With the Conversations report, you can view the history of all exchanges that have taken place on your site.
To access it from the iAdvize administration, go to the Reports > Conversations tab.
In this report, you will find all exchanges conducted with your visitors, whether they are conversations with your Assistant, workflows, or your agents.
The List of visitors button, above the conversation log, opens the project's visitor list. It is only enabled when a single project is selected in the Projects filter.
Metrics
| Closed conversations |
Number of conversations closed during the period, taking filters into account The percentage shows the share of filtered conversations out of all conversations closed during the period, for the selected projects and channels. Other filters do not change this total. |
| Conversations volume | Trend in the number of conversations closed over the period, by hour, day or month. Click a point on the chart to refocus the whole report on that period, with a finer level of detail (from month to day, from day to hour). |
| Date | Closing date of the conversation |
| Duration | Total duration of the conversation. |
| Conversation ID | Conversation identifier. You can click on it to quickly copy/paste and share it with support for example. |
| Agent | Name of the respondent who closed the conversation (Assistant, workflow, or agent). If another respondent was involved in the conversation, a "+1" will be displayed. (+X depending on the number of additional respondents involved) |
| CSAT (Customer Satisfaction) |
Score indicating customer satisfaction regarding the conversation.
If a comment has been added, it will be visible when hovering over the message icon. |
| NPS (Net Promoter Score) | Score indicating the likelihood that the customer would recommend your company. |
| Revenue | Amount of the sale attributed to the conversation, meaning the full transaction total when at least one discussed product was purchased. |
Filters
Filters allow you to refine your analysis based on a specific period or a particular project. You can add additional filters to further narrow down your results.
When you select several values in the same filter, the conversations displayed match at least one of these values. When you add several filters, they match all the filters at once. Two exceptions, detailed below: the Only mode of the Respondents filter, and the Keywords filter.
Respondents
This filter isolates conversations by the type of respondent who handled them: Agent, AI Shopping Assistant or Workflow.
Two modes are available at the top of the filter:
| Mode | Definition |
| Contains | The conversation involved at least one of the selected respondents, whatever other respondents took part. |
| Only | The conversation involved the selected respondents only. No other respondent type took part. |
The two modes cannot be combined: applying one replaces the other. In Only mode, the filter shows as Respondents (only).
Resolution status
This applies only to your conversations handled by your AI Shopping Assistant (fully or partially automated). Conversations handled only by your agents have no status: they appear under Unclassified.
It allows you to analyse the quality of the responses provided by your Assistant and assess the extent to which it was able to address the visitor's request.
The level is determined automatically by artificial intelligence after the closure of each conversation, based on the entirety of the exchanges between your Assistant and the visitor.
Here are the three possible resolution statuses:
| Status | Definition |
|---|---|
| Resolved | The Assistant answered the visitor. The answer is either complete, or useful but still requiring an action. |
| Missing information or transfer | The Assistant gave no solution: no matching product or information was found, or the conversation was handed over to a human agent. |
| Unclassified | Everything else: an answer that missed the question, or a conversation with nothing to solve for your brand. This value also includes conversations with no status, such as those handled only by your agents. |
Automation level
| Value | Definition |
|---|---|
| Fully automated | The conversation was handled entirely by an AI Assistant (or a workflow), with no human intervention. |
| Partially automated | The conversation was handled both by the Assistant (or a workflow) and by a human agent. |
| Not automated | The conversation was handled exclusively by a human agent. |
Assisted sale
This applies to all your conversations, whoever the respondent: Assistant, workflow or agent.
| Value | Definition |
|---|---|
| Yes | The conversation generated at least one sale. |
| No | The conversation did not generate any sale. |
Pre-sale / Post-sale
The classification is assigned automatically by artificial intelligence after analysing the entire conversation, based on the category detected in the exchanges.
| Value | Definition |
|---|---|
| Pre-sale |
Value assigned by default to any conversation that does not explicitly relate to after-sales. The visitor has not yet placed an order. They are looking to gather information or make a purchase decision.
The following conversations are classified as pre-sale:
A visitor who mentions an "order", a "delivery", or a "warranty" in a general sense remains classified as pre-sale, as long as they are not referring to a purchase already made. For example, "can I return it if it doesn't suit me?" is a pre-sale question: it is not a return request for an item already received.
|
| Post-sale |
The visitor is referring to an order or a purchase already made, or explicitly expresses the wish to contact after-sales service to report a problem.
The following conversations are classified as post-sale:
|
Assistant Topics
This applies only to conversations handled by your AI Shopping Assistant.
The topic of each conversation is assigned automatically by artificial intelligence after it closes, from 17 topics: for example Order status, Questions about a product, Availability and stock or Discount and special offers. A conversation whose topic could not be determined is classified under No identified topic.
To use it, click Add a filter, then in the Automation section, choose Assistant Topics.
Keywords
This filter displays the conversations whose messages contain the words you search for. The search covers the visitor's messages and the respondents' messages (agent, Assistant or workflow).
To use it, click Add a filter, then in the Conversations section, choose Keywords. Type your search in the Type a keyword field and confirm. The filter shows as Keyword.
If you type several words, the conversations displayed contain all of them, wherever they appear in the conversation. You can also add several searches: unlike the other filters, a conversation must then match all the searches to be displayed.
Case does not matter, accents do: "café" and "cafe" are two different searches.
For a more precise search, you can use the following special characters:
| Character | Effect | Example |
|---|---|---|
| * | Finds the words starting with the letters placed before the *. | order* finds order, orders, ordering, ordered… |
| + | All the words must be present. This is the default behaviour: a space has the same effect. | delivery + late |
| | | At least one of the words must be present. | hello | hi |
| - | Excludes the conversations containing the word placed right after it. | delivery -late finds the conversations containing delivery but not late. |
| " " | Searches for an exact sequence of words, in this order, within a single message. | "user account" |
| ( ) | Groups part of the search so it is processed first. | (order | ordered) "delivery date" finds the conversations containing order or ordered, and the sequence delivery date. |
A few searches to copy and paste:
| To find | Search |
|---|---|
| Bug reports | cookie | browser | bug | "error message" |
| Questions about your product pages | "product page" | "product information" | descript* | dimensions |
| Unhappy customers | "buy somewhere else" | unacceptable | angry | unhappy |
Conversation details
By clicking on "Read" or on the conversation banner, you can view the exchanges between your visitor and the respondent.
Messages in grey are from your visitor, in black from the respondent.
You can view elements not visible to the visitor such as: the engagement strategy triggered, the visitor's page change during the conversation with the URL, transfers made, etc.
To view the details of a conversation, click on the icon.
You will have access to the following elements:
- Conversation ID
- Transactions
- CSAT and open comment
- NPS
- Context: conversation status (pending, ongoing, snoozed or closed) and its date, duration, project, channel, respondents who took part in the conversation (including users deleted since), discussion types, notes left by agents
- Visitor information: last name, first name, email, phone, address, city, postal code, country and shopper IDs (internal and external), or username and profile page for conversations outside your site. Your custom data does not appear here. The Edit button opens the visitor’s contact record in a new tab, so you can update it.
- Visitor blocking: if you no longer wish to offer iAdvize to a visitor (repeated contacts, abusive language, etc.), you can block them. This option is reserved for administrators and managers, for conversations on your site. The block applies immediately, with no confirmation: turn off the Block visitor toggle to lift it.
If the visitor changes computer, browser, or deletes their cookies, we will no longer be able to identify them and apply a block. If they manage to start a new conversation, you will need to block the visitor again from that conversation.
Analysing messages
Conversation starters
If the visitor's first message is preceded by a star icon, it is a starter from your Conversation Starters widget.
Response analysis
The "Analyse" button appears below your Shopping Assistant's responses. It allows you to understand how your Assistant generated its response: the interpretation, the knowledge used, and any custom behaviour triggered, etc.
The Shopping Assistant generates its responses in real time, drawing on:
- the most recent artificial intelligence models and a general knowledge base
- the conversation context
- the page context (if it is a product or category page)
- the data available in your knowledge bases (product catalogue and useful content)
This operating mode gives it genuine flexibility: it can rephrase its responses, adapt its language, or suggest new formulations, as long as they remain faithful to the information at its disposal.
Is the interpretation incomplete or incorrect?
You can share your feedback by clicking on Report below the interpretation, then write a comment giving us as much detail as possible.
Your feedback will be sent to our Product team and will contribute to the continuous improvement of our AI Shopping Assistant.
Adding or editing knowledge
If your Assistant's answer is incomplete, you can fix its knowledge directly from the analysis panel:
- Add, next to Knowledge: create new helpful content. Choose the knowledge base, fill in the Topic and Content, then click Save. This content will apply to all your products. The button appears when the answer relies on at least one helpful content entry.
- Edit, on a Helpful content card: change the topic or the content, or delete the entry. Only content created manually can be edited: content synced from an external tool or a file (Zendesk, Mayday, CSV…) is read-only.
- See in catalog, on a Product card: opens the product sheet in your catalog, in a new tab.
Custom behaviours
If your Shopping Assistant triggered a custom behaviour that you created (and not the default ones), its name will be displayed in the response analysis panel.
You can click on "Edit" to go directly to the Assistant editor in a new tab to view and adjust the behaviour if needed.
Exporting conversations
To export conversations, click on the icon:
You will then be prompted to enter your email address and confirm. Your address is pre-filled; you can change it. An email allowing you to download a .CSV file will be sent to you.
By enabling the "include messages" option, you will then receive a second file containing the conversation content. This option only appears if your profile has the permission to export messages.
The download links in the email are valid for 7 days.