Abandoned carts in your dashboard
Looking at a photo and closing the tab is not abandonment yet. A cart is abandoned once an item has gone into the basket, and one step further is a started checkout, abandoned by someone who got as far as the address form and gave up there. Count these stages separately. Leaving the form says something different from leaving a cart after checking a price.
In the Tuesday example, 28 of 40 carts ended without a purchase, which is 70 percent. The numbers are invented for the sum and say nothing about your industry. Set a time window too. If someone put shoes in the basket on Monday evening and paid on Wednesday morning, a one-day window counts them as abandoned and a three-day window does not.
Research from the Baymard Institute points to two groups of reasons. Some people were not ready to buy. They wanted to see the total with delivery or left the cart as a note for later. Others wanted to buy and something stopped them. Baymard's research on reasons was done with shoppers in the US, so it tells you what to look for in your own shop, but it cannot explain why your customers leave.
Why do customers abandon their carts?
You have little influence over the first group. The second is more interesting, because that is where the fixable things sit.
One customer sees PLN 149 on the product page, clicks through three screens and only on the last step learns what the courier costs. Another is told to create an account when all they wanted was one pair of socks. Someone makes a typo in the postcode and gets a red box with no hint about what to correct. A bank declines a card. The customer comes back to the shop and instead of a button to try again finds an empty screen.
A separate case is a question that came up on the product page and stayed unanswered. The customer wants to know whether size 42 in this model runs like size 42 in other shoes, or whether the charger fits a laptop from five years ago, and there is nobody to ask. So they put the decision off. Or they go and look for the answer somewhere else.
Walk through a purchase on your phone as a guest
Research will not tell you which obstacle works in your shop, so take your phone and buy something in your own store as someone with no account. Watch the moment you learn the full price with delivery. And the moment you learn how long the parcel will take. If delivery cost depends on what is in the basket, show the rules for working it out earlier, on the product page or in the cart, before the customer reaches the last screen.
At the form, note every place where you have to guess what to type. Make a typo in the postcode and see whether the shop shows which field to correct. Then interrupt the payment and try to get back to the order the way an annoyed customer with a declined card would. Every screen that does not say what to do next goes on the list of fixes.
After a change, compare the same step at a similar level of traffic. Write down whether a campaign was running at the time that brought in completely different people, because then a drop in abandonment may have nothing to do with the shorter form.
An unanswered question halfway through a purchase
It is 22:30 and a customer asks in the chat whether a bed frame will fit under a bed with 25 cm clearance. The product page has no height, so nobody can answer well, neither you in the morning nor a bot straight away. Start with the data. Put dimensions and compatibility with specific models where the customer will see them, even if they have no wish to talk to anyone.
Only then does a chat advisor have something to work with. CartBot looks for answers in your catalogue and in your shop's knowledge base, shows a product in the conversation with a photo and price, and on WooCommerce, Shopify, PrestaShop and Shoper lets the customer add it to the cart without leaving the chat. It cannot fix a form or a payment. If your test showed that people drop out at the postcode field, chat changes nothing there.
Be careful with delivery times. If you order part of your stock from a supplier only after the customer buys, write separate fulfilment rules for it. Otherwise the bot will take the general time from your delivery page and promise dispatch tomorrow on something you are waiting a week for yourself.
Preventing abandonment versus winning a cart back
Everything above is prevention, which means removing the obstacle before the customer leaves. Recovery starts later, when the cart already sits abandoned and you try to remind the customer about it, most often by email.
A reminder email is a separate tool, and you should judge its results on your own numbers. There is no reason to assume in advance that it works worse than chat or that it works on every customer. CartBot does not send such emails, so for that you need another tool.
Check consent before you switch sending on. If the reminder counts as a commercial message or direct marketing, you need the recipient's prior consent, which follows from Article 398 of the Polish Electronic Communications Law. An address typed into an order form does not automatically mean consent to marketing. The legal basis for processing the data is a separate question, assessed under the GDPR.
Test two more cases that are easy to forget, the customer who bought after the first email and the customer who unsubscribed from marketing. In both, further reminders should stop. In your report, keep orders after an email apart from orders after a chat conversation so that you do not count one sale twice.