Statistics and reports

What it is

ZeroChats produces statistics and reports at several levels so you understand what your AI is doing and where to improve: the Statistics page (how your conversations evolve), the Performance page (which post or origin brings the leads who convert), a per-lead report and a weekly report sent to your email.

This page explains each statistic as it appears in the app: what it counts, which leads it looks at, which date it uses, whether it follows the date picker and how to read it. If two numbers don’t match, go straight to Why don’t two numbers match?.

Main dashboard

In the dashboard you see at a glance:

Conversations active right now.
Lead distribution by funnel state.
AI performance and recent conversions.
Quick access to the leads panel and open conversations.

How to read any statistic

The chips next to the title

Not every section looks at the same people, the same period or the same date. That’s why each section carries chips next to its title. Hover over (or tap) any chip to see its explanation.

Who it counts:

Chip	What it means
AI only	Only counts the people your AI actually talked to. Anyone who never exchanged a message with it is left out, so the percentages reflect real conversations.
All leads	Counts everyone you reached, whether they replied or not, and even if your AI never talked to them. It measures reach, not AI performance.

Which period:

Chip	What it means
Range	The figure is recalculated with the selected date range.
All time	The figure covers the account’s whole lifetime and doesn’t change with the date picker.

Which date places each lead on a day:

Chip	Which date it uses
By state change	The day of the lead’s last state change (when it moved to Proposal, Booked, Paid, Lost…). If it never changed state, the day it came in. If it moves on tomorrow, it moves to tomorrow: you always see the current picture.
By send date	The day the person received the welcome, the sales link or the content. That day never changes. If they received it twice, it counts once.
By arrival date	The day and hour the lead first wrote to you, which is exactly what that section measures: when people write to you.
Mixed dates	Welcomes (outbound) count on the day they were sent; leads who wrote to you count on the day of their last state change.

Replied

Across all statistics, “Replied” means the same thing: the lead sent you a direct message (DM) after a message of yours.

On welcomes (outbound, you wrote first): they answered your welcome.
On inbound (they wrote first): they wrote again after your reply. Their first message alone is not a reply.
Comments don’t count. Commenting on a post is not replying. If you answered a comment privately and the lead then wrote by DM, that does count.
It counts even if the lead was later lost. A lead who replied and ended up in Lost still counts as replied.
It doesn’t depend on the AI. It counts even if your AI never talked to that person (for example, because it was off or out of credits).

It’s the same rule in Statistics, Performance and Zero, so those numbers always match.

Conversion rate

The conversion rate is, out of every 100 conversations, how many are in Booked or Paid: Booked plus Paid, divided by the conversations.

Proposal made is everyone who got a proposal (now in Proposal, Booked or Paid). Booked and Paid each count only their own state, never twice: a lead who booked and then paid counts only in Paid.

Across the app, a conversion means being in Booked or Paid today; a Proposal is not a conversion yet. It’s the rate used by the Overview, Conversations per day, AI performance, the tag section, the content and sales link sections, and the Conversions on the Performance page. Figures that do include Proposal have their own name: Proposal made and Qualified.

Which date each section uses

By default, each lead counts on the day of its last state change, and leads that never changed state count on the day they came in. That way the same range describes the same leads in Statistics, Performance and Zero.
Everything about welcomes (daily chart, templates, the Outbound tab of the final funnel and the Welcome row in Performance) uses the day the welcome was sent.
Sales links and content use the day each one was sent.
Day of week and hour of day use the moment the lead came in.

Explain with Zero

Every statistic has an Explain with Zero button. Pressing it opens Zero with that section already loaded to explain it to you: what you’re looking at, why a figure is what it is and what you can do about it. It’s the quickest way to settle a specific question about a number without leaving the page.

The How it works button

The How it works button, on both Statistics and Performance, opens the in-app reference with three tabs:

What each counts — a table with who each section counts, its period and its date.
Lead states — the nine funnel states, with the same colors as the charts.
Why don’t they match? — the differences that confuse people most.

Statistics

Each section has its own date picker, with the ranges Last 7 days, Last 30 days and Last 3 months, plus Custom range. Named ranges are rolling: they move forward with you every day. All days are calculated in your business timezone.

The sections appear in this order.

Overview

What it counts: five figures since you started: Conversations since the start, Proposal made, Booked, Paid and All-time conversion rate.
Who: AI only.
Period: All time. It doesn’t change with the date picker.
How to read it: Booked and Paid never overlap and Proposal made includes them; the rate is, out of every 100 conversations, how many are in Booked or Paid (see Conversion rate). Don’t compare it with the All tab of the final funnel: unanswered welcomes don’t count here and there’s no range.
It can go down: it counts the state each lead is in today. If a lead leaves Proposal, Booked or Paid (because you move it by hand or it moves on its own to another state, such as Lost), it stops counting in those figures. That’s why it doesn’t have to match what you counted for a period.

Conversations per day

What it counts: for each day, four bars side by side: Conversations, Proposal made (Proposal, Booked or Paid), Booked and Paid. On top, the same range totals and the Conversion rate ((Booked + Paid) / conversations).
Who: AI only.
Date: By state change. Period: Range.
How to read it: a lead that moves to Booked today moves to today’s bar. That’s why a past day can lose conversations when its leads move on: you’re seeing the current picture. Proposal made, Booked and Paid count as in the Overview, and the conversion rate is the same.

AI performance

What it counts: one card per AI Agent with Messages, Total, Proposal made, Booked, Paid and Conversion ((Booked + Paid) / Total). Each lead counts for the agent assigned to it; leads with no agent appear under Default.
Who: AI only.
Period: All time.
How to read it: compare agents by their Conversion, not their volume: an agent that gets more traffic will always have more conversations. Since it’s all time, a recent prompt change takes a while to show here; use Conversations per day for that.

Tag conversion rate

What it counts: for each day and each tag, how many leads with that tag are in Booked or Paid. It only appears if you have tags.
Who: AI only.
Date: By state change. Period: Range.
How to read it: a lead with several tags counts in each of them, so the bars of different tags don’t add up. Use it to see which kind of lead (according to your Tags) moves forward most.

Content conversions per day

What it counts: for each day and each piece of content, how many people received it (full bar) and how many of them are today in Booked or Paid (solid part).
Who: AI only.
Date: By send date. Period: Range.
How to read it: each bar answers “of what I sent that day, where do those people stand today”. A person who received three pieces of content counts in all three, so don’t add content up. The most recent days always look worse: those people haven’t had time to decide yet.

Sales link conversions per day

What it counts: the same as the previous one, but per sales link: for each day, how many people received each link and how many are today in Booked or Paid.
Who: AI only.
Date: By send date. Period: Range.
How to read it: only a link that reached the lead counts: if the platform refused the message, it isn’t counted. Links sent by the AI, follow-ups, comment replies and the ones you type yourself, including from the Instagram app, all count. As with content, a person counts once per link they received.

Welcome messages

A welcome is the first message when you start the conversation (outbound): from the Instagram app, with another tool or with another app connected to your account.

What it counts: Sent, Responded and Response Rate for the range, and a per-day chart you can switch between Sent, Responded and Response rate (the last one as a heatmap per template).
Who: All leads. It counts everyone you wrote to, whether they replied or not.
Date: By send date (the day the welcome went out). Period: Range.
How to read it: “Responded” follows the definition of Replied. The sends of a closed week don’t change, but its replies can keep going up if someone answers late.

By template

Below, the same section groups each welcome into its template: the same text with names and personal details replaced by a placeholder. That way you compare the welcomes you’re testing instead of seeing them all mixed together.

One row per template with its text —placeholders are marked— and its funnel: how many people responded and how many reached Qualified, Booked and Paid, with each stage’s percentage over the sent ones.
If two templates are really the same message, select them and press Combine: their figures add up. The app also suggests the ones that look like the same message. You can Separate a variant afterwards.
The No template row gathers welcomes sent before templates were recorded, those that couldn’t be grouped and freshly sent ones still being classified (a few minutes). It counts in the totals all the same: the rows always add up to the figures above.

Final funnel

The complete funnel, from every lead in the range to those who booked or paid.

Tabs:
- Outbound — leads you wrote to. Inbound — leads who wrote to you. All — both together. All three tabs count every lead, whether your AI spoke with them or not.
Date (Entry and state): All and Responded count the leads that came in during the period (for welcomes, the send day). Qualified and Booked and paid count the leads whose last state change fell in the period, like the other state-based sections.
Stages: All → Responded (per Replied) → Qualified (Proposal, Booked or Paid) → Booked and paid.
Views: Funnel (the stages as a funnel, with the states in each one), Steps (what percentage moves on to the next step) and By state (bars with Not advanced, Unqualified, Proposal and Booked and paid).
Period: Range.
How to read it: the drop between All and Responded is the people who didn’t write to you after a message of yours: in Outbound, those who didn’t answer the welcome; in Inbound, those who didn’t write again after your reply. Inbound almost always looks better than Outbound because Outbound carries everyone who never answered: compare each tab with its own evolution, not one against the other. Because the first two stages are dated by entry and the last two by state change, in a short range a lead who came in earlier and moved on now counts in Qualified even if it isn’t in All.

Follow-up response rate

Which share of the leads who received a follow-up replied within the 24 hours after each step. It has three tabs, one per kind of follow-up, sharing the same date range:

24-hour follow-ups — the automatic ones, sent on their own within 24 h of the lead’s last message.
Manual · 7 days — those a person sends from a manual step (Pending, Chats or Zero), up to 7 days after the lead’s last message. A message typed by hand without starting from a manual step doesn’t count.
Welcome — those sent on their own to people who didn’t answer your first message.

In each tab:

What it counts: Total (distinct leads who received any step), Responded and Response rate, overall and per step.
Who: All leads who received the follow-up.
Date: the day each step was sent. Period: Range.
How to read it: someone who received several steps counts in each bar, but only once in the Total. Only steps that reached the lead count: a step the platform refused is left out.

Your leads’ reply time

What it counts: how long each lead takes to answer your last message, split into brackets (< 5 min, 5 – 60 min, 1 – 6 h, 6 – 24 h, > 24 h), and the figure “Half of your leads reply in less than…”. It measures the latest reply of each lead active in the last 7 days; if there are more than 5,000, a random sample of 5,000.
Who: every recently active lead who answered a message of yours. Comments don’t count.
Period: always the last 7 days. It doesn’t follow any date picker.
How to read it: it tells you when to schedule your follow-ups. If half of your leads reply in under an hour, a follow-up at 2 h arrives when most of those who were going to answer already have.

Conversations by day of week and by hour of day

What it counts: the leads who wrote to you (inbound), grouped by the day of the week (Monday to Sunday) and the hour of the day (0-23) when they wrote, with four bars: Conversations, Proposal made, Booked and Paid.
Who: AI only, and inbound only: an outbound lead comes in when you write to them, not when they write.
Date: By arrival date. Period: All time.
How to read it: use them to know when your people show up and to adjust your office hours or when you post. Being all time, they show stable patterns; don’t expect them to change from one week to the next.

Performance

The Performance page answers a different question: which post or origin brings the leads who convert. It has a single date picker for the whole page (by default, the last 30 days), and every card shows the range it describes.

On Performance, Conversions means being in Booked or Paid today, as everywhere else in the app. The sections are dated by state change, except the Welcome row, which uses the day the welcome was sent.

The four figures at the top

All four look only at leads attributed to an Instagram post, as the note next to the date range reminds you: “Only leads that came from stories, posts, reels and ads”. Inbound leads, Welcome leads and those in a personalized origin are not included: you’ll find them in Performance by Welcome and Inbound and in the Performance Breakdown.

Card	What it counts
People	Distinct people attributed to any post in the range. If the same person engaged with three posts, they count once.
Conversions	How many of those people are in Booked or Paid today.
Conversion Rate	Conversions divided by People, as a percentage.
Unique Origins	How many distinct posts brought at least one lead in the range.

Performance by Welcome and Inbound

What it counts: two rows and their total. Welcome: leads you wrote to. Inbound: leads who wrote to you directly, with no Instagram post or personalized origin behind them. Columns: Total, Answered, Proposal made, Booked and Paid, each with its percentage over the Total.
Who: All leads.
Date: Mixed dates: Welcome by the day the welcome was sent; Inbound by state change. Period: Range.
How to read it: the Welcome row counts the same leads as the Welcome messages section for the same range, and Answered follows the definition of Replied. Here Booked and Paid are separate: a lead in Paid doesn’t appear in Booked.

Top 10 Origins

What it counts: the ten posts with the best results, sortable by People, Conversions or Conv. Rate.
Who: All leads. Date: By state change. Period: Range.
How to read it: sort by Conv. Rate to see which content converts best, and by People to see which attracts most. A post with few people and a very high rate may be a fluke: look at both.

Daily Trends by Origin

What it counts: people per day, grouped by origin type (Story, Ad, Post, Reel).
Who: All leads. Date: By state change. Period: Range.
How to read it: use it to see which format is driving your acquisition at any moment, for example whether reels start bringing more people than stories.

Performance by Origin

What it counts: people, conversions and rate added up by origin type.
Who: All leads. Date: By state change. Period: Range.
How to read it: it tells you which format works best overall, beyond any single post.

Performance Breakdown

What it counts: one row per post (and per personalized origin), with Type, Published (publication date), Total, Answered, Proposal made, Booked and Paid, each with its percentage. Click a column to sort.
Who: All leads. Date: By state change. Period: Range.
How to read it: a lead who touched three posts counts in all three rows, so the Total column doesn’t add up to your leads. Each stage has a See the leads link that opens the list of that post’s leads in that stage and range (if there are many, it tells you it’s only showing the first ones).
Export CSV downloads these metrics, meant for uploading to ChatGPT or Claude and asking them for an analysis. It downloads figures, not the content of the conversations; for that use the leads panel or MCP.

Where a lead came from

Each conversation is grouped by where it came from. If the lead came from a specific post, you’ll see which one; if not, you’ll see how the conversation started:

Origin	What it means
Story	The lead came from a story.
Ad	The lead came from an ad.
Post	The lead came from a feed post (photo or carousel).
Reel	The lead came from a reel.
Inbound	The lead wrote to you first and there’s no identified post behind it.
Welcome	Your business started the conversation: the first message came from your account, not the lead’s.

With the Personalized origins button you can create your own origins to group leads by the CTA of their first message (for example, everyone who writes “GUIDE”) and update the leads you already have.

Each lead’s origin also appears in its report, and in the notification email when a lead books.

Why don’t two numbers match?

It’s almost never a bug. Before writing to support, check these six causes:

They count different people. AI only sections only count those who talked to your AI; All leads sections count everyone you reached. That’s why welcomes sent don’t match conversations, the Overview doesn’t match the All tab of the final funnel, and Performance shows more leads than Statistics.
They use different dates. What’s dated by send date (welcomes, content, sales links) places each person on the day they received the message; what’s dated by state change places them on the day of their last state change. For the same range they can be different people: a lead who got the welcome on day 1 and booked on day 10 counts on day 1 in the welcome chart and on day 10 in Conversations per day.
All time versus range. The Overview, AI performance and the day-of-week and hour-of-day charts cover the account’s whole lifetime; the rest follow the range. And Your leads’ reply time always looks at the last 7 days. Also, the Overview counts the state each lead is in today, so it goes down when a lead leaves Proposal, Booked or Paid.
A lead counts once per item. In the content, sales link and tag charts, and in the Performance breakdown, a lead who received three pieces of content or touched three posts counts in all three. Adding those bars up gives more leads than you have.
People versus conversations in Performance. The figures at the top of Performance count distinct people attributed to posts (and only to posts), while the table counts each person under every post they touched.
The weekly report is a snapshot. Its figures are calculated when it’s generated, at the send time, so its last day only covers up to that hour. Afterwards, Statistics keeps changing as leads change state, and the same 7 days can show different figures later. That’s expected, not a bug.

Zero Assistant: ask the AI about your data

The old GPT page no longer exists: now you ask Zero, from Zero Assistant in the sidebar or with the Explain with Zero button on each statistic. Some examples:

“What are the most common objections this week?”
“What kind of lead usually buys?”
“Conversion rate for the last 30 days”

Zero uses the same figures and the same rules you see in Statistics, Performance and Benchmark, can make charts for you and, if you ask, can also make changes for you. Everything it can do is in Zero Assistant.

Per-lead report

Each lead has an automatic report available from its panel. It includes:

Lead quality and interest level.
Detected urgency and intent.
Red flags (objections, lack of budget, etc.).
A summary of the conversation.
List of content and links already sent.

Ideal for preparing a closing call without having to reread the whole chat.

Weekly report by email

You can receive a complete summary of the business in your inbox every week:

Open the Zero Assistant and ask it to turn on the weekly report.
Tell it which day and at what time you want to receive it.
If you want your team to get it too, give it their emails.

The same way, you can ask it how it’s set up, to change the day, time or recipients, or to turn it off.

The report comes with an AI-generated analysis of the whole week’s activity: volume, conversions, states, most-used content and recommendations.

Its figures are frozen when it’s generated, at the send time: its last day only includes up to that hour, and if you later look at those same days in Statistics you may see different numbers because leads kept changing state. More in Why don’t two numbers match?.

Best practices

Before comparing two figures, check their chips: if they don’t share the population, the period and the date, they have no reason to match.
Judge a recent week carefully in send-dated sections: its replies and conversions can still go up.
Review the weekly report before touching the prompt: decisions are made with data, not hunches.
Use the Zero Assistant to spot patterns quickly, especially objections and funnel drop-off points.
Share the report with the sales team by adding their emails as recipients of the weekly report.
Before closing a sale, open the lead report and use it as a script: the AI has already done the qualification.