Why carts get abandoned
70% of carts never turn into an order. The reasons split into structural and situational.
Structural — fixed by optimizing the checkout:
– An unexpected shipping cost, revealed only at the last step
– Mandatory registration before purchase
– A long form with fields nobody needs
– Too few payment options
Situational — handled through personalization:
– The shopper is putting the decision off — they need a reminder
– They are comparing prices — they need a reason to come back (a promo code, a guarantee)
– They got distracted — they need a trigger on the next visit
The metric: cart abandonment rate
Cart Abandonment Rate = 1 − (Completed orders / Add-to-carts)
Example: 1 − (200 / 1000) = 80%
Read it by device and by channel — mobile is typically 10–20 pp worse than desktop.
Recovery strategies
| Tool | Audience | Best moment |
|---|---|---|
| Onsite pop-up | Every visitor, anonymous included | On the next visit, within 7 days |
| Exit-intent pop-up | Visitors leaving the cart page | As the cursor moves to close the tab |
| Email trigger | Signed-in shoppers with marketing consent | 1 hour, 24 hours, 72 hours after abandonment |
| Push notification | Push subscribers | 1–3 hours after |
| Remarketing | Cookied audience | The first 24–72 hours |
Email, push and remarketing run in the channels the retailer already owns — an ESP, a mobile app,
an ad platform. Gravity Field covers the onsite rows of this table, and feeds personalized product
recommendations into the abandoned-cart emails your own ESP sends.
Tip: do not lead with a discount — try a plain reminder first. An instant discount teaches
shoppers to abandon the cart on purpose in order to collect a promo code.
Common mistakes
- One scenario for everyone — a shopper with a $10 cart and a shopper with a $1,000 cart need
different offers - No deduplication — if the shopper did place the order in the end, reminders about the
“abandoned” cart keep arriving - An instant promo code — it devalues the brand and teaches people to postpone the purchase