#Reports

#What it is

Reports are detailed pages about how your workspace performed over a period you choose. They live under Reports in the navigation (Admins only):

Report	Answers
Inbox performance	How fast does my team reply, who handles the most, how much does the AI resolve?
Campaigns	How did my broadcasts do?
Automations	Do my automations work, and what goes wrong?
Usage Report	What does the AI cost, and where is it used?

Social posting has its own report under Social, Report. See Social Media.

#The problem they solve

The Dashboard tells you how you are doing. When something looks wrong (slow replies, falling delivery, rising AI cost), you need to dig in: which day, which person, which campaign, which automation. Reports let you do that, and download the data to share.

#How they work (all four)

• Choose 7 days, 30 days, 90 days, or your own From and To dates (up to a year), then Apply.
• Each number is shown next to the period just before, so you see whether it is better or worse.
• Dates follow your workspace time zone.
• Export CSV downloads the data. Files are safe to open in a spreadsheet: anything that could be mistaken for a formula is saved as plain text. Every export is recorded in the Activity log.

#1. Inbox performance

Reports, Inbox performance.

Cards

• First reply (median): how long until anyone, the AI included, answers.
• First reply from a person (median): only conversations a person answered.
• Time to resolve (median): from opening to resolved.
• Conversations opened and resolved.
• Reply target met: the share of first human replies sent within your target. The target is the number of minutes you set under Settings, Workspace, Alert when a customer has waited. With no target set, the card says so.

Below the cards

• A daily chart of the reply target.
• A team table: for each person, conversations assigned, resolved, replies and median first reply.
• Conversations by channel and by label.

Export CSV gives one row per conversation opened in the range, with reply and resolve times in seconds.

Why use it: to spot slow hours, to see who needs help, and to prove your response times.

#2. Campaigns

Reports, Campaigns. Every campaign created in the period, with recipients, sent, delivered (and %), read and failed. Each row links to that campaign's own report with the delivery funnel and the reasons for failures. Totals sit at the top. Export CSV gives the list.

Why use it: to compare campaigns and find which channel, time or message works best.

#3. Automations

Reports, Automations.

• Runs, and how many are running.
• Success rate: of runs that ended on their own, how many completed.
• Failed and cancelled.
• Waiting now: runs paused on a wait or a question, right now.
• Runs per automation, with how many failed.
• The most common errors, so you can fix a cause once.

Why use it: an automation that fails quietly is worse than none. This page shows which ones need fixing.

#4. Usage Report (AI cost)

Every time the AI does something, Threadline records it. There are two kinds of AI work:

• Chat. Writing an answer to a customer, a draft for an agent, an email or a social post plan.
• Embedding. Reading your documents and turning them into a searchable form. This happens when you add or change content in a knowledge base.

AI providers charge by tokens. A token is a small piece of text, roughly three quarters of a word. Prompt tokens are what the AI was given to read. Completion tokens are what it wrote back.

Cards

• Requests: how many times the AI was called.
• Prompt tokens and Completion tokens.
• Estimated cost: what your AI provider is expected to bill.
• Answer time: the average, and the time within which 95% of answers arrived.
• Error rate: how often the AI provider failed.

Below, the numbers are broken down by day, provider, model, operation and by assistant.

Filtering. Choose dates, then optionally a Provider, Model or Operation, and click Apply filters. If you see "No usage in this range", widen the dates or clear a filter.

Your plan may cap AI usage per month. When it is used up, the AI stops answering until next month or an upgrade. See Plans and Billing.

#5. Ads

If you run Click-to-WhatsApp (or Click-to-Messenger and Instagram) ads on Meta, Reports → Ads shows which ad each conversation came from and what came of it. For each ad in the period:

• Conversations that started from it, and the new contacts it brought. A person is credited to the first ad they ever came from.
• Deals those people opened, how many were won, and their value.
• Orders from a connected store, matched to those people, and their value.
• The median time to the first reply.

Meta tells us an ad's id and headline, but not the name you gave it in Ads Manager. Press the pencil next to an ad to give it your own name. Click an ad to see the people it brought in Contacts.

#Good to know

• Reports are for Admins. Agents see only their own numbers on the Dashboard.
• A report counts what happened in the period, so changing a past record later (for example reopening a conversation) can change old numbers slightly.
• An export covers the filters you set.

#Common questions

Why is a number a dash? There was nothing to measure in that period (for example no replies yet).

Why does "Resolved" differ from "Closed"? Closing a conversation also resolves it, but a conversation can be resolved without being closed.

The AI cost looks high. Open the Usage Report and look at the breakdown by assistant, model and operation: adding large documents (embedding) is a one-off cost; many long answers (chat) is ongoing.