Analytics
See what customers ask, and what to teach the AI next
EzyConn reports show how many conversations the AI handled, how many went to a person, what customers ask about most and what it could not answer. Then a reply your team wrote becomes the answer it gives next time, once you approve it.
- On every plan
- Every number defined
- Export to CSV
A worked example of the reports screen for the last 30 days: 412 conversations, 296 handled by the AI, 58 escalated to the team and 41 led by a person. Questions are charted by day and grouped by topic with an answer rate for each. Fitting and installation has the lowest answer rate, and three answers written by the team are waiting for approval.
Worked exampleThe shop and every figure are invented. The layout and labels are the portal’s own.
What it measures
Six questions every owner asks, answered from real conversations
For yesterday, or the last 7, 30 or 90 days, each compared with the period before.
How it performed
Every conversation in the period, split by who handled it, against the period before.
ConversationsHandled by AIEscalatedPerson-led
How fast
How long a customer waited for the first answer, from the assistant and from your team.
First reply, AIFirst reply, a person
What they ask
Questions grouped into topics as they come in. A low answer rate is where the assistant needs more material.
Questions by dayQuestions by topicAnswer rate per topic
What to improve next
The specific things to fix, in the order they cost you most.
Reasons it could not answerHours it does worst
What it achieved
Leads, bookings and other outcomes the assistant logged, counted per event.
Outcomes recorded
What customers thought
The thumbs customers left on a reply, counted up. A quick signal on every answer.
Customer ratings
From gap to answer
Every question it missed becomes one it can answer
Reports show you which questions still need an answer. Your team’s replies fill them in, and nothing goes live without you.
It could not answer
A customer asks something your content does not cover. The assistant says so and offers a person instead of guessing.
Your team answers it
A teammate replies in the inbox. The customer is helped, and the reply is captured alongside the question.
You approve it
The pair waits for your approval. None of it is live until you say so.
Next time, it answers
Once approved, the assistant uses that wording for the same question. The gap closes and the topic’s answer rate shows it.
Drawn from the knowledge screenThe headings are the portal’s own. Anything the assistant added by itself is listed separately for you to check.
Clear numbers
Every number says what it actually counts
A figure you can explain to your team is a figure you can act on.
Every figure says what it counts
The definition beside each headline number comes from the system that calculates it, not from a marketing page. Per message or per conversation, and which channels it covers.
Missing means missing
When a figure was not returned for a period, the report says so, instead of showing a zero that looks like a result.
Take it with you
Download the numbers, or stream every event
A CSV for the monthly report, and live events for the tools you already use.
A CSV with stable headings
Column names stay the same between exports, so the spreadsheet formulas you build on them keep working.
Events as they happen
Bookings, conversations and contacts leave through Zapier or a signed webhook to your own endpoint, as current as the last thing that happened.
Every run, step by step
Each process keeps a record of the steps it ran, the answers it collected and how it ended, so one conversation can be explained in full.
Plans
Reports are on every plan. AI Insights starts higher.
- Reports, topics and outcomesOn every plan
- Approving answers your team wroteOn every plan
- CSV export and the event streamOn every plan
- AI Insights: why answers went wrong, and the topics most often handed to a personFrom Professional
FAQ
Questions about analytics, answered
Including how the AI learns from your team.
What can I see in EzyConn analytics?
For any period from yesterday to the last 90 days: conversations, how many the AI handled, how many were escalated or led by a person, first reply times, questions by day and by topic, the answer rate per topic, the reasons it could not answer, outcomes recorded and customer ratings, each compared with the period before.
How do I find the questions the AI could not answer?
The reports group questions into topics, worst answer rate first, and list the reasons it could not answer. When a teammate answers one of those questions in the inbox, the reply is captured as a draft answer for you to approve.
Does the AI learn from my team’s replies?
Only with your approval. A reply your team sent to a question the assistant could not answer waits as a draft. Once you approve it, the assistant uses that wording next time. Anything it added on its own is listed separately for you to check.
Can I export the numbers?
Yes. The reports download as CSV, and 17 events (bookings, conversations, contacts) leave as they happen through Zapier or a signed webhook to your own endpoint.
Is there a Google Analytics integration?
Yes, for chat events. Add your Google Analytics tag and API secret in the developer settings, and chat events are sent to your own Analytics property. For the reports themselves, use the CSV export or the event stream through Zapier or a webhook.
Which plan do I need?
Reports and analytics are on every plan, the free one included. AI Insights, which shows why answers were marked wrong and the topics most often handed to a person, is from Professional.
Start free
Find the questions your AI is missing
Add your content, let customers ask, and the reports show you exactly where to add more.
Free plan, no card.