# Setting up quality evaluation forms and policies in Genesys Cloud Systems: Genesys Cloud CX For Genesys Cloud quality or contact-center administrators setting up or rebuilding quality evaluation forms and policies. Canonical: https://warmtransfer.net/guides/genesys-cloud-quality-evaluation-setup Last verified: 2026-10-02 In Genesys Cloud, evaluation forms are created from Menu > Conversation Intelligence > Quality Management > Evaluation Forms, and a policy can create evaluations by evaluators, by agents, by interaction, or as calibration evaluations[^49][^56]. This guide takes a standard Evaluation Form through publishing, policy assignment and calibration, in which several evaluators review the same interaction so a calibrator can compare their scores[^50][^30]. ## Before you start - Creating and publishing an evaluation form requires a Genesys Cloud CX 1 WEM Add-on II, CX 2, CX 3 or CX 4 license plus Quality administrator permissions[^47]. - AI Scoring requires a Genesys Cloud CX 3, CX 4, CX 2 WEM Add-on I, CX 1 WEM Add-on II or Genesys Cloud EX license plus Quality administrator permissions[^6]. - Quality management uses 2 default roles: Quality Administrator, who manages policies and forms, and Quality Evaluator, who reviews and scores interactions[^33]. - Creating a policy requires the Recording > Retention Policy Add, Edit and View permissions[^57]. - Call recording requires Recording to be enabled on the SIP trunk[^63]. - Recordings that match no policy are retained indefinitely, or up to the maximum interaction data retention time if one is set[^64]. - Evaluation forms offer 3 scoring features: default to highest scoring answer, fatal question and critical question[^62]. See also [Genesys Cloud licensing](https://warmtransfer.net/knowledge/genesys-cloud-licensing). See also [Genesys Cloud CX telephony and BYOC](https://warmtransfer.net/knowledge/genesys-cloud-telephony). ## What changes by situation - How should the policy assign evaluations? A fixed number per evaluator; A fixed number or percentage per agent; Every matching interaction. - Should AI Scoring pre-fill answers on the form? No, evaluators answer every question; Yes, AI Scoring pre-fills answers and a human evaluator reviews them. - Can agents dispute evaluations? Yes, allow disputes; No, evaluations are final. ## Step 1: Confirm licensing, roles and recording **No, evaluators answer every question** ### Do Confirm that the org has a Genesys Cloud CX 1 WEM Add-on II, CX 2, CX 3 or CX 4 license and that you have Quality administrator permissions[^47]. Give form and policy owners the Quality Administrator role and the people who score interactions the Quality Evaluator role[^33]. Give whoever creates policies the Recording > Retention Policy Add, Edit and View permissions[^57]. Confirm that Recording is enabled on the SIP trunk, which call recording requires[^63]. ### Verify Suggested check: place a test call through a queue you plan to evaluate and confirm it has an agent-side recording. A policy does not create an evaluation for an interaction that has no agent-side recording, although such an evaluation can still be assigned manually[^1]. ### Rollback Suggested rollback: remove the role and permission grants you added in this step. **Yes, AI Scoring pre-fills answers and a human evaluator reviews them** ### Do Confirm that the org has a Genesys Cloud CX 1 WEM Add-on II, CX 2, CX 3 or CX 4 license and that you have Quality administrator permissions[^47]. Also confirm a license that includes AI Scoring: Genesys Cloud CX 3, CX 4, CX 2 WEM Add-on I, CX 1 WEM Add-on II or Genesys Cloud EX[^6]. Give form and policy owners the Quality Administrator role and the people who score interactions the Quality Evaluator role[^33]. Give whoever creates policies the Recording > Retention Policy Add, Edit and View permissions[^57]. Confirm that Recording is enabled on the SIP trunk, which call recording requires[^63]. Confirm that voice interactions are transcribed with Speech and Text Analytics, which using AI Scoring on calls requires[^9]. Give evaluators Quality > Evaluation View Sensitive Data plus the Speech and Text Analytics and Recording view permissions, which seeing AI scoring in an evaluation requires[^10]. Budget AI Experience tokens: AI Scoring is metered at 20 evaluations scored with AI scoring per Genesys AI Experience token[^17]. An AI Scoring charge is incurred whenever a form with at least 1 AI-Scoring-enabled question is used to evaluate an interaction, but not when the evaluation-level AI request errors[^3]. ### Verify Suggested check: place a test call through a queue you plan to evaluate and confirm it has an agent-side recording and a transcript. A policy does not create an evaluation for an interaction that has no agent-side recording, although such an evaluation can still be assigned manually[^1]. AI Scoring requires the conversation transcript to contain at least 300 characters[^8]. ### Rollback Suggested rollback: remove the role and permission grants you added in this step. ## Step 2: Design the scoring model on paper ### Do Group the questions into question groups, and decide whether to enable Scaled Scoring, which assigns a numeric weighting factor to each question group[^61]. The evaluation score is the sum of each group's earned points times its weight divided by the sum of each group's possible points times its weight[^65]. Decide which questions are critical: critical questions must be scored Yes or No and produce a separate critical score alongside the overall evaluation score[^32]. Decide which questions are fatal: a fatal question scored No makes the whole evaluation score 0[^46]. Decide where N/A is allowed: a question answered N/A is ignored in a group's total and possible points, and an N/A group is excluded with the remaining group weights re-proportioned[^53]. ### Verify Suggested check: score a few sample interactions by hand with your formula, including one with an N/A group and one with a failed fatal question, and confirm the results match your intent. ## Step 3: Create the form and its question groups ### Do Go to Menu > Conversation Intelligence > Quality Management > Evaluation Forms and select Create[^49]. Choose the standard Evaluation Form type rather than the Agent Auto-Evaluation type[^50]. Check N/A Enabled Question Group on any group whose questions evaluators should be able to skip with an N/A box[^52]. Use a group's Answers Default To setting to default its questions to their highest scoring answer or to N/A, which evaluators can override[^51]. If you use Scaled Scoring, enter a numeric weighting factor for each question group[^61]. ### Verify An evaluation form holds at most 200 questions, with question text limited to 500 characters and form and question group names to 250 characters[^48]. ### Rollback Suggested rollback: delete the unpublished draft, or leave it unpublished. Division assignments cannot be changed after the form is published[^41], and a published form cannot be edited directly[^42]. ## Step 4: Add questions, critical and fatal flags, and automation **No, evaluators answer every question** ### Do For each fatal question, give it exactly 2 answers valued 1 (correct) and 0 (incorrect), and select both the Critical Question and Fatal Question checkboxes[^45]. Set each question's automation to No Automation rather than AI Scoring or Evaluation Assistance[^59]. ### Verify Suggested check: review every fatal question and confirm it also carries the critical flag and has only the correct and incorrect answers. A fatal question must have exactly 2 answers and both the Critical Question and Fatal Question checkboxes selected[^45]. ### Rollback Suggested rollback: edit or delete questions while the form is still a draft. Once a form is published, changes are made in a new form created from the last published version[^42]. **Yes, AI Scoring pre-fills answers and a human evaluator reviews them** ### Do For each fatal question, give it exactly 2 answers valued 1 (correct) and 0 (incorrect), and select both the Critical Question and Fatal Question checkboxes[^45]. Set automation to AI Scoring on the questions AI can judge, noting that multi-select questions support only Evaluation Assistance[^59]. Keep AI Scoring to at most 50 questions per form[^13]. Give each AI-scored question at most 5 answer options, with 500 characters per question or answer and 6,000 characters of help text[^7]. Genesys advises writing AI-scored questions as objective, transcript-only checks of a single behaviour with binary or specific answers and help text that says when each answer applies[^12]. Keep questions about queue routing, timestamps, duration, platform metrics, tone, attitude or inferred meaning on No Automation, because AI scoring cannot evaluate them[^2]. Spell answer labels consistently, including accents and special characters, because AI Scoring matches the model's response to answer options by exact text[^5]. ### Verify Suggested check: review each AI-scored question and confirm none depends on metadata, tone or call length. A form may contain at most 50 questions with AI scoring enabled[^13], and an AI-scored question may have at most 5 answer options[^7]. ### Rollback Suggested rollback: set a question back to No Automation while the form is still a draft. Once a form is published, changes are made in a new form created from the last published version[^42]. An AI Scoring charge is incurred whenever a form with at least 1 AI-Scoring-enabled question is used to evaluate an interaction[^3]. ## Step 5: Configure disputes on the form **Yes, allow disputes** ### Do On the form's Settings tab, turn on the Allow disputes toggle, which needs Quality > Evaluation Form Edit or Add permission[^37]. Set the maximum number of disputes per evaluation, up to a system limit of 5[^39]. Choose whether a disputed evaluation is reassigned to the original evaluator, to a specified person, or left unassigned for Process Automation[^36]. Grant agents Quality > Evaluation Dispute, and grant reviewers Quality > Evaluation Edit or Edit score and Quality > Evaluation Decline[^40]. ### Verify Suggested check: reopen the Settings tab and confirm Allow disputes is on, with your chosen maximum and assignee. ### Rollback Suggested rollback: turn Allow disputes off while the form is still a draft. Once a form is published, changes are made in a new form created from the last published version[^42]. **No, evaluations are final** ### Do Leave the Allow disputes toggle off on the form's Settings tab, since disputes are turned on per evaluation form with that toggle[^37]. Do not grant agents Quality > Evaluation Dispute, which disputing needs[^40]. ### Verify Suggested check: reopen the Settings tab and confirm Allow disputes is off. ## Step 6: Publish the form ### Do Set the form's divisions before publishing, because division assignments of an evaluation form cannot be changed after it is published and a new form version is required[^41]. Publish the form, which requires a Genesys Cloud CX 1 WEM Add-on II, CX 2, CX 3 or CX 4 license plus Quality administrator permissions[^47]. ### Verify Suggested check: confirm the form is listed as published and can be selected when you build a policy. ### Rollback A published evaluation form cannot be edited directly; changes are made in a new form created from the last published version[^42]. After a new version of a form referenced by policies is published, new interactions use the new published version[^58]. ## Step 7: Create the evaluation policy **A fixed number per evaluator** ### Do Create the policy with matching criteria drawn from conversation direction, users, work teams, time sets, queues, wrap-up codes, date range, conversation duration and customer participation[^55]. Add the action that creates evaluations by evaluators[^56]. Set a fixed number of evaluations per evaluator per time interval; this method evaluates the first agent connected to the interaction[^25]. Expect policy-driven evaluation creation to be capped at 50 evaluations per agent per day, 175 per week and 700 per month[^21]. Enable the policy, noting that policies generate evaluations only for interactions that occur after the policy is enabled[^54]. ### Verify Suggested check: after matching interactions occur, confirm each evaluator receives evaluations up to the configured count. ### Rollback Suggested rollback: disable or delete the policy. **A fixed number or percentage per agent** ### Do Create the policy with matching criteria drawn from conversation direction, users, work teams, time sets, queues, wrap-up codes, date range, conversation duration and customer participation[^55]. Add the action that creates evaluations by agents[^56]. Include users, teams or queues in the matching criteria, and sample either a fixed number of evaluations per period or a percentage of interactions with a minimum number[^23]. Select the evaluators; policy-created evaluations are spread evenly among them and any remainder is distributed randomly[^44]. This method evaluates the last agent that participated, excluding monitors and coaches[^24]. Expect policy-driven evaluation creation to be capped at 50 evaluations per agent per day, 175 per week and 700 per month[^21]. Enable the policy, noting that policies generate evaluations only for interactions that occur after the policy is enabled[^54]. ### Verify Suggested check: after matching interactions occur, confirm each matched agent receives the configured number of evaluations. A by-agents policy prorates the evaluation count for a new agent from the date of their first interaction[^22]. ### Rollback Suggested rollback: disable or delete the policy. **Every matching interaction** ### Do Create the policy with matching criteria drawn from conversation direction, users, work teams, time sets, queues, wrap-up codes, date range, conversation duration and customer participation[^55]. Add the action that creates evaluations by interaction[^56]. This method evaluates every matching interaction, suits compliance monitoring or full coverage of a subset of agents, and evaluates the first agent connected[^26]. Expect policy-driven evaluation creation to be capped at 50 evaluations per agent per day, 175 per week and 700 per month[^21]. Enable the policy, noting that policies generate evaluations only for interactions that occur after the policy is enabled[^54]. ### Verify Suggested check: after matching interactions occur, confirm each one with an agent-side recording produced an evaluation, within the policy limits. A policy does not create an evaluation for an interaction that has no agent-side recording[^1]. ### Rollback Suggested rollback: disable or delete the policy. ## Step 8: Calibrate the evaluators ### Do Run a calibration, in which several evaluators review the same interaction so a calibrator can compare their scores and standardise evaluation[^30]. For a one-off calibration, open an interaction's Quality Summary tab and choose a form, evaluators and optionally an expert evaluator to index scores against[^28]. For ongoing calibration, create a policy with the Assign Calibration action, naming the form, evaluators, an expert evaluator and a calibrator; the policy's creator is the calibrator by default[^29]. Calibration policies assign at most 50 evaluations per evaluator per day[^27]. ### Verify The Calibration Results tab shows each evaluator's score with the high, low and average score and the variance[^31]. Suggested check: where the variance is wide, review the wording of those questions with the evaluators. ### Rollback Suggested rollback: delete the one-off calibration, or disable the calibration policy. ## Step 9: Test end to end **No, evaluators answer every question** ### Do Place test interactions that match the policy criteria after the policy is enabled, because policies generate evaluations only for interactions that occur after the policy is enabled[^54]. ### Verify Suggested check: include a test interaction that should fail a fatal question and confirm an evaluation is created for it. A fatal question scored No makes the whole evaluation score 0[^46], and critical questions produce a separate critical score alongside the overall evaluation score[^32]. ### Rollback Suggested rollback: delete or ignore the test evaluations, and narrow the policy criteria if unexpected interactions matched. **Yes, AI Scoring pre-fills answers and a human evaluator reviews them** ### Do Place test interactions that match the policy criteria after the policy is enabled, because policies generate evaluations only for interactions that occur after the policy is enabled[^54]. Sources disagree on the bar for comparing AI and human answers: the AI scoring best-practices article says to test a form across at least 25 interactions and aim for 85 percent or more agreement[^16], while the AI Scoring FAQ says testing needs at least 20 interactions and shows 80 to 100 percent agreement as green[^15]. ### Verify An AI-prefilled answer shows an AI icon and the reason the answer was chosen[^14]. AI Scoring prefills an answer only where the model's answer has high confidence, so some questions may be left for the evaluator[^11]. Inconsistent accents or special characters in answer labels cause mismatches, because AI Scoring matches the model's response to answer options by exact text[^5]. ### Rollback Suggested rollback: move questions with low agreement back to No Automation. Once a form is published, that change is made in a new form created from the last published version[^42]. ## Step 10: Evaluate and release to agents **No, evaluators answer every question** ### Do Evaluators answer the questions and, before submitting, pick when the evaluation is released to the agent, no more than 45 days in the future[^60]. ### Verify Suggested check: confirm the agent can see the evaluation after the chosen release date. **Yes, AI Scoring pre-fills answers and a human evaluator reviews them** ### Do Evaluators review each AI-prefilled answer and its reason and can change it before submitting[^14]. Before submitting, the evaluator picks when the evaluation is released to the agent, no more than 45 days in the future[^60]. AI scoring is limited to 50 evaluations per agent per day, which customer care can raise to 100 or 200 on request[^4]. ### Verify Evaluators can modify or override any AI-provided answer[^11]. Suggested check: open a submitted test evaluation and confirm it shows the evaluator's final answers, including any overrides, and that the agent can see it after the release date. ## Step 11: Run the dispute workflow **Yes, allow disputes** ### Do Have a test agent with Quality > Evaluation Dispute dispute a released evaluation[^40]. A submitted dispute moves the evaluation to the In Review state, where the evaluator either revises the score or declines the dispute[^38]. ### Verify The assignee of a disputed evaluation gets an inbox notification[^36]. After the configured maximum number of disputes, the evaluation cannot be disputed[^39]. ### Rollback Suggested rollback: decline the test dispute. Declining a dispute needs Quality > Evaluation Decline[^40]. **No, evaluations are final** ### Do Confirm that the form's Allow disputes toggle is off and that agents lack Quality > Evaluation Dispute, which disputing needs[^37][^40]. ### Verify Suggested check: sign in as a test agent and confirm a released evaluation offers no way to dispute it. ## Step 12: Report and maintain ### Do Open the Agents Evaluation Summary view from Analytics Workspace, filter by date and user attributes, and export it[^43]. When questions change, publish a new form version; after a new version of a form referenced by policies is published, new interactions use it[^58]. ### Verify The view's form filter covers only the most recent published form name, so filter older evaluations by date[^43]. ## Applicability Applies to: Genesys Cloud CX. Deployments: multi-tenant. Sources checked 2026-10-02. This guide covers the standard Evaluation Form type, not the Agent Auto-Evaluation type[^50]. An Agent Auto-Evaluation form submits automatically with Virtual Supervisor as evaluator, and every question must use AI Scoring or Evaluation Assistance or the form cannot be published[^19]. Quality policies do not support Agent Auto-Complete evaluations, which are generated by agent scoring rules, the Evaluations API or ad hoc[^18]. Auto-complete evaluation data is not currently included in Quality Management reports[^20]. ## What remains uncertain Sources disagree on the AI agreement bar: one Genesys article sets at least 25 interactions and 85 percent agreement[^16], and the FAQ sets at least 20 interactions with 80 to 100 percent shown as green[^15]. Whether evaluations a policy has already created are affected when that policy is disabled or deleted is not covered by the sources below. How Evaluation Assistance answers are configured, beyond its being one of the automation options, is not covered by the sources below. Screen recording and survey policies are not covered by the sources below. How long an agent has to open a dispute, or a reviewer to resolve one, is not covered by the sources below. ## Sources [^1]: A policy does not create an evaluation for an interaction that has no agent-side recording, although such an evaluation can still be assigned manually. Source: [Create a policy](https://help.genesys.cloud/articles/create-recording-policy/), Create a policy, evaluation actions note. Checked 2026-10-02. [^2]: AI scoring cannot evaluate queue routing, timestamps, duration, platform metrics, tone, attitude or inferred meaning. Source: [AI scoring best practices](https://help.mypurecloud.com/articles/ai-scoring-best-practices/?theme=simplified), AI scoring best practices, limitations. Checked 2026-10-02. [^3]: An AI Scoring charge is incurred whenever a form with at least one AI-Scoring-enabled question is used to evaluate an interaction, but not when the evaluation-level AI request errors. Source: [AI Scoring FAQs](https://help.genesys.cloud/faqs/category/ai-scoring/), AI Scoring FAQs, billing question. Checked 2026-10-02. [^4]: AI scoring is limited to 50 evaluations per agent per day, which customer care can raise to 100 or 200 on request. Source: [AI Scoring FAQs](https://help.genesys.cloud/faqs/category/ai-scoring/), AI Scoring FAQs, evaluation limits question. Checked 2026-10-02. [^5]: AI Scoring matches the model's response to answer options by exact text, so inconsistent accents or special characters in answer labels cause mismatches. Source: [AI Scoring FAQs](https://help.genesys.cloud/faqs/category/ai-scoring/), AI Scoring FAQs, default answer or error despite evidence question. Checked 2026-10-02. [^6]: AI Scoring requires a Genesys Cloud CX 3, CX 4, CX 2 WEM Add-on I, CX 1 WEM Add-on II or Genesys Cloud EX license plus Quality administrator permissions. Source: [Automated scoring options in evaluation forms](https://help.genesys.cloud/articles/automated-scoring-options-in-evaluation-forms/), Automated scoring options in evaluation forms, prerequisites. Checked 2026-10-02. [^7]: An AI-scored question may have at most five answer options, with 500 characters per question or answer and 6,000 characters of help text. Source: [Automated scoring options in evaluation forms](https://help.genesys.cloud/articles/automated-scoring-options-in-evaluation-forms/), Automated scoring options in evaluation forms, limits. Checked 2026-10-02. [^8]: AI Scoring requires the conversation transcript to contain at least 300 characters. Source: [Automated scoring options in evaluation forms](https://help.genesys.cloud/articles/automated-scoring-options-in-evaluation-forms/), Automated scoring options in evaluation forms, limits. Checked 2026-10-02. [^9]: Using AI Scoring on calls requires the interactions to be transcribed with Speech and Text Analytics. Source: [AI Scoring FAQs](https://help.genesys.cloud/faqs/category/ai-scoring/), AI Scoring FAQs, question on whether STA is required. Checked 2026-10-02. [^10]: Seeing AI scoring in an evaluation requires Quality > Evaluation View Sensitive Data plus Speech and Text Analytics and Recording view permissions. Source: [Evaluate an interaction](https://help.mypurecloud.com/?p=4296), Evaluate an interaction, prerequisites. Checked 2026-10-02. [^11]: AI Scoring prefills an answer only where the model's answer has high confidence, and evaluators can modify or override any AI-provided answer. Source: [Automated scoring options in evaluation forms](https://help.genesys.cloud/articles/automated-scoring-options-in-evaluation-forms/), Automated scoring options in evaluation forms, AI Scoring section. Checked 2026-10-02. [^12]: Genesys advises writing AI-scored questions as objective, transcript-only checks of a single behaviour with binary or specific answers and help text that says when each answer applies. Source: [AI scoring best practices](https://help.mypurecloud.com/articles/ai-scoring-best-practices/?theme=simplified), AI scoring best practices, question writing principles and help text. Checked 2026-10-02. [^13]: An evaluation form may contain at most 50 questions with AI scoring enabled. Source: [Create and publish an evaluation form](https://help.genesys.cloud/articles/create-publish-evaluation-form/), Create and publish an evaluation form, limits list. Checked 2026-10-02. [^14]: An AI-prefilled answer shows an AI icon and the reason the answer was chosen, and the evaluator can change it before submitting. Source: [Evaluate an interaction](https://help.mypurecloud.com/?p=4296), Evaluate an interaction, AI scoring section. Checked 2026-10-02. [^15]: The AI Scoring FAQ says agreement-score testing needs at least 20 interactions and shows 80 to 100 percent agreement as green (disputed). Source: [AI Scoring FAQs](https://help.genesys.cloud/faqs/category/ai-scoring/), AI Scoring FAQs, agreement score question. Checked 2026-10-02. [^16]: The AI scoring best-practices article says to test a form across at least 25 interactions and aim for 85 percent or more human-AI agreement (disputed). Source: [AI scoring best practices](https://help.mypurecloud.com/articles/ai-scoring-best-practices/?theme=simplified), AI scoring best practices, testing recommendations. Checked 2026-10-02. [^17]: AI Scoring is metered at 20 evaluations scored with AI scoring per Genesys AI Experience token. Source: [Genesys Cloud tokens model](https://help.genesys.cloud/articles/genesys-cloud-tokens-model/), Genesys Cloud AI Experience metering details table, AI Scoring row. Checked 2026-10-02. [^18]: Quality policies do not support Agent Auto-Complete evaluations; they are generated by agent scoring rules, the Evaluations API or ad hoc. Source: [AI Scoring FAQs](https://help.genesys.cloud/faqs/category/ai-scoring/), AI Scoring FAQs, quality policies question and generating auto-complete evaluations question. Checked 2026-10-02. [^19]: An Agent Auto-Evaluation form submits automatically with Virtual Supervisor as evaluator, and every question must use AI Scoring or Evaluation Assistance or the form cannot be published. Source: [Agent auto evaluation form](https://help.genesys.cloud/articles/agent-auto-complete-form-type/), Agent auto evaluation form, overview and configuration rules. Checked 2026-10-02. [^20]: Auto-complete evaluation data is not currently included in Quality Management reports. Source: [AI Scoring FAQs](https://help.genesys.cloud/faqs/category/ai-scoring/), AI Scoring FAQs, auto-complete data in reports question. Checked 2026-10-02. [^21]: Policy-driven evaluation creation is capped at 50 evaluations per agent per day, 175 per week and 700 per month. Source: [Create a policy](https://help.genesys.cloud/articles/create-recording-policy/), Create a policy, evaluation automation limits note. Checked 2026-10-02. [^22]: A by-agents policy prorates the evaluation count for a new agent from the date of their first interaction. Source: [Create evaluations by agents](https://help.genesys.cloud/articles/create-evaluations-per-agent-overview/), Create evaluations by agents, prorating rules. Checked 2026-10-02. [^23]: A by-agents policy samples either a fixed number of evaluations per period or a percentage of interactions with a minimum number, and requires users, teams or queues in its matching criteria. Source: [Create evaluations by agents](https://help.genesys.cloud/articles/create-evaluations-per-agent-overview/), Create evaluations by agents, configuration and sampling options. Checked 2026-10-02. [^24]: Assign evaluations by agents is the choice for an even distribution and a specific number of evaluations per agent, and it evaluates the last agent that participated, excluding monitors and coaches. Source: [Which evaluation policy should I use?](https://help.genesys.cloud/faqs/which-evaluation-policy-should-i-use/), Which evaluation policy should I use?, Assign evaluations by agents. Checked 2026-10-02. [^25]: Assign evaluations by evaluators is the choice for a fixed number of evaluations per evaluator per time interval, and it evaluates the first agent connected to the interaction. Source: [Which evaluation policy should I use?](https://help.genesys.cloud/faqs/which-evaluation-policy-should-i-use/), Which evaluation policy should I use?, Assign evaluations by evaluators. Checked 2026-10-02. [^26]: Assign evaluations by interactions evaluates every matching interaction, suits compliance monitoring or full coverage of a subset of agents, and evaluates the first agent connected. Source: [Which evaluation policy should I use?](https://help.genesys.cloud/faqs/which-evaluation-policy-should-i-use/), Which evaluation policy should I use?, Assign evaluations by interactions. Checked 2026-10-02. [^27]: Calibration policies assign at most 50 evaluations per evaluator per day. Source: [Create a calibration policy](https://help.genesys.cloud/articles/create-calibration-policy/), Create a calibration policy, note. Checked 2026-10-02. [^28]: A one-off calibration is created from an interaction's Quality Summary tab by choosing a form, evaluators and optionally an expert evaluator to index scores against. Source: [Conduct a calibration](https://help.mypurecloud.com/articles/conduct-calibration/), Conduct a calibration, procedure. Checked 2026-10-02. [^29]: A calibration policy uses the Assign Calibration action, naming the form, evaluators, an expert evaluator and a calibrator, and its creator is the calibrator by default. Source: [Create a calibration policy](https://help.genesys.cloud/articles/create-calibration-policy/), Create a calibration policy, procedure. Checked 2026-10-02. [^30]: A calibration has several evaluators review the same interaction so a calibrator can compare their scores and standardise evaluation. Source: [Calibration overview](https://help.genesys.cloud/articles/calibration-overview/), Calibration overview, introduction. Checked 2026-10-02. [^31]: The Calibration Results tab shows each evaluator's score with the high, low and average score and the variance. Source: [Conduct a calibration](https://help.mypurecloud.com/articles/conduct-calibration/), Conduct a calibration, Calibration Results tab. Checked 2026-10-02. [^32]: Critical questions must be scored Yes or No and produce a separate critical score alongside the overall evaluation score. Source: [Evaluation form weights](https://help.genesys.cloud/articles/evaluation-form-weights/), Evaluation form weights, critical score formula. Checked 2026-10-02. [^33]: Quality management uses two default roles: Quality Administrator, who manages policies and forms, and Quality Evaluator, who reviews and scores interactions. Source: [About quality management](https://help.genesys.cloud/articles/about-quality-management/), About quality management, roles. Checked 2026-10-02. [^34]: The Manage disputed evaluations article says agents can open a dispute within 45 days, after which disputing is blocked (disputed). Source: [Manage disputed evaluations](https://help.genesys.cloud/articles/manage-disputed-evaluations/), Manage disputed evaluations, introduction. Checked 2026-10-02. [^35]: The Allow evaluation disputes article describes the 45-day dispute timeframe as the number of days the evaluator has to review a dispute (disputed). Source: [Allow evaluation disputes](https://help.mypurecloud.com/articles/allow-evaluation-disputes/), Allow evaluation disputes, dispute timeframe setting. Checked 2026-10-02. [^36]: A disputed evaluation can be reassigned to the original evaluator, to a specified person, or left unassigned for Process Automation, and the assignee gets an inbox notification. Source: [Allow evaluation disputes](https://help.mypurecloud.com/articles/allow-evaluation-disputes/), Allow evaluation disputes, dispute assignment options. Checked 2026-10-02. [^37]: Disputes are turned on per evaluation form with the Allow disputes toggle on the form's Settings tab, which needs Quality > Evaluation Form Edit or Add permission. Source: [Allow evaluation disputes](https://help.mypurecloud.com/articles/allow-evaluation-disputes/), Allow evaluation disputes, procedure. Checked 2026-10-02. [^38]: A submitted dispute moves the evaluation to the In Review state, where the evaluator either revises the score or declines the dispute. Source: [Manage disputed evaluations](https://help.genesys.cloud/articles/manage-disputed-evaluations/), Manage disputed evaluations, dispute and review procedures. Checked 2026-10-02. [^39]: The administrator sets the maximum number of disputes per evaluation, up to a system limit of 5, after which the evaluation cannot be disputed. Source: [Allow evaluation disputes](https://help.mypurecloud.com/articles/allow-evaluation-disputes/), Allow evaluation disputes, Disputes settings. Checked 2026-10-02. [^40]: Disputing needs Quality > Evaluation Dispute; revising a disputed score needs Quality > Evaluation Edit or Edit score; declining needs Quality > Evaluation Decline. Source: [Manage disputed evaluations](https://help.genesys.cloud/articles/manage-disputed-evaluations/), Manage disputed evaluations, prerequisites per procedure. Checked 2026-10-02. [^41]: Division assignments of an evaluation form cannot be changed after it is published; a new form version is required. Source: [Create and publish an evaluation form](https://help.genesys.cloud/articles/create-publish-evaluation-form/), Create and publish an evaluation form, publish section note. Checked 2026-10-02. [^42]: A published evaluation form cannot be edited directly; changes are made in a new form created from the last published version. Source: [Edit an evaluation form](https://help.genesys.cloud/articles/edit-evaluation-form/), Edit an evaluation form, procedure. Checked 2026-10-02. [^43]: The Agents Evaluation Summary view is opened from Analytics Workspace, filters by date, user attributes and only the most recent published form name, and can be exported. Source: [Agents Evaluation Summary view](https://help.genesys.cloud/articles/agents-evaluation-summary-view/), Agents Evaluation Summary view, navigation, filters and export. Checked 2026-10-02. [^44]: Policy-created evaluations are spread evenly among the selected evaluators and any remainder is distributed randomly. Source: [Create evaluations by agents](https://help.genesys.cloud/articles/create-evaluations-per-agent-overview/), Create evaluations by agents, distribution paragraph. Checked 2026-10-02. [^45]: A fatal question must have exactly two answers valued one (correct) and zero (incorrect) and must have both the Critical Question and Fatal Question checkboxes selected. Source: [Configure a fatal question](https://help.genesys.cloud/articles/configure-a-fatal-question/), Configure a fatal question, procedure. Checked 2026-10-02. [^46]: A fatal question scored No makes the whole evaluation score zero. Source: [Evaluation form weights](https://help.genesys.cloud/articles/evaluation-form-weights/), Evaluation form weights, fatal question section. Checked 2026-10-02. [^47]: Creating and publishing an evaluation form requires a Genesys Cloud CX 1 WEM Add-on II, CX 2, CX 3 or CX 4 license plus Quality administrator permissions. Source: [Create and publish an evaluation form](https://help.genesys.cloud/articles/create-publish-evaluation-form/), Create and publish an evaluation form, prerequisites block at top of article. Checked 2026-10-02. [^48]: An evaluation form holds at most 200 questions, with question text limited to 500 characters and form and question group names to 250 characters. Source: [Create and publish an evaluation form](https://help.genesys.cloud/articles/create-publish-evaluation-form/), Create and publish an evaluation form, limits list. Checked 2026-10-02. [^49]: Evaluation forms are created from Menu > Conversation Intelligence > Quality Management > Evaluation Forms by selecting Create. Source: [Create and publish an evaluation form](https://help.genesys.cloud/articles/create-publish-evaluation-form/), Create and publish an evaluation form, procedure step 1. Checked 2026-10-02. [^50]: When creating a form the administrator chooses between the standard Evaluation Form type and the Agent Auto-Evaluation type. Source: [Create and publish an evaluation form](https://help.genesys.cloud/articles/create-publish-evaluation-form/), Create and publish an evaluation form, procedure, form type selection. Checked 2026-10-02. [^51]: A question group's Answers Default To setting can make its questions default to their highest scoring answer or to N/A, and evaluators can override the default. Source: [Add a question group to an evaluation form](https://help.genesys.cloud/articles/add-question-group-evaluation-form/), Add a question group to an evaluation form, Answers Default To option. Checked 2026-10-02. [^52]: When N/A Enabled Question Group is checked, evaluators can skip every question in that group with an N/A box. Source: [Add a question group to an evaluation form](https://help.genesys.cloud/articles/add-question-group-evaluation-form/), Add a question group to an evaluation form, N/A Enabled Question Group option. Checked 2026-10-02. [^53]: A question answered N/A is ignored in a group's total and possible points, and an N/A group is excluded with the remaining group weights re-proportioned. Source: [Evaluation form weights](https://help.genesys.cloud/articles/evaluation-form-weights/), Evaluation form weights, N/A questions and N/A groups examples. Checked 2026-10-02. [^54]: Policies generate evaluations only for interactions that occur after the policy is enabled. Source: [Create a policy](https://help.genesys.cloud/articles/create-recording-policy/), Create a policy, note on policy effectiveness. Checked 2026-10-02. [^55]: Policy matching criteria include conversation direction, users, work teams, time sets, queues, wrap-up codes, date range, conversation duration and customer participation. Source: [Create a policy](https://help.genesys.cloud/articles/create-recording-policy/), Create a policy, Matching Criteria section. Checked 2026-10-02. [^56]: A policy can create evaluations four ways: by evaluators, by agents, by interaction, or as calibration evaluations. Source: [Create a policy](https://help.genesys.cloud/articles/create-recording-policy/), Create a policy, Actions to Perform, evaluation actions. Checked 2026-10-02. [^57]: Creating a policy requires the Recording > Retention Policy Add, Edit and View permissions. Source: [Create a policy](https://help.genesys.cloud/articles/create-recording-policy/), Create a policy, prerequisites. Checked 2026-10-02. [^58]: After a new version of a form referenced by policies is published, new interactions use the new published version. Source: [Edit an evaluation form](https://help.genesys.cloud/articles/edit-evaluation-form/), Edit an evaluation form, note on policies. Checked 2026-10-02. [^59]: Each question's automation is set to No Automation, AI Scoring or Evaluation Assistance, and multi-select questions support only Evaluation Assistance. Source: [Create and publish an evaluation form](https://help.genesys.cloud/articles/create-publish-evaluation-form/), Create and publish an evaluation form, add questions, automation options. Checked 2026-10-02. [^60]: Before submitting, the evaluator picks when the evaluation is released to the agent, no more than 45 days in the future. Source: [Evaluate an interaction](https://help.mypurecloud.com/?p=4296), Evaluate an interaction, release and submit step. Checked 2026-10-02. [^61]: With the optional Scaled Scoring setting enabled, the administrator assigns a numeric weighting factor to each question group. Source: [Create and publish an evaluation form](https://help.genesys.cloud/articles/create-publish-evaluation-form/), Create and publish an evaluation form, question group settings, Scaled Scoring. Checked 2026-10-02. [^62]: Evaluation forms offer three scoring features: default to highest scoring answer, fatal question and critical question. Source: [Evaluation scoring](https://help.mypurecloud.com/?p=17497), Evaluation scoring, feature list. Checked 2026-10-02. [^63]: Call recording requires Recording to be enabled on the SIP trunk. Source: [Quality policies overview](https://help.genesys.cloud/articles/about-quality-policies/), Quality policies overview, recording note. Checked 2026-10-02. [^64]: Recordings that match no policy are retained indefinitely, or up to the maximum interaction data retention time if one is set. Source: [Quality policies overview](https://help.genesys.cloud/articles/about-quality-policies/), Quality policies overview, retention note. Checked 2026-10-02. [^65]: The evaluation score is the sum of each group's earned points times its weight divided by the sum of each group's possible points times its weight. Source: [Evaluation form weights](https://help.genesys.cloud/articles/evaluation-form-weights/), Evaluation form weights, scoring formula. Checked 2026-10-02.