Skip to main content

Documentation Index

Fetch the complete documentation index at: https://setup.cevro.ai/llms.txt

Use this file to discover all available pages before exploring further.

Where Drivers tells you what players contacted you about, Signals tells you what is happening inside the conversation. You define the patterns (“frustration”, “competitor mentions”, “responsible-gaming risk”) and Cevro detects every occurrence — across every channel, every brand, every agent. Signals is built for compliance teams, CS leadership, and anyone who needs to monitor in-conversation behavior at scale. URL: https://app.cevro.ai/insights/signals

​
What Is a Signal?

A Signal is a named pattern — a thing you want to detect across conversations. Each signal has:
  • A name (“Frustration”, “Pix Down”, “Competitor: Stake”)
  • An optional description
  • A list of reference phrases — examples of what the signal looks like in real conversation
  • An enabled toggle
When a message comes in, Cevro compares it against your signal phrases. Anything close enough in meaning is recorded as a detection and the ticket is tagged with the matched signal — so you can find it later in any chart or filter.
Signals run across both player and agent messages, so you can track agent compliance the same way you track player sentiment.

​
Creating a Signal

Open Manage Signals in the top-right of the Signals page to open the configuration panel.
1

Create the signal

Enter a name and an optional description, then click Create Signal. Pick a clear, descriptive name. Examples that work well:
  • Frustration — “I’m so frustrated”, “this is ridiculous”, “wasting my time”
  • Competitor mention — “Stake has”, “I’ll just use bet365”
  • RG risk — “I can’t stop”, “lost everything”, “borrowing money”
  • Refund request — “I want my money back”, “give me a refund”
2

Add reference phrases

Expand the signal and add example phrases — one at a time. 5–15 well-chosen phrases is usually enough. You can mix phrases in any language; matching is meaning-based, not keyword-based.
You only paste the phrase text — Cevro handles the rest.
3

Enable the signal

Toggle the switch in the row header. Detections start showing up on new messages within about a minute, and existing messages are backfilled in the background.
If you need to tune detection sensitivity for a specific signal, ask your CSM.

​
Boards & Charts

Detections become useful when you visualize them. Signal Boards are saved dashboards — each one contains an ordered list of charts you configure once and revisit whenever you need.

​
Adding a chart

Click + Add chart to open the chart editor. Each chart is configured by:
FieldOptions
Chart NameFree text
Chart TypeBar, 100%-stacked bar, Line, Area, 100%-stacked area, Pie, Matrix
MeasureTickets (raw count) or Tickets % (share of total)
Grouped bySignals, Day, Week, Month, Contact driver, Status, Agent, Browser, Platform, Language, Channel, Polarity, State, Contact type, CSAT Rating
Stacked by (optional)Same options as Grouped by, plus None
Look for…Ticket-level filters (same set as Drivers’ advanced filters) + Player metadata
Top NTruncate to Top 10 / 20 / 50 / 100 or show All — the long tail rolls into a single “Other” bucket
Brand and Date Range are board-level filters at the top of the page — they apply to every chart on the board. Line and area charts require a time dimension (Day / Week / Month) on Grouped by. 100%-stacked charts require a non-empty Stacked by. Common board recipes:
  • “What’s trending?” — Group by Day, stack by Signal — shows volume over time
  • “Which brand has the most frustration?” — Filter by signal=Frustration, switch the page-level brand filter
  • “Where do agents most often trigger a signal?” — Filter by signal=…, group by Agent
Click any bar, slice, or table row to drill into the underlying conversations — same Conversations Modal pattern as the rest of Insights.
When you make changes (add/remove charts, edit a config), a Save view button appears in the action bar. The board has unsaved changes until you click it.

​
Signals vs Content Shield

Signals are an analytics overlay, not a moderation gate — detections run after the fact so you can analyze and report on them. To block or filter compliance-critical content inline before a message goes out, use Content Shield.

​
Signal vs Driver: When to Use Which

QuestionUse
”What are players contacting us about?”Drivers (per-ticket classification)
“How often does the word frustrated appear?”Signals (per-message detection)
“Which Theme has the most volume this week?”Drivers
”How many tickets had RG language this week?”Signals
”Why is CSAT dropping in withdrawals?”Both — start with Drivers, drill into Signals
A useful mental model: Drivers = what the conversation was about, Signals = what was said in the conversation.

​
Common Use Cases

​
Compliance & Responsible Gaming

Define an RG risk signal with phrases drawn from your jurisdiction’s regulator guidance (UKGC, MGA, eCOGRA…). Watch the time series — any spike is a flag for a compliance review.

​
Competitor monitoring

A Competitor mention signal helps you spot churn risk early. Pair with the brand filter to see which brands lose players to which competitors.

​
Agent quality

Signals catch off-script behavior, missing greetings, or compliance gaps in agent responses. Group by Agent to see which agents over-index on a signal, then combine with QA Scoring for a full picture.

​
Product issue detection

A Pix not working or App crash signal can spot operational fires before they show up in Drivers — because you don’t have to wait for a ticket to close to detect a phrase.

​
Tips

  • Start broad, narrow over time. Add a wide phrase list at first, watch for false positives, prune.
  • Phrases in multiple languages work. Matching is meaning-based; one signal can carry phrases in EN, ES, PT, etc.
  • Don’t over-create signals. 5–15 well-tuned signals is more useful than a long list of noisy ones.
  • Drivers — ticket-level VoC classification (the “what was the ticket about” view)
  • Content Shield — real-time inline moderation
  • QA Scoring — scorecards for systematic quality evaluation
  • Sentinel — agent performance against scorecards