thread.
That is conversation commerce: the purchase happens inside the thread instead of a search-and-click funnel. For re-orders it removes a step. A shopper who replies “same as last time” can be answered in the same conversation, by an agent who knows what last time was.
As separate metrics, on the same cohorts, against a holdout that keeps your current messages. Repeat-order rate and order interval show whether timing worked. Average basket size and items per order show whether the offers worked. Funnel completion shows whether the imagery reduced hesitation.
Watch the margin alongside the basket. A basket that grows only because of a discount is a cost, not a result. And give retention time: a change in reorder timing shows up over several cycles, not in the first week.
To see your own cohorts mapped onto this model, book an e-commerce growth assessment.
Published 2026-10-14 · Whilter.AI Research
