analytics
Customer support analytics
that show what AI actually resolves.
Most support analytics were designed before AI answered anything. HelpYap measures the questions that matter now: what the AI resolves on its own, why conversations escalate, and what your AI actually costs, down to tokens per model per project.
Analytics retention: 7 days on Starter, 90 days on Pro, 365 days on Business.
This week
Mon–Sunwhy, not just how many
Escalation reasons point at the fix
Knowing your escalation rate is trivia. Knowing that half of escalations happen because your docs have no coverage on one topic is a to-do list. HelpYap breaks every escalation down by reason.
- No KB coverage, user requested, complaint, AI judgment
- Top intents show what customers actually ask about
- Source coverage tracked on every AI reply
Why conversations escalate
Every bar is a doc you could write
ai costs, itemized
See what the AI costs before the invoice does
Every AI call is metered with its estimated cost, broken down by feature, model, and project. If your advanced model is answering questions the fast one could handle, you will see it here first.
- Cost per model, per feature, per project
- Recent calls listed with their usage
- CSV export for everything
Usage and estimated cost, split by feature, model, and project
What metrics does HelpYap track?
The dashboard covers the full support funnel: how much comes in, how much gets resolved and by whom, how fast, and how customers felt about it. Every number is computed directly from your conversations.
How do I measure what the AI resolves on its own?
The AI-only resolution rate counts conversations resolved without any human touching them, separate from the overall resolution rate. It is the honest number for how much work the AI is actually taking off your team.
Because HelpYap includes AI resolutions in the seat price, this number going up costs you nothing extra. On per-resolution pricing, it would be your bill.
Why do my conversations escalate to humans?
Analytics breaks escalations down by reason, derived from the source coverage of the AI's answers. "No KB coverage" means the agent had nothing in your docs to answer from, which points directly at articles worth writing.
Top intents show what those escalations were about across 10 categories such as billing, bug, and account, so you can see whether gaps cluster in one area of your product.
Can I see what my AI actually costs?
Yes. AI usage tracking reports usage and estimated cost broken down by feature, by model, and by project. You can see exactly how much the chat agent, the Copilot, and each project consume.
That makes model choices concrete: you can check what moving a project to a more capable model actually changed in both quality and spend.
How is customer satisfaction measured?
CSAT comes from conversation ratings customers leave in the widget after a chat. Ratings roll up in analytics so you can watch satisfaction move as you change your docs, your models, or your team process.
On Business, negative ratings also feed the Learning Loop, which turns them into lessons the AI applies to future conversations.
Can I export the data and how long is it kept?
Every report exports to CSV, so the numbers can go into your own spreadsheets and BI tools. Retention depends on plan: 7 days of history on Starter, 90 days on Pro, and 365 days on Business.
- CSV export on all reports
- Starter: 7 days of retention
- Pro: 90 days of retention
- Business: 365 days of retention
faq
Common questions
What is AI-only resolution rate?
The share of conversations resolved entirely by the AI with no human involvement, tracked separately from the overall resolution rate. It measures how much work the AI takes off your team.
How long does HelpYap retain analytics data?
7 days on Starter, 90 days on Pro, and 365 days on Business.
Can I export HelpYap analytics?
Yes. Reports export to CSV so you can analyze the data in spreadsheets or your own BI tools.
Does HelpYap track AI token usage and cost?
Yes. AI usage tracking reports tokens and estimated cost broken down by feature, by model, and by project.
What is source coverage in analytics?
Every AI answer is labeled grounded, partial, or none based on how well your knowledge base backed it. Analytics aggregates these labels and uses them to explain why conversations escalated, including a "No KB coverage" reason.
Try it on your own docs
Every plan starts with a 7-day free trial. No credit card required.