Selzee Dashboard

• the number of visitors reached,
• the number of AI messages,
• conversation length.

You can also browse the progression of past conversations in this tab.

To display detailed information about conversations, press the < button. A conversation can have a status of Active or Inactive depending on whether it is currently in progress. In addition to conversation duration and message count, you can also view a short summary of its progression.

In the case that the chatbot's responses in a given conversation were not in accordance with your expectations, you can use the option of leaving feedback — in the feedback itself you describe the expected response or correction of inaccurate information that the chatbot provided. Feedback is subsequently incorporated into the chatbot's feedback data.

Static — These are all information that are not a product catalogue. Typically categorised here are:

• general information — contact, privacy policy, store information, cookies, etc.
• purchase information — shipping options, payment options, terms & conditions, contract withdrawal, complaints procedure, etc.
• services — app, articles, contest rules, discounts, promotional flyers, news, loyalty programme, etc.

Product catalogue — The complete product catalogue is uploaded (Google Feed, Heureka Feed, etc.) in .json or .xml format. In case the product catalogue needs to be updated, the catalogue needs to be uploaded as a URL link to .xml.

Data can be uploaded in three ways:

• manual upload,
• automatic upload,
• data extraction.

You upload data by clicking the button in the top-right corner of the screen marked with a plus sign. You have a choice of three data upload options — manual upload, automatic upload, or data extraction.

Data can be uploaded as a file, text, or via a URL link. Store owners on the Shopify and Shoptet platforms can have this part of the process automated via the Import from Shopify and Import from Shoptet functionality.

When adding a data source, it is necessary to select the chatbot into which the data will be uploaded, and the type of data.

In the Data tab you can filter uploaded data by status, type, and the chatbot they are assigned to. Through the preview icon, data sources can be individually opened and their content browsed. If you are not sure whether a specific piece of information is already among the data the chatbot draws from, verify it using the Test Search feature.

The Appearance category enables visual customisation of the chatbot. Here you configure:

• Agent name — What will your chatbot be called?
• Agent personality — A short description under the name that further specifies the character of your chatbot.
• Gender — When selecting the gender male, your chatbot will respond in the masculine gender. When selecting the gender female, your chatbot will respond in the feminine gender.
• Welcome message — This message is displayed to every user who decides to use the chatbot's chat window. It usually serves as a short introduction of the chatbot and a welcome to the user in the e-shop. The text that the chatbot displays is editable and supports Markdown.
• Starter prompts — These are short message suggestions that the user will see before they ask the chatbot a question. These prompts should address the most common questions, should direct toward recommending specific products or categories, or possibly address questions regarding the selection itself. After selecting one of the buttons in the starter prompts, this text will be sent in the same format to the chatbot. The text must be understandable and direct toward a specific action.
• Currency — The currency displayed in this field will be taken into account for product amounts.
• Brand colour — Colour selection is possible in two ways: Color Picker or import of a hexadecimal colour code. The shade of colour you choose will also be used when displaying the colour of starter prompt buttons and quick reply suggestion buttons.
• Avatar image — Supported formats: JPEG, JPG, PNG, GIF.
• Chat button — In the case that this option is enabled, when the chatbot window is closed, the chat button text will appear next to the chatbot image. When this option is disabled, after closing the chat window only the chatbot photo will be displayed on the page without the chat button text.
• Input placeholder — Text displayed in the text input field.
• Position — Offset of the chatbot box placement, specified in px (e.g. 24px).
• Custom CSS — Add custom CSS to customise the chatbox widget style (font, text colour, etc.).

The Intelligence category serves to determine the response logic and boundaries that the chatbot must follow. Here you manage:

• Language — The chatbot will preferentially communicate with users in the language selected in this setting. In the case that a user communicates in a language other than the selected one, the chatbot will adapt to the language of communication.
• System instructions — Core setting that determines how the chatbot should respond, react to certain situations, what procedures it must follow, and what situations it must avoid.

In the case that the chatbot takes into account information other than the desired one in its responses, you have the option to reshape the logic of its responses with new system instructions.

• Tone of voice — Set guidelines for the tone in which the chatbot will communicate with the user. Consider here the level of formality of responses (informal address, formal address) and adapt the expression to your brand voice.
• Qualification — Describe the demographics of your audience, customer profile, problems that trouble them, and solutions they are looking for. Determine what questions the chatbot should ask the user for their qualification.
• Escalation — Define cases in which the required next step is escalation to a customer service operator, phone contact, or email. The AI, based on keywords such as "I want to connect with a person" or "I don't want to talk to a robot", evaluates the situation and escalates the communication as needed.
• Personality — Choose one of the options: formal, balanced, friendly.
• Product preferences — In this field, take into account brands, products, and product categories that you want to prioritise in product recommendations. When choosing among multiple suitable products, products listed in this window will have priority.
• Quick reply suggestions — Instructions for generating quick reply suggestions directed from the user to the chatbot after each generated chatbot message. They help maintain a smooth conversation flow and serve to prevent entry barriers into the sales process. In this setting, describe the logic by which the chatbot generates the content of clickable buttons.
• Products to recommend — Choose the maximum number of products that will be displayed to users when a need matches.

Proactive popups — Enable proactive outreach using the toggle button. When this feature is enabled, the chatbot, based on the user's activity on the page, will reach out to the user at the right moment — welcome on the page, product comparison, selection guidance, cart abandonment, etc.

The Actions category offers options for interaction with third parties.

• Operator handoff — Keep this setting enabled in the case that the chatbot serves on your e-shop as a channel designated for direct contact with the customer and you plan to actively enter customer conversations and communicate with users. During the need for handoff, AI responses will be deactivated and your team will receive a notification (dashboard — sound and visual notification, browser notification, email). The suitability of escalation is evaluated by the AI, which reacts to prompts such as "I don't want to talk to a robot", "I want a person", etc.
• Cart — The Actions category also enables adding products to the cart directly through chat. If you use the Selzee add-on for your Upgates / Shoptet / Shopify e-shop, you can use this option without additional implementation. For other platforms, custom technical integration is required, which through global objects and API calls will enable identification of your system and communication with the cart (you can find details in the technical documentation).
• Order tracking — Through order tracking, the chatbot can handle requests of the type "Where is my order?" directly in the chat. The customer, to verify the status of their order, enters the order number and a verification detail (email) and the chatbot instantly displays its status. If you use the Selzee add-on for your Upgates / Shoptet / Shopify e-shop, you can use this option without additional implementation. For other platforms, it is necessary to make available an API endpoint that retrieves the current status based on order number and email; the detailed procedure is in the technical documentation.

At the top of the dashboard you will find three key indicators: total number of conversations, number of AI messages sent by the chatbot, and the estimated number of hours saved in customer support.

The section displays the visitor's journey from arriving on the page to an active conversation. You see the number of visitors, the number of popup impressions (including the percentage share of total visitors), the number of chat opens, and the number of messages actually sent. These values allow you to identify at which point the biggest drops in interest occur.

This table shows the performance of individual proactive popup versions. For each popup you see how many times it was displayed (Displayed), how many times a visitor opened it (Opened), and the resulting open rate (Open Rate). Also listed is the source of conversations — via popup or directly via the chat button.

This section compares the purchasing behaviour of visitors who used the chat with those who did not use it. Tracked metrics include:

• Conv. Rate (conversion rate) — Separately for users with chat and without chat, and the overall conversion rate of the e-shop.
• Chat Uplift (Δ conv.) — The difference in conversion rate between users with chat and without it. Shows by how many percent the chat increases the probability of purchase.
• AOV (average order value) — Comparison of average order value for customers who chatted, compared to those who did not communicate via chat.
• ARPU (average revenue per user) — Average revenue generated by one user, separately for the chat and non-chat segments.
• Estimated Revenue Uplift — Estimated additional revenue directly attributed to the chatbot compared to the baseline value of the site's conversion rate.

The section provides a deeper look at the quality and content of conversations. It includes an overview of average conversation length, in-chat conversion rate, the percentage of cases when operator escalation was needed (Human Needed), click-throughs (Click-throughs), and feedback (Reactions). In addition, you see the visitor mood distribution (Visitor Mood), conversation outcome (Conversation Outcome), most frequent user intents (Most Frequent Intents), most common complaints (Most Common Complaints), and the products that are most discussed and most clicked.