What to prepare before installing
First decide who is going to answer. Live chat makes sense if someone really sits at the panel during the hours your customers write, because a green dot beside an empty desk only irritates them. A bot answers by itself. But it has to know your products and rules, otherwise it starts pointing to the terms and conditions or guessing. And if for now it is enough that the customer leaves a question and you reply in the morning, Shoper has a built-in “Zapytaj o produkt” (Ask about the product) module that you add to the product page.
The second thing is the template technology. Shoper has Storefront templates and older ones, in the RWD technology, and that decides where you can paste code at all. Instructions for Storefront usually start in “Wygląd i treści > Wygląd sklepu > Edycja szablonu graficznego” (Appearance and content > Store appearance > Graphic template editing). If you do not know which technology your store has, ask Shoper support before you start copying forum guides written for the old template.
Also write down the rules you or the bot will answer from, meaning delivery with the free shipping threshold, and returns, ideally the way you explain them to customers over the phone. Add the hours when someone replies, because without that even a well-connected chat ends up giving vague answers.
Which method your template and plan allow
The simplest route is an app from the Shoper App Store, which you install in “Dodatki i integracje > Aplikacje” (Add-ons and integrations > Apps) and then configure in “Moje aplikacje” (My apps). The catalogue includes Smartsupp, a live chat with a chatbot and an AI assistant, and AI bots connected to store data, such as Chat AI from Witly or Sensbot AI Chatbot. Shoper advises Storefront owners to look in the App Store for apps marked as Storefront. Some appear in the store on their own after activation, while others, like Smartsupp, still have to be enabled as a module in Shoper Visual Editor, in the “Pop-upy i dodatki” (Pop-ups and add-ons) tab.
Before installing, read the pricing in the app description. Some charge by the number of conversations per month, so in December the bill can look different from February.
When the chat provider gives you only a piece of code, Google Tag Manager is what is left on Storefront. Shoper has a built-in GTM module in the template editor, so in the store admin you paste only the container ID, and you add the chat code itself in the GTM panel.
Custom JavaScript can be entered directly into the Storefront template through a custom module, but Shoper makes custom modules available only on the Shoper Premium plan, so on the other plans this route is closed. On an older template it is simpler, because you paste the code in “Integracje własne” (Custom integrations), in the page header or footer field. In Storefront those fields do nothing, because, as Shoper writes, the template in this technology ignores content entered into Custom integrations.
CartBot is not an app from the Shoper App Store. On Storefront we install it through GTM, and on an older template through the header field in Custom integrations, and in both cases it is the same script, which recognizes the template generation by itself.
How to add a chat in Shoper through GTM
Before you start, check whether the chat provider has a separate GTM instruction. Google warns that Tag Manager minifies the code of custom HTML tags, and in our test it also dropped extra attributes of the script tag, so code that works pasted straight into a page may fail to run through GTM. The Shoper instructions together with Google's help give this order:
- In GTM create a container for the site and copy its ID in the GTM-XXXXXXXX format.
- In the Shoper admin go to “Wygląd i treści > Wygląd sklepu > Edycja szablonu graficznego” and click “Edytuj układ i styl szablonu” (Edit template layout and style).
- In the “Integracje” (Integrations) tab, in the “Wbudowane” (Built-in) section, open the “Google Analytics and Google Tag Manager” module, paste the ID, turn on sending data through GTM and the module's activity, and save.
- In GTM add a tag of the “Custom HTML” type with the chat code and a trigger for all pages.
- Publish the changes with the “Submit” button and give the store ten or fifteen minutes, because in our test a new container version reached Storefront with just that kind of delay.
A chatbot on Shoper has to know the products
The bubble alone is half the job. A live chat works straight away, because you do the answering, but an AI bot needs products with prices and availability, and it gets them through the store API or from a product file. App Store bots such as Witly describe the connection to Shoper data as part of the installation, and a provider from outside the catalogue may ask you for a Webapi account.
You create such an account in “Ustawienia > Ogólne > Administratorzy sklepu” (Settings > General > Store administrators). Through “Więcej” (More) you add an administrator group with the “dostęp do webapi” (webapi access) type, and in the “Uprawnienia” (Permissions) tab you tick only the sections the integration needs. Then in the “Administratorzy” (Administrators) tab you create an account for the provider and hand over the login details. Shoper advises limiting the permissions of such a group to the necessary minimum, and a bot that is only meant to talk about products does not need the right to change prices or orders.
For CartBot we ask for a read-only account with access to the product range and to sales. From the product range the bot knows the catalogue with variants. Sales lets it give the customer an order status from the number and email address, and count how many orders came in after a conversation. Setup and day-to-day running are included in a subscription that starts at PLN 50 net a month, and we work out the exact amount from the size of the catalogue and the traffic.
Test on a phone before the first customer
Once the bubble is there, walk through the store on your phone like a customer. See whether the chat window covers the add-to-cart button and whether you can close it with a thumb without aiming. Then ask the bot about a product you do not carry and check whether it admits it does not know, instead of offering something similar as if it were the same item.
The most revealing test is a product with variants, for example a plate in three colours, so ask the bot to add it to the cart and open the store's cart right away. The bot should ask about the colour or send you to the product page, where the customer picks it themselves. A Storefront store can refuse to add a product without a chosen variant in a way a chat easily misses, and then in the conversation you see “added” while the cart counter in the header stays where it was. That is why you should look in the cart after every attempt.
Finally, change the price of one product in the admin and note how long it takes the bot to quote the new one. A bot that pulls the catalogue on a schedule sees changes with a delay, so ask the provider how often it refreshes the data.
Taking over a conversation and contacting support
A bot cannot handle every matter, because a complaint or an upset customer needs a person with the authority to decide. Check with the provider whether, when you take over a conversation, you see its whole history, and what the bot tells the customer when nobody is in the store.
After hours the bot should collect a contact. The response time it quotes to the customer has to match your schedule, so if you sit down to the inbox in the morning before packing, it should say exactly that. When the “Zapytaj o produkt” module runs alongside, its messages go to the address in the “Główny adres e-mail” (Main email address) field in “Ustawienia > Ogólne > Dane firmy” (Settings > General > Company details). Questions from two channels then flow into two places, so decide who reads which.
If you want to see how a bot copes with your products, send us the address of your Shoper store, and we will show it running on your own catalogue before you pay anything.
