Analytics

See what's happening in your inbox right now, what's been happening over the last 7, 30, or 90 days, and write your own SQL widgets when the built-in metrics aren't enough.

TL;DR

• What it is — a live operations panel (active chats, unanswered, waiting times, agents available, calls in progress) plus historical charts for response times, resolution, tickets, calls, and bots — and a custom-query builder where you can save and run your own SQL widgets, with an AI assistant that writes the SQL for you from plain English.
• Who it's for — team leads watching the day in real time; managers reviewing weekly and monthly trends; admins building org-specific KPIs the standard dashboard doesn't ship.
• Top outcome — answer "how are we doing right now?" in two seconds, "how were we doing last month?" in five, and ship a custom widget without writing SQL by hand.

At a glance

How to find it

Sidebar → Analytics.

Direct URL: https://app.lodgestory.com/crm/insights

📸 [SCREENSHOT: analytics-nav.png — sidebar with Analytics highlighted]

What is Analytics?

The problem it solves

Most teams need three different answers from their analytics: what's happening this minute?, how have we been doing this week?, and the standard dashboard doesn't break it down the way I think about my business — can I get a custom view?. The first is a real-time operations question. The second is a trend question. The third is "we run a different kind of business and the canned KPIs don't fit." A traditional analytics page answers one well and the others poorly.

The Analytics module answers all three from one page. The top of the page is a live operations panel that refreshes every 30 seconds. Below it, historical metrics for chats, tickets, calls, and bot performance over a date range you choose. And in a separate tab, a custom-query builder where you write SQL once, save it, and run it any time — with the data automatically scoped to your organisation.

What you get

• Live operational tiles that refresh on a 30-second loop — active chats, unanswered chats, the waiting-time distribution against SLAs you configure, oldest waiting chat, agents available, calls in progress, bot sessions in progress. The active and unanswered tiles exclude resolved & closed chats, so a closed thread never inflates your waiting counts.
• Configurable waiting SLAs — pick a tight (5m / 15m / 30m), default (5m / 1h / 4h), or relaxed (1h / 4h / 24h) preset, or set custom thresholds. Tiles re-bucket immediately.
• Historical charts — message volume, first response time (median), average response time, average resolution time. Each on a date range you choose.
• Date range picker — 24-hour, 7-day, 30-day, 90-day presets, plus a custom-range option for any window you want.
• Agent leaderboard — every active agent with their name, avatar, messages sent, chats handled, calls handled, talk time, FRT median, and ART average. Click an agent to drill into their profile.
• Per-agent profile — six KPI tiles for one agent plus their first-response-time trend.
• Tickets, Calls, and Bots tabs — purpose-built charts for each surface (created vs resolved, call volume + handle time, bot containment rate).
• Availability tab — total available vs. unavailable time and a per-break breakdown over your date range, plus a per-agent comparison and detail table. Admins see the whole team; an agent sees their own.
• Custom queries — write SQL with PostgreSQL syntax highlighting, autocomplete for the org-readable table list, and pick the visualisation type (KPI, bar, line, pie, or table). Save it, activate it, and run it any time.
• Ask AI — describe a metric in plain English and the assistant returns a complete query (SQL + name + description + visualisation type) ready to drop into the editor. Multi-turn — refine ("now group by channel", "and only the last 7 days") in conversation.
• Download CSV — every result preview and the agent leaderboard exposes a one-click CSV export (UTF-8 with a BOM so Excel opens cleanly).

How it's different

• Live + historical on the same page. Most analytics pages are one or the other. This one shows the day-to-day right now and the trend underneath in one scroll.
• Configurable SLAs, not fixed buckets. The waiting-chat tiles bucket against thresholds you set. Hospitality teams running 24/7 want different SLAs than agencies with strict business hours; the dashboard adapts.
• Custom queries built in. Write your own dashboard widgets in SQL, save them, share visualisations. No second tool needed for org-specific KPIs.
• AI writes the SQL for you. The Ask AI assistant has been pre-trained on the full analytics data model — every readable table, the JSON shapes of fields like the latest message blob and bot session variables, and a library of canonical query patterns the built-in metrics already use. You describe what you want; it generates SQL that matches the same patterns.
• Tenant safety on by default. Every saved query is scoped to your organisation automatically — there's no path that lets one organisation read another's data, even if a query author writes a sloppy WHERE clause.

Customer scenarios

• Daily standup. Team lead opens Analytics first thing. Live tiles show unanswered count and oldest waiting time; historical chart confirms first-response time is trending in the right direction. They click into the agent leaderboard, sort by FRT, and spot the agent who's slipping.
• Quarterly review. Operations manager picks a 90-day range, screenshots the four overview charts, and pastes them into the deck. They flip to the Tickets tab for ticket volume + resolution time, screenshot those too.
• Custom KPI for marketing. Marketing wants to track "chats per UTM source per day". They write a SQL query against the conversations and contacts data, save it as "Chats by source", set it to a bar chart, and pin it. They share the URL with their team and check it weekly.
• SLA breach hunt. A hospitality customer has chats waiting > 4 hours during weekends. They open the SLA settings, switch to the relaxed preset, and the tiles instantly show the new buckets — they can quantify the problem before staffing weekend shifts.

How it fits with the rest of Lodgestory

This module is a consumer, not a producer. Every metric is derived from data owned by other modules: message volumes from Home, chat states from Workflow Lifecycle Stages, assignments and identities from Team Members, tickets from Tickets, call activity from Calls, bot sessions from Bot Journeys. Downstream: Reports exports raw CSVs when you want to do work in a spreadsheet or BI tool.

📸 [SCREENSHOT: analytics-landing.png — Analytics overview with live tiles and historical charts]

Core concepts

Quick Start — read the dashboard in two minutes

Step 1 — Open Analytics

Sidebar → Analytics.

Step 2 — Glance at the live tiles

Top of the page. They refresh every 30 seconds; a pulsing green dot and "last update HH:MM:SS" confirm they're alive.

Step 3 — Tune the waiting SLAs once

The gear icon next to "Right now" opens a small settings panel. Pick a preset that fits your business, or type your own thresholds in minutes. Tiles re-bucket immediately.

Step 4 — Pick a date range

Top right — 24h, 7d, 30d, 90d, or Custom. Custom opens a popover with two date pickers.

Step 5 — Read the four overview charts

Message volume, FRT (median), ART (average), and average resolution time, all over the selected window.

Step 6 — Drill into an agent

Open the Agents tab. Click any row to see that agent's full profile — six KPI tiles plus their FRT trend.

Step 7 — (Optional) Write a custom query

Open the Custom queries tab → + New query. Write SQL, choose a visualisation, toggle Active, save, and click Run.

How it works

The right-now tiles read directly from your live data — there's no separate snapshot. Every 30 seconds the page asks for fresh values; the database returns them in well under 100 milliseconds.

The historical charts run aggregations across the date range you've chosen. Standard time-bucketed views (per hour, per day) come back in milliseconds for typical organisations.

Custom queries run under a locked-down read-only role with a five-second hard time-out and a 10,000-row cap. Tenant filtering is enforced automatically — your queries only see your organisation's data.

Features in depth

Right-now tiles

• Active chats — open conversations with state machine still active.
• Unanswered — chats whose latest message is from a guest.
• Waiting tiles — one tile per configured SLA threshold, showing how many unanswered chats are older than each. Severity colour shifts from neutral to warning to danger as the threshold is crossed.
• Oldest waiting — the age of the longest-waiting unanswered chat, formatted as 5m, 2h, 3d.
• Agents available — team members with availability turned on.
• Active calls — calls connected right now.
• Active bot sessions — bot journeys mid-flight.

SLA configuration

The gear icon opens a settings popover with three preset templates and three minute-input fields:

• Tight — 5m / 15m / 30m
• Default — 5m / 1h / 4h
• Relaxed — 1h / 4h / 24h

Pick a preset, or type your own. You can configure between 1 and 5 thresholds. They're sorted ascending automatically. Save updates the right-now tiles immediately.

Date range

Four presets — 24h, 7d, 30d, 90d — plus a Custom button that opens a popover with two date pickers. The active selection is highlighted; when a custom range is picked, the button shows the actual dates.

Overview charts

• Message volume — bar chart, inbound vs outbound counts per hour or day.
• First response time (median) — area chart of median seconds between a guest message and the first agent reply, per hour.
• Average response time — area chart of mean seconds across every guest→agent gap, per hour.
• Average resolution time — line chart of mean seconds from chat open to closure, per day.

All four use a single primary colour for clarity. Hovering shows the bucket value.

Agent leaderboard

A sortable table covering every agent active in the window:

Click any row to open the per-agent profile.

Per-agent profile

Header with avatar, name, and email. Six KPI tiles: messages, chats handled, calls handled, talk time, FRT p50, ART average. Below them, a chart of that agent's FRT trend over the same date range.

A back arrow returns to the leaderboard.

Tickets, Calls, Bots tabs

• Tickets — created per day (bar) and average resolution time per day (line).
• Calls — total per hour (bar, grouped by direction) and average handle time per hour (line).
• Bots — bot containment rate per day (line). Containment = completed sessions that did not hand off to a human, divided by total completed sessions.

Each tab has its own date range picker.

Availability

The Availability tab turns each agent's availability history into a duration report over your chosen date range:

• Three headline tiles — total time the team was available, total time unavailable, and the availability rate (available ÷ total).
• Break breakdown — a horizontal bar chart of total time per break label (Lunch, Meeting, …), grouped by the label recorded at the time.
• Available vs. unavailable by agent — a stacked bar per agent, sorted by unavailable time. (Admins only.)
• Per-agent detail table — each agent's available and unavailable totals, with a chip per break and its duration. (Admins only.)

Admins see the whole organisation; an agent without admin rights sees only their own numbers. Durations are computed from the gaps between availability changes, so the report is only as complete as the history — an agent who has never changed their status has nothing to show yet. Break labels come from Break Types, and agents set their status from the Home / Unified Inbox dropdown.

Custom queries

The Custom queries tab lists every query for your organisation plus library templates published by Lodgestory. Library templates have a faded border so they're easy to spot.

+ New query opens an editor with:

• A name and description
• A SQL editor with PostgreSQL syntax highlighting, line numbers, bracket matching, code folding, and ctrl-space autocomplete. Autocomplete knows the table and column names of every readable table in your organisation's data.
• A visualisation picker — KPI, bar, line, pie, or table.
• An Active toggle. New queries default to Draft; activate once you've reviewed the SQL.
• A Run button that executes the query and renders the result inline. The chosen visualisation drives the rendering — KPI shows the first numeric value as a big number, bar/line/pie chart the data appropriately, table shows everything in a sortable grid.

Tenant filtering is automatic — your queries only ever see your organisation's data. The convention is to reference $1 in your WHERE clause; Lodgestory binds your organisation ID to it before the query runs.

Every result preview also exposes a CSV button that downloads the rows as a UTF-8 CSV (Excel-compatible).

📸 [SCREENSHOT: analytics-custom-editor.png — custom query editor with SQL highlighted and a pie chart preview]

Ask AI — generate queries from plain English

Click the ✨ Ask AI button on the editor. A side panel opens with a chat input. Describe the metric:

"Top 10 contacts by inbound message volume in the last 7 days"
"Median first response time per hour, last 24 hours"
"Tickets opened vs resolved by priority over the last 30 days"
"Bot containment rate per day for the last 2 weeks"
"Average call handle time per hour, inbound only"

The assistant returns a complete widget — SQL, suggested name and description, and a visualisation type — rendered in the panel for review. Click Insert into editor to populate every field at once.

The conversation is multi-turn, so you can refine without retyping:

"Now restrict to WhatsApp only" — adds the channel filter.
"Group by hour of day instead" — switches the bucketing and the chart type.
"Exclude archived chats" — adds the soft-delete filter.
"Only open chats" / "exclude closed chats" — the assistant knows the resolved-&-closed flag and adds the resolvedAt IS NULL filter; ask for "resolved chats" or "chats closed by agent" to go the other way.

The assistant has been instructed to:

• Generate read-only SQL only — no writes, no schema changes.
• Always scope queries to your organisation.
• Refuse out-of-scope requests, attempts to read tables outside the analytics allowlist, and prompt-injection attempts.
• Surface caveats in Warnings when a query has an obvious limitation (for instance, agent display names are resolved by Lodgestory after the query runs, not by the SQL itself).

If a generated query has a lint warning, the panel surfaces it inline so you can fix or accept before inserting. Generated queries land in Draft state — review them and toggle Active before running.

📸 [SCREENSHOT: analytics-ask-ai.png — Ask AI panel with a chat transcript and the generated SQL preview]

Roles and permissions

Every team member sees every other member's numbers in the agent leaderboard. Custom-query authoring is admin-only by default; if you want a user-facing builder for non-admins, request it.

Cross-module workflows

"Are we keeping up right now?" daily check

Team lead opens Analytics. Live tiles show Unanswered: 3 and Oldest waiting: 12m. They sort the agent leaderboard by FRT to confirm coverage looks right, then move on with their day. Total time: 30 seconds.

Modules involved: Analytics → Home (to actually act on the unanswered chats).

"Why is response time creeping up?" investigation

Manager picks a 30-day range, sees FRT trending upward in the last week. Switches to 7d, sees it's spiking on weekends. Opens the agent leaderboard for last weekend, identifies whose numbers slipped. Checks Team Members for their schedule.

Modules involved: Analytics → Team Members → optionally Reports for deeper export.

"Custom KPI for our hospitality model"

Operations team needs to track "chats per property" — a custom field on contacts. They write a SQL query that joins conversations to contacts, groups by the property field, and renders as a bar chart. Save → Activate → Run. Pin the URL. Check weekly.

Modules involved: Analytics (custom queries) → Contacts (where custom fields are defined).

"I don't know SQL — let the AI write it"

Marketing manager wants a "messages per UTM source per day" widget. They open Custom queries → New → click ✨ Ask AI → type "messages per day grouped by the utm_source custom field on the contact". The assistant returns a query joining conversations to contacts, extracting customFields->>'utm_source', with a bar visualisation pre-selected and a name like "Daily messages per UTM source". They click Insert into editor, toggle Active, hit Run. If the chart looks right they're done; if they want a refinement ("only the last 30 days, only inbound"), they re-open the panel and add a follow-up turn — the assistant edits the existing query rather than starting fresh.

Modules involved: Analytics (custom queries + Ask AI) → Contacts.

"SLAs are breaking on weekends"

Manager opens SLA settings, switches from Default (5m / 1h / 4h) to a Tight preset to see the breakdown more granularly. The waiting tiles re-bucket; they screenshot. Then they switch back to Default for the next staffer who picks up the dashboard.

Modules involved: Analytics (live tiles + SLA configuration).

Limits a user will run into

Errors & FAQ

Frequently asked

How real-time are the right-now tiles?
They refresh every 30 seconds. The page header shows the last update time so you can tell at a glance.

Can I get a custom date range?
Yes — click Custom in the date range picker. Two date inputs let you pick any window.

Can my organisation have its own SLAs?
Yes — the gear icon next to the Right Now header opens the SLA settings. Pick a preset or type your own thresholds in minutes. The tiles re-bucket immediately.

Why don't I see my colleague's calls under their agent profile?
You should — the leaderboard counts both messages and calls. If a number looks zero, the agent didn't have activity in the chosen date window. Try widening the range.

Why is there no live ticket count tile?
The Right Now panel focuses on chat-level operations. For ticket volume use the Tickets tab; the open ticket count is on the roadmap.

Can I share a custom query with another organisation?
Not directly. Library templates published by Lodgestory are visible to every organisation. If you want a custom query promoted into the library, ask support.

Can I export results from a custom query?
Yes — every custom-query result preview and the agent leaderboard has a CSV download button. Output is UTF-8 with a BOM so Excel auto-detects encoding. For very large or scheduled exports, Reports is the dedicated tool.

How does the AI know about our data?
The Ask AI assistant is briefed on the full analytics data model — every readable table, the JSON shapes of fields like the latest message blob and bot session variables, the canonical query patterns the built-in metrics use, and a list of known limitations (e.g. agent display names need to be resolved by Lodgestory after the query runs). It does not see your actual data — only the static schema and your prompt.

What if the AI generates a wrong query?
The assistant lands queries in Draft by default — you must toggle Active explicitly before running. If the syntax is broken, the editor's pre-flight check rejects it with a clear error before any data is read. If the logic is wrong, refine in the chat panel ("now group by channel", "exclude archived") or edit the SQL directly.

My custom query worked yesterday but is failing today with a column error.
Likely the underlying field structure changed. SQL queries reference fields by name; if a name was renamed elsewhere in the product, your query needs to be updated. Open the query and re-run — the error message points at the missing field.

Related modules & next steps

• Reports — raw CSV export for anything beyond the dashboard.
• Home (Unified Inbox) — the primary data source for chat metrics.
• Tickets — ticket creation, resolution, and assignment that feed the Tickets tab.
• Calls — call activity behind the Calls tab.
• Team Members — drill-down on individual agents.
• Bot Journeys — sessions behind the Bots tab.
• Campaigns — broadcast-level progress (separate from this page).