Overview

The Analytics page shows how your AI agent is performing across all your channels. Every widget updates together based on two filters at the top of the page:

Time period — 7 days, 30 days, 90 days, or all time
Channel — all channels combined, or a single channel (WhatsApp, Telegram, Instagram, Messenger, Email, and so on)

Change either filter and every widget recalculates for that scope. You can also export the data to CSV for your own reporting.

KPI Cards

The KPI cards give you the headline numbers for the selected period, each with a week-over-week change so you can see the direction of travel:

New leads — how many new contacts started a conversation
Messages — total messages exchanged
AI replies — how many replies the AI sent
Qualified rate — the share of leads classified as HOT or WARM (or moved to a qualified/booked status)
Response latency — how quickly the AI responds
Bookings — appointments scheduled through the agent

How to act: a falling qualified rate usually points to classification or knowledge-base tuning. Rising latency is worth checking against a large or recently changed knowledge base.

Conversion Funnel

The funnel shows how leads move through the journey and where they drop off:

All leads → first AI reply → engaged → qualified (HOT/WARM) → booked → completed

Each stage shows how many leads reached it, so you can spot the biggest leak.

How to act: a big drop between engaged and qualified suggests the conversation isn't moving people toward a decision — review your AI personality, knowledge base, or follow-up. A drop between qualified and booked points at the booking step or calendar setup.

Per-Channel Breakdown

A table comparing your channels side by side — leads, messages, qualified rate, and other key metrics per channel.

How to act: use it to see which channels earn their keep and where to focus. A channel with lots of messages but a low qualified rate may need channel-specific prompt or knowledge-base overrides.

Response-Latency Percentiles

Latency is shown as percentiles (not just an average) so a few slow outliers don't hide the typical experience. Percentiles tell you what most customers actually wait.

How to act: if the higher percentiles climb, check knowledge-base size and whether tools (calendar, catalog) are adding round-trips on some replies.

Hourly Activity Heatmap

A heatmap of activity by hour, so you can see when your customers actually message you.

How to act: align staffing, quiet hours for alerts, and follow-up timing with your real peak hours.

Handoff Reasons

A breakdown of why conversations were handed to a human — for example, which keywords triggered a handoff most often.

How to act: recurring handoff reasons are a to-do list for your knowledge base. If people keep asking the same thing the AI can't answer, add it so the AI handles it next time.

Credit Consumption

Shows how your credits were spent over the period, broken down by the kinds of events that consume them (replies, follow-ups, media, and so on).

How to act: use it to forecast top-ups and to spot unexpected spikes in usage.

Knowledge Base Performance

Reports on how the knowledge base is contributing — how often retrieved content is used in replies.

How to act: low usage can mean your documents aren't matching real questions. Revisit content, coverage, and phrasing so the AI can find the right passages.

Lead Sources

Shows where your leads are coming from, so you can attribute results back to your channels and campaigns.

How to act: double down on the sources producing qualified leads, and reconsider the ones that only produce volume.

Exporting Data

Use the Export to CSV option to download the analytics data for the selected period and channel. This is handy for board reports, spreadsheets, or combining Waslo data with other sources.

FAQ

Does changing the channel filter affect every widget? Yes. The time period and channel filters apply to the whole page, so every widget reflects the same scope.

What counts as a "qualified" lead? Leads classified as HOT or WARM, or leads moved into a qualified or booked status.

Why show latency as percentiles instead of an average? An average can be dragged around by a few slow replies. Percentiles show what most customers actually experience.

My qualified rate is low — where do I start? Review your classification criteria and your knowledge base. If the AI can't answer common questions well, fewer conversations progress. The handoff-reasons and knowledge-base widgets are good places to find the gaps.

Can I get the raw numbers out? Yes — use the CSV export for the current period and channel selection.

Analytics & Insights