Cross-Sell / Complementary Offer

Build this when order data shows some buyers of the main product later buying one specific complementary product on their own, while most never learn it exists. Before launch, verify the compatibility of the main product and complement pairing, and track purchase, return and product-active status, what the customer already owns, and the separate permissions for push, email and WhatsApp on one customer identity. If the complement is not selling because the main product has not yet been delivered or put to use, tie the wait to delivery and first use; if compatibility cannot be verified, do not build an offer on a complementary claim.

Incremental complementary purchase rate vs control

What counts in a cross-sell is how many more complementary purchases the offered group makes than the control group that gets no offer; purchases that would have happened anyway, and clicks, do not.

How it is counted: Take all main-product buyers assigned to the two groups after the three-day checks as the denominator; the numerator is those who buy the defined complementary product within the same follow-up window after assignment, and the difference between the two rates is what the journey added. Delivery or clicks on a message do not change the denominator.

Also track

• Contribution per touchTrack complementary purchases after the push, the email and the WhatsApp message separately; if the WhatsApp step adds no sales, consider ending the cascade after two touches.
• Share closed for ownershipRecords closed without a message at the three-day check because the complement was already owned, as a share of entries; if it is near zero, verify that ownership data is really being read.
• Return and inactivity exitsExits caused by a returned or inactive main product, as a share of entries; if they rise, the problem is the main product, not the cross-sell.
• Unsubscribes and complaintsWatch push opt-outs, email unsubscribes and WhatsApp complaints or blocks per channel; the cost of the direct channel shows up here.

• Trigger and entryBuying the main product starts the journey, and the offer window opens three days later. Randomly assign the eligible customers who pass the three-day checks to the journey group or to a control group that receives no offer from this flow.
• Who must not enterA customer who already owns the complementary product gets no offer; if ownership shows at the three-day check, the journey closes without a message. Returned or inactive main products stay out too. Scope is one purchased main product with one instance; a repeat purchase of the same product belongs to Predicted Replenishment.
• Timing and channelsThree days after purchase the journey checks that the product is active and unreturned and that the complement has not been bought, then sends a push. If there is no purchase after 1 day an email follows; after 2 more days, with WhatsApp permission, a WhatsApp message; the final check comes 2 days later.
• Exit and handoffOnce the complementary product is bought the journey ends and queued messages are cancelled. If the main product is returned or inactive, communication stops. Without WhatsApp permission, or with no purchase at the final check, no further offers go out; there is no handoff to another journey.
• Frequency and consentAt most three offers per main product: push, email and, only with permission, WhatsApp. Push and email are also used only where that channel's consent exists; the WhatsApp step does not open without passing the permission check.

1. 1Failing to beat the control group's complementary purchase rate means the offer pulled forward purchases that would have come anyway or had no effect; opens and clicks do not change that. A high margin alone does not justify the journey either, because the rule is to explain what genuinely completes the first product.
2. 2This result holds for this main product and complement pairing, the three-day wait and the push, email, WhatsApp order; do not transfer it to another product pair. A repeat purchase of the same product is Predicted Replenishment's outcome, a cart left behind belongs to Cart Reminder, and a benefit redeemed through Unused Loyalty Reward is not credited here.
3. 3When the gap is small or unclear, compare each group's customers and complementary purchases in the A/B test significance calculator; if the gap is not significant, let more main-product sales accumulate before changing the cascade.