A simple weekly support report template: which metrics to track, how to read them, and how to turn chat data into decisions for your next week.
Most weekly support reports fail in one of two ways: they are a wall of numbers nobody reads, or a vague paragraph that says "it was a busy week." Both leave the team without a clear next step. This article gives you a one-page template you can copy, explains what each section is for, and shows how to fill it in from website chat, WhatsApp and Messenger conversations in about thirty minutes.
What a weekly support report is actually for
A weekly report has one job: help someone decide what to change on Monday. If a number does not lead to a decision, it does not belong on the page.
Think of three readers:
- The support lead, who needs to know whether staffing and coverage held up.
- The product or operations owner, who wants to hear what customers keep asking about.
- The person who maintains your AI agent, who needs a list of answers to fix or add.
Write for all three on one page. If the report runs longer than a single screen of scrolling, cut something.
The one-page template
Copy this structure into a document, spreadsheet or wiki page.
Section 1: Headline (three sentences)
- What happened this week, in plain words.
- The single biggest change versus last week.
- The one decision you need from the reader, if any.
Section 2: Core numbers
| Metric | This week | Last week | Note |
|---|---|---|---|
| Conversations started | By channel | ||
| Resolved without a human | Share of total | ||
| Handed to a human | Reasons | ||
| Unanswered or flagged | Topics | ||
| Customer rating (thumbs up/down) | Count, not just share | ||
| Sentiment | Direction of change | ||
| Leads or orders captured | If relevant |
Section 3: Top topics
List the five most common question themes, with a one-line comment on each: is the answer good, outdated, or missing?
Section 4: Failures and fixes
Three to five conversations that went badly, each with the cause and the fix you applied or plan to apply.
Section 5: Handoff and coverage
How many handoffs happened, when they clustered, and whether anyone was on shift.
Section 6: Actions for next week
A short list with an owner and a due date for each item.
Choosing the metrics that matter
Resist the urge to track everything. Start with these, and add more only when a question demands it.
Volume, by channel
Count conversations started on each channel: website chat, WhatsApp, Messenger and any others. Channel mix matters because behaviour differs. WhatsApp customers often expect a reply like a message from a friend, while website visitors are often mid-purchase and impatient.
Resolution without a human
This is the share of conversations the AI agent handled to the end. Be careful with the definition. A conversation that ended because the customer gave up is not resolved. Pair this number with ratings and with a manual read of a few samples.
Handoffs and their reasons
Do not just count handoffs. Tag each one with a reason: the customer asked for a person, the agent lacked the answer, the issue was sensitive, or the request needed an account action. The reason column is where improvement ideas come from.
Customer feedback
Thumbs up and down are crude but fast. Report the raw counts alongside any percentage, because "80% positive" means little if only five people voted.
Sentiment
Treat sentiment as a direction, not a verdict. A drop in a week with a known outage is expected. A drop in a quiet week deserves a look at the underlying chats.
Unanswered questions
These are your most valuable data. Every question the agent could not answer is a gap in your knowledge base or your policies.
How to read the numbers without fooling yourself
Numbers move for boring reasons. Before you write a story about them, check the basics.
- Holidays and campaigns. A promotion changes both volume and question types.
- Small samples. If you receive 40 chats a week, a swing of five chats is noise.
- Definition drift. If someone changes how "resolved" is counted, the trend line breaks. Note any change in the report.
- One loud customer. A single long conversation can distort averages. Look at medians or counts where you can.
Here is a hypothetical to show the habit. Suppose a shop receives 300 chats in a week and 60 are handed to a human. If last week it was 300 chats and 45 handoffs, the headline is not "handoffs rose by a third." The headline is the reason column: perhaps 20 of the 60 were about a delayed shipment, which is a communication problem, not an agent problem.
A chart like this tells you where to act. Here the first fix is a proactive delivery notice, and the second is a faster path to a human.
A repeatable weekly routine
The report is only as good as the routine behind it. Here is a thirty-minute version.
- Pull the numbers (5 minutes). Open your analytics and conversation logs for the last seven days and copy the core figures into the table.
- Scan the flagged and low-rated chats (10 minutes). Read, do not skim. Pick the three that teach you the most.
- Cluster the topics (5 minutes). Group the week's questions into five themes.
- Decide on fixes (5 minutes). For each failure, choose one: add or correct a knowledge source, adjust the agent's instructions, change a handoff rule, or fix something outside support.
- Write the headline last (5 minutes). Once you know what happened, three sentences are easy.
- 01Collect the numbers
- 02Read the worst chats
- 03Fix the answers
- 04Report and review next week
The loop matters more than any single report. The fixes you make this week should show up as better numbers next week, and the report is where you check.
Common mistakes to avoid
- Reporting averages only. Averages hide the bad days. Add a note on the worst day.
- No owners on actions. A list of "we should" items without names does not get done.
- Measuring speed, ignoring quality. A fast wrong answer is worse than a slightly slower right one.
- Skipping the follow-up. Open each report by checking last week's actions. Closing the loop is what builds trust in the process.
- Hiding bad news. A report that always looks good will stop being read.
Doing this in AI SNS CHAT
Most of the data for this template already sits inside the dashboard. Here is where to find each piece.
Volume and topics. Under Analytics you will find Chats, Topics, Question history and Sentiment. Advanced analytics require the Pro plan, so on lower plans, lean on the Chat logs for deeper reading. If you run phone agents, Voice has its own Analytics page.
Conversations across channels. Open Activity → Chat logs. Every channel except email appears there, including website chat widget, WhatsApp and Messenger conversations. Email conversations live in the Inbox. In Chat logs you can filter, flag, escalate, resolve, bulk-select and export, and you can see thumbs up and down ratings. Export gives you a clean way to paste figures into your report spreadsheet.
Failures and fixes. Open a poorly rated conversation, take it over if it is still live (the bot pauses), and edit a reply into the ideal answer to train the agent. That turns a failure in Section 4 into a fix in the same sitting. When the cause is missing content, add it under Data sources, which accepts Files, Text snippets, Website, Q&A, Notion and more. The Suggestions and Tickets sources can help surface gaps.
Handoff and coverage. Helpdesk → Scheduling lets you set shifts per person and timezone, and handoffs are assigned to whoever is on shift. Handoff alerts can go to Slack, Teams, LINE WORKS, WhatsApp, email or a webhook. When a handoff happens, the customer message includes a "what happens next" line based on your rota, and the system never invents a wait time. Helpdesk needs Standard or higher.
Leads. If your agent collects leads, they appear in Activity → Leads, and each lead also creates a contact you can export as CSV.
Credits and usage. The Usage page shows credits per agent, which is useful as a line at the bottom of the report. One reply uses one credit.
Who changed what. If an answer shifted unexpectedly, check Workspace settings → Activity (owners and admins), which records every create, edit, delete or upload with name, email, feature and time.
Getting help building the report. Synthia, the in-dashboard assistant, can walk you through screens with guided tours, so a new team member can learn where each number lives. You can compare plans and limits on the pricing page and see the full feature list on the features page.
One honest limit: there is no built-in scheduled report email in the facts above, so plan on a manual export or copy-paste each week. A shared document with the template above is enough.
Conclusion
A good weekly report is short, consistent and tied to decisions. Use the template, keep the definitions steady, read the worst conversations yourself, and assign an owner to every action. After a month, you will have a trend you can trust and a support agent that gets measurably better each week.
Frequently asked questions
How long should a weekly support report be?
One page. A headline, a core numbers table, top topics, a few failures with fixes, handoff notes and a short action list. If readers have to scroll through charts to find the point, trim it.
Which single metric matters most?
None on its own, but the unanswered and handed-off questions with their reasons are the most actionable. They tell you exactly which content or process to fix next. Pair them with ratings so you know whether customers were satisfied.
Can I report on WhatsApp and Messenger separately from website chat?
Yes. In AI SNS CHAT the Chat logs include conversations from every channel except email, and you can filter them. Report volume by channel in your core table, since customer behaviour often differs between them.
What if I only receive a small number of chats each week?
Read every conversation instead of relying on percentages. With small volumes, a handful of chats can swing any ratio, so counts and direct examples are more reliable than rates.


