Generation 4 is Voyc's newest Scorecard builder. The core process is the same you already know: Groups, Statements and Options with Conditions. What has changed is the builder itself: a cleaner canvas, a more intuitive design and a faster setup experience. The biggest new additions are Smart Statements and Smart Detection Conditions on your Options.
Some moments are harder to capture with exact words, as with any keyword-only Scorecard: a discovery question that agents phrase differently every time, a compliance check where exact wording varies or a nuanced response that could come out a hundred different ways. Smart Statements handle those moments by reading for meaning rather than matching words. You get both in a single builder: Keyword Statements where the language is predictable and Smart Statements where it isn't. You decide which approach fits each part of the conversation.
The result is more accurate scoring and when paired with Smart Detection Conditions on your Options, more reliable call routing too.
This is the complete setup guide for Generation 4 Scorecards. For background on Groups, Statements and scoring logic, see How to Create a Scorecard on Voyc.
Scorecard Terminology
Get your bearings before you start building. The table below covers every term you will come across on the canvas.
The suggested build order of your Scorecard is:
Create your Groups —> Create your Statements and connect them —> Validate —> Publish —> Activate —> Refine
Term | What it means |
Canvas | The visual workspace where you build and arrange your Scorecard. |
Group | A container for a section of your Scorecard, representing a main category of your call, for example Introduction or Call Closure. |
Statement | A Keyword Statement that detects specific words and phrases. If a Keyword is found in the transcript, this will note this Statement as said or passed, contributing to the Group score. |
Smart Statement | An AI-powered Statement that scores by context and meaning using Voyc's LLM, rather than matching specific keywords. |
Option | A branching tool that routes a call down a different path based on conditions. Used when a conversation can take more than one direction. |
Node | The general term for any individual item on the canvas, whether a Statement, Smart Statement or Option. You will see this term in the Smart Nodes counter and in validation messages. |
Connections | The links between Statements and Options that map the flow of the conversation from start to finish. |
Statement Type | Classifies what a Statement is measuring: Sales, Compliance or both. Setting a Statement to Compliance unlocks Auto-Fail and Negative Scoring. |
Assessment Criteria | The description the LLM reads to decide whether a Smart Statement has been met. Written in plain language describing what should happen in the conversation. |
Scripting | The phrase the agent should speak at a given point in the conversation. For Smart Statements, scripting also helps the LLM understand what an ideal response looks like. |
Auto-Fail | A setting that automatically fails the entire conversation if a Compliance Statement is not detected. |
Validation | A check that runs before publishing to confirm every Statement has what it needs and all Statements and Options are correctly linked. |
Smart Nodes counter | Displayed in the top right of the canvas, this shows how many Smart Statements and Smart Conditions you have used against your allowance of 75 Smart Nodes. |
Creating a Scorecard
Go to your Channel and click Configurations.
Open the Scorecards tab.
Click + Add Gen4 Scorecard in the top right corner of the page.
In the Create New Scorecard modal, give your Scorecard a Name. A Description is optional but useful when you are managing multiple Scorecards across a Channel.
Click Create.
You are now in the Generation 4 Scorecard canvas.
💡 Tip: You can edit your Scorecard name and description at any time by clicking the Scorecard name in the top left of the canvas.
Navigating the Scorecard Canvas
Think of the canvas as a visual map of every route a conversation could take. Every Scorecard has a clear starting point and a clear end, but in between, calls can branch off in different directions depending on what was said. Mapping those routes is how you show Voyc what to listen for along each path.
Here's what you'll find on the canvas.
Scorecard name and status
In the top left, your Scorecard name sits alongside a pencil icon. Click either to edit the name and description at any time. Next to the name is a status pill showing the current state of your Scorecard, for example Draft or Active.
Validate and Publish and Activate Actions
Directly below the Scorecard name are your Validation, Publishing and Activation tools. These change depending on your Scorecard's status.
Validating checks whether your Scorecard is ready to publish and flags any issues along the way. Run it as often as you like while you build to stay on track.
Adding items to your Canvas
You can add items to the canvas by dragging from the left panel or using the quick-add + menu on a Group or at the top right of any existing statement. Either way, the four building blocks are:
Group: a container for a section of your Scorecard, think of it as a main category of your call. Statements cannot exist outside of a Group
Statement: a Keyword Statement that detects specific words and phrases within a Group
Smart Statement: an AI-powered Statement that scores by context and meaning
Option: a branching tool that routes your call down a different path
The quick-add + menu also includes Duplicate, which copies the current statement including its name, assessment criteria and scripting. This works for both Keyword and Smart Statements.
Zoom and navigation
At the bottom left of the canvas, use the zoom controls to zoom in and out, fit the full Scorecard to your screen, or lock the items on the Canvas while you work. You can also use your mouse scroll wheel to zoom from wherever your cursor sits on the canvas. To move around the canvas, click and drag on any empty area of the background.
Smart Nodes counter and search phrases
In the top right, the Smart Nodes counter shows how many Smart Statements and Smart Conditions you have used against your allowance of 75 Smart Nodes. Once you reach your limit, new Smart Statements and Smart Conditions cannot be added until your allowance is increased.
The search bar in the top right lets you search for any keyword or phrase already used across your Scorecard. Use it to check whether a phrase is already covered elsewhere before adding it to a new statement, or to quickly find which statement contains a particular keyword.
Moving items
To reposition a Group or Statement, click and drag the drag handle (the six-dot icon at the top of the item). To reposition an Option, click and drag it directly. Click on a Group to extend it by dragging the blue dot on the bottom right.
Reading the Canvas at a Glance
Once you have Statements on the canvas, each card gives you a quick summary of what that Statement is doing without you needing to open it.
Card colour: A purple header on a Statement card means it is a Smart Statement. Keyword Statements have a plain white card.
Dot colour next to the name: Indicates the Importance level set for that Statement, from blue for Low through to red for Critical. A quick way to spot where the heaviest-weighted Statements sit across your Scorecard.
Circled exclamation mark: Auto-Fail is enabled. If Voyc does not detect this Statement in a call, the conversation will fail automatically.
Card description text: Your Assessment Criteria or description appears directly on the card, so you can read what each Statement is checking without opening it.
Pills at the bottom of the card:
Compliance or Sales: the Statement Type
Scripting Verbatim: the agent must read the script word for word
X Keywords: how many keywords are currently saved on this Statement
Once you are familiar with the Canvas, you can begin creating your Scorecard.
Creating Groups
Groups represent the sections of your agent's guide document and will be visible in your Grading Workflow. Each Group is a container for the Statements and Options that belong to that part of the conversation. For example, a Scorecard for an insurance call might have several Groups which may include: Introduction, Policy Discussion and Affordability Check.
Drag the Group icon from the left action panel onto a blank area of the canvas.
Give the Group a clear Name that reflects the section it represents.
Set the Group points to control how much this section contributes to the overall Scorecard score. We generally suggest a score that reflects how critical the group is. The numbers don't need to add up to anything in particular, as long as you weight them consistently across the Scorecard.
Click Save.
Repeat until you have a Group for each section of your conversation.
Adding Statements
Statements are how you teach Voyc what to listen for. Every call is scored based on what was said in the transcript and what wasn't, matched against the Statements you have built.
Generation 4 Scorecards support two Statement types:
Keyword Statement: Detects specific words or phrases an agent uses. Best for fixed scripts, compliance-critical language or predictable phrases.
Smart Statement: Uses Voyc's LLM to understand context and meaning rather than matching keywords. Better for variable phrasing, complex criteria or nuanced conversation moments. Each Smart Statement includes visible AI reasoning, so you can see exactly why a score was given.
Choosing the Right Statement Type
The right choice comes down to one question: how predictable is the language for this moment in the call?
Use Keyword when the required phrasing is straightforward and you can list the likely variations without much effort. A greeting, a compliance disclosure or a specific product mention typically fall here. For example:
An agent introduction can be reliably detected with a small set of keywords: hello, good morning, good afternoon, speaking to, my name is.
A recording disclosure can be covered with a short list: "This call may be recorded," "Calls are recorded," "Being recorded."
Use Smart when the required behaviour is nuanced, context-dependent or too variable to capture with keywords alone. For example:
Confirming all policy details are accurate before proceeding: the phrasing varies too widely per agent to cover with a keyword list.
Checking whether a customer has already received a competitor quote: this is a discovery moment that can surface naturally in conversation in too many forms to pin down with keywords.
A quick rule of thumb: if you can write the keyword list in a minute and feel confident it covers the variation, use Keyword. If you find yourself writing long lists and still feeling uncertain, use Smart.
You can mix both types freely within the same Scorecard and easily switch between Smart and Keyword.
📝 Note: You have an allowance of 75 Smart Nodes. The Smart Nodes counter on the canvas shows how many Smart Statements and Smart Conditions you have used against that allowance.
Adding a Statement
Click the + next to a group name, or drag a Statement or Smart Statement from the left panel into the group. The Edit Statement panel will open.
Fill in the Details section first. Once your Details are complete, use the Keyword / Smart toggle at the top of the panel to set how the statement detects whether it has been met.
Then configure the remaining sections covered below.
Switching to Smart: A confirmation modal appears showing your remaining Smart Statement allowance. It asks you to confirm your Assessment Criteria before switching. Click Switch to Smart to confirm.
Switching to Keyword: A confirmation modal explains that the statement will now score calls by matching exact words and phrases instead of using AI assessment, so scoring will change. Your name, assessment criteria and scripting are all kept. Click Switch to Keyword to confirm. You can switch back at any time.
💡 Tip: You can switch the Statement between Keyword and Smart at any point. Switching retains all your information, it just changes how Voyc assesses that Statement.
Click on your Statement to edit it and toggle between the two options at the top of the panel. Check your Criteria is accurate and make the switch.
Configuring a Statement
Every Statement uses the same five-section accordion layout, whether Keyword or Smart: Details, Keywords, Impact, Connections and Groups. Click any section header to expand or collapse it.
Details
The Details section is where every Statement starts, regardless of type.
Name (required): A clear label anyone reading the Scorecard can understand at a glance.
Assessment Criteria (required): For Smart Statements, this is what the LLM reads to decide whether the Statement has been met. Write what should happen in clear, natural language. For Keyword Statements, use this to describe the intent of the Statement for your team's reference.
A simple structure to follow:
[Who] + [What should happen] + [Any context that matters]
For example: "The agent states that all policy details are accurate before closing the call."
📝 Note: Criteria, not a prompt. Write it the way you would explain what should happen to a colleague: describe the behaviour in the conversation, not how it feels. A sentence or two is enough.
When editing a Smart Statement, you will find a How to write your criteria guide built into the Details section. It includes Do's and Don'ts for each principle. Here is a summary:
Tip | Do | Don't |
Describe the behaviour, not the words | The agent acknowledges the client's frustration before offering a solution. | The agent says "I understand your frustration." |
Name who is acting | The agent confirms the client's date of birth before discussing the account. | Date of birth is confirmed. |
Add a condition or example | When the customer mentions a complaint, the agent provides the ombudsman details within the same call. | Agent mentions ombudsman. |
Write out lists with numbers | The agent asks at least 3 of the following 4: (1) full name, (2) date of birth, (3) postcode, (4) address. | The agent asks the customer the DPA questions. |
Keep one criterion per statement | The agent states the call is being recorded. | The agent states the call is recorded, verifies identity, and explains the reason for the call. |
Scripting: The phrase the agent should speak at this point in the conversation. For Smart Statements, scripting also helps the LLM understand what an ideal response looks like, which improves scoring accuracy. Enable Scripting Verbatim if the agent must read it word-for-word.
Statement Type: Toggle Sales, Compliance or both. This tells the system what type of behaviour this Statement is measuring, and unlocks Auto-Fail for Compliance Statements.
💡Tip: Keep each Smart Statement focused on one behaviour. Avoid combining multiple requirements into a single Smart Statement unless they are genuinely inseparable.
Keywords
Keywords aren't just for Keyword Statements. Both types use them, but the role they play is different. For Keyword Statements, they are what Voyc actually uses to detect whether the Statement was said: essential, not optional. For Smart Statements, they act as examples that help the LLM recognise what you are listening for. Not required, but strongly recommended, around five solid phrases can make a real difference to accuracy.
There are three ways to add keywords:
Add manually: Type keywords directly into the block provided and select Add or Enter/Cmd, separating multiple entries with commas. You can add up to 100 keywords per statement.
Ask Joyc for Suggestions: Joyc uses your Assessment Criteria to generate AI-powered keyword suggestions, each with a relevance score. Select the ones that fit. Hover over any suggestion to edit the wording before accepting it. Click Accept Selected Keywords to add them in one go.

Search conversations: Type a keyword or phrase to search your real call transcripts. Voyc returns matches showing the actual conversation snippets where that phrase was said, so you can see exactly how agents phrase it in practice. Click the + next to any result to add that phrase as a keyword. Added keywords appear below and can be edited or deleted.
You can delete or edit a keyword by selecting the pencil or the delete icon on each Keyword.
❗Important: Keep keywords specific and unique to each statement. Avoid generic words that could appear anywhere else in a conversation. For example, "security" is too broad: it might be picked up from "I have a security system that keeps going off" when you only intended it to signal a recording disclosure. Whole phrases or distinctive word combinations work better than single common words.
Impact
The Impact section is where you decide how much weight this Statement carries in the overall Scorecard score.
Importance: Select Low, Medium, High or Critical. This helps Voyc understand how important this is in your script or conversations and how much it should contribute to scoring.
Auto-Fail: When enabled, a conversation will fail compliance if this Statement is not detected. Auto-Fail is available for Compliance Statements only. Enable the Compliance toggle in Statement Type to activate it.
Negative Scoring: When Auto-Fail is active, enter a percentage to deduct from the conversation scoring if the Statement is missed.
❗ Important: Only enter positive values in the Negative Scoring field. Voyc does not process negative input values. For example, entering 40 deducts 40% from the agent's overall score.
Connections
Connections define where the conversation flows after this Statement. Click on the Target field and select the Statement that should follow this one. You can also connect Statements and Options by selecting the dark dot on either side of the Statement or Option until a plus sign appears, then click and drag to the desired Statement or Option.
Groups
The Groups section lets you move a Statement to a different Group after creating it. Select the destination from the Group dropdown. You can also drag the Statement into the new Group.
💡Tip: Need to space out your Scorecard? You make your groups bigger by clicking the group and dragging the bottom right corner outwards (the blue dot).
Adding Options
Options are branch points in your Scorecard that represent the different directions a call can take. Not every conversation follows the same script: an agent might ask which plan a customer prefers, or whether they want to proceed or cancel, and each answer leads the call somewhere different. An Option captures each of those possible directions.
When Voyc grades a call, it analyses every possible route through your Scorecard and works out which one the conversation most likely followed. This is called routing. The route Voyc selects determines which Statements are checked and how the call is scored. Adding Conditions to your Options gives Voyc stronger signals when working out which route a call most likely followed. The more context Voyc has about a path, the more reliably it can identify the right one.
💡Tip: For full setup instructions on each Condition type, see How to Add Conditions to your Scorecard Option Nodes.
Click the + next to a Statement, or drag an Option from the left panel into the Group. The Edit Option panel will open.
In the Details section, give the Option a clear name, for example Outbound Call or Option A, and click Save.
Open the Routing section. This is where you tell Voyc what signals indicate a call followed this path. Click Add a condition, then + Add a Condition in the modal. Choose from the four available types and configure it, then click Close.
Phrase: specific words or phrases that signal this route was taken
Length: whether the call fell within a certain duration
Attributes (Beta): a known call property, such as call direction, applied before the conversation is analysed
Smart Detection (Beta): the meaning of what was said, described in plain language rather than specific words
You can add one Smart Detection Condition and multiple of the other Condition types using + Add a Condition again. All Conditions use AND logic: every one must be true for the route to match.
Use the Connections section to link the Option to the next Statement, or draw a connection line on the canvas. Repeat for each possible path.
Validate, Publish and Activate
Once your Scorecard is built, publish it and set it as active to put it live on your Channel. You can validate first to check everything is in place, but it isn't required before publishing.
❗Important: You can demote an Active Scorecard back to Published, but doing so will leave your Channel with no Active Scorecard and conversations will stop being auto-scored. You do not need to demote first to switch Scorecards. Simply promote any Published Scorecard directly and it will replace the active one.
Step 1: Publish
Click Publish on the left panel. A Publish Scorecard modal will confirm the Scorecard has been published and is ready to be promoted to active. Your Scorecard name will include a version number, for example Sales v3.
When you select Publish Voyc will run a validation first, where it checks each Statement and Option to ensure it has everything it needs to function before publish. If it fails, go to your Scorecard and it will tell you exactly what needs to be fixed. Statements that pass highlight in green. If something is wrong, the colour around that Statement or Option changes. If you open the item it will explain why it failed. Resolve any issues and select Validate to check if needed.
Step 2: Promote to Active
Click Promote to Active on the left panel. An Activate Scorecard modal will confirm the Scorecard is now live and scoring calls on your Channel.
Once active, the status pill changes to Active - read-only. To make changes, copy the Scorecard from the Scorecards tab, edit the copy, then validate, publish and promote the new version to active.
💡Tip: Once your Scorecard is active, run a handful of real calls through it before relying on it for reporting. Check whether Statements are being detected as expected, spot any Keywords that are too generic or missing, and refine as you go. A small amount of testing early saves a lot of manual correction later.
The Gist
Generation 4 gives you a redesigned, faster Scorecard builder: Keyword Statements for the predictable moments, Smart Statements for the nuanced ones, improving scoring accuracy by reading for meaning where exact phrasing varies. Pair Smart Statements with Smart Detection Conditions on your Options and the same intelligence applies to call routing too: conversations land on the right branch based on what was actually said, not just which keywords appeared.




















