Přeskočit obsah

How Quality Works

Daktela measures quality from two sides:

  • Internal quality (QA) — your team leaders, or AI, score how well an agent handled a call, chat or email against a checklist you define: a QA form. The result is a QA review with a score.
  • Customer satisfaction — the customer rates you. Daktela has two separate mechanisms: a built-in NPS survey for emails and web chats, and CSAT forms for surveys that your own integration submits.

This page explains how the pieces fit together and how to start. The field-by-field reference lives on the pages linked below.

The Quality Loop

flowchart TD
    A[QA form] --> B[Linked to queues]
    B --> C{Who reviews?}
    C -->|Team leader| D[Scores an activity]
    C -->|AI QA| E[Reviews activities automatically]
    D --> F[QA review]
    E --> F
    F --> G[QA rating statistics]
    F --> H[Agent sees own reviews]
    G --> I[Coaching and form updates]
    H --> I
    I --> A
  1. Build a QA form in Settings → Workflows → QA Forms: the questions, the answers and how many points each answer is worth. You can start every review from an Initial score and add or subtract points, set a Minimum final score, and mark answers as Disqualifying (the review drops to 0 or the minimum score) — see QA Forms.
  2. Link the form to queues. A form can only be used on activities from the queues it is linked to. Link it from the form's Queues relation in the QA form list (or from the queue's QA forms relation).
  3. Review activities. A reviewer opens Listings → Activities, clicks the QA reviews action on an activity, picks a form and answers each question while listening to the recording or reading the transcript. Each review records an absolute score, a maximum score and a relative score.
  4. Let AI do part of the work (optional). AI QA fills in the same QA forms automatically for the activities you choose. Its reviews appear in the same list and a team leader can correct them.
  5. Read the results. Every review is listed in Listings → QA Reviews. For averages, use the Rating QA by queue and Rating QA by user statistics in Reporting → Statistics: they show how many activities there were, how many were rated and the average score, and can be filtered by QA form.
  6. Close the loop with agents. Give agents the Read own reviews permission and they can open their own reviews (read-only) in Listings → QA Reviews, so feedback is based on the same evidence the reviewer used.

The Building Blocks

Building block What it does Where to set it up
QA Forms The scoring checklist: questions, answers, points, disqualifying answers, linked queues. Settings → Workflows → QA Forms
QA Reviews The list of all completed reviews, with filters by form, reviewer and reviewed agent. Listings → QA Reviews
AI QA Automatic reviews by AI using an AI QA Form, per queue. Queue → AI functions — see Queue Basics
Smart Call Transcript Turns recorded calls into text, which AI QA needs for calls. Queue settings (Speech-to-Text)
NPS survey templates A 0–10 "would you recommend us" survey added to emails or shown at the end of a web chat. Settings → Workflows → Templates — see Templates
CSAT Forms The structure of satisfaction surveys that an integration submits through the API. Settings → Workflows → CSAT Forms
CSAT Surveys The list of submitted CSAT surveys. Listings → CSAT Surveys

AI QA

AI QA reads an activity and answers the questions of a QA form for you. It is set up per queue, in the queue's AI functions: turn on Use AI QA Reviews, choose the AI QA Form and the AI Language, and optionally limit which activities are reviewed with AI QA conditions and AI QA Frequency (from All activities down to every 30th activity).

Before you turn it on:

  • The form must be an AI QA form. Tick AI QA Form on the QA form, and give every question and answer a clear Description — that text is what the AI is instructed with. Linear Scale questions are not supported.
  • Calls must be recorded and transcribed. AI QA only reads text. For call queues, set When to record calls automatically (recording is off by default) and turn on Smart Call Transcript (Speech-to-Text), which is licensed separately.
  • It costs Power Pack tokens: 15 tokens for every question, for every activity analysed. A 10-question form on 1,000 calls a month uses 150,000 tokens. Use AI QA conditions and AI QA Frequency to review a sample rather than everything. Consumption is shown in the AI Hub.

Customer Satisfaction: NPS vs CSAT

NPS survey CSAT forms
What the customer answers One question, a score from 0 to 10; email respondents can also leave a comment Any fields you define (text or numbers)
Channels Emails (added at the bottom of every email sent from the queue) and web chats (shown at the end of the chat) Any — your integration decides when and how to ask
How it is collected Built in. Select an NPS survey template on the email queue or the web chat connector. Only through the REST API. You need an integration or a Daktela Event that submits the survey.
Where you see single answers Listings → Activities (filter by NPS score) Listings → CSAT Surveys and the CSAT Surveys tab of the activity or ticket
Aggregate reporting Rating NPS statistics by queue, user and time: detractors (0–6), passives (7–8), promoters (9–10) and the Net Promoter Score = (promoters − detractors) ÷ answered None in Daktela. Export the surveys or analyse them in your own tools.
Calls Not available. A Robocaller campaign can collect a keypad rating into a custom field instead. Possible, if your integration submits the survey after the call

Set up NPS: create an Email NPS survey or Chat NPS survey template (Templates), then select it in NPS survey template on the email queue or on the web chat connector.

Set up CSAT: build the form in CSAT Forms, then have your developer or integration partner submit surveys against it through the API.

Set Up Quality in 4 Steps

Recommended starting practice

The steps below are a recommendation for a first quality programme, not product defaults. Adjust the numbers to your team size.

  1. Turn on recording on the queues you want to review, so reviewers have something to listen to. See When to record calls automatically in Inbound Call Queue.
  2. Build one short QA form with 5–10 questions about the behaviours you care about most (greeting, identifying the customer, solving the problem, closing the call). Use Multiple choice questions with clear points, mark at most one or two Disqualifying answers (for example a compliance breach), and link the form to your main queues. See QA Forms.
  3. Agree a review rhythm: for example 2–4 reviews per agent per month, picked at random from Listings → Activities. Give agents Read own reviews so they see their scores, and discuss reviews in one-to-ones.
  4. Add the customer's voice: put an NPS survey on your email queues and web chats, and check the Rating NPS statistics monthly next to Rating QA. When the team is used to the form, consider AI QA on a sample of activities to raise coverage.

Permissions

What each person can do is set in their user access (Settings → Users → Accesses); which queues and users they work with comes from their user rights. See How Users & Rights Work.

Who Set in Permission
Admin who builds QA forms User access The QA forms module
Reviewer (team leader) User access QA reviews Create and Read all (QAs), and Update / Delete if they may change reviews
Agent User access Read own reviews (QAs) only
Anyone running AI QA on demand from an activity User access AI ondemand actions
Admin who builds CSAT forms; anyone reading CSAT surveys User access CSAT — see CSAT Forms → Permissions

Reviewers only see activities, and reviews of activities, from the queues and users their user rights cover.