Criteria
A criterion is a question you want answered on every conversation — for example, “Did the rep offer a follow-up meeting?” or “Which competitor did the client mention?”. Aelo asks it of each conversation in the criterion’s scope, once the conversation is transcribed and categorized — in its own step, separate from the analysis, so an answer does not depend on the analysis succeeding — and stores the answer with the conversation.
Not the same as evaluation profiles. A profile is the scorecard Aelo applies to a whole category of conversations, with its own set of metrics. A criterion is one question asked of every conversation, answered separately from the profile.
Your own criteria are included in every plan, up to the plan’s criteria limit. On top of them there are up to eight built-in criteria worded by Aelo; they do not count toward the limit. Answering criteria, built-in ones included, is part of the analysis and is covered by the per-minute price — there is no separate charge.
Demo data includes two demo criteria with sample answers on the demo conversations; demo data loaded before they were added comes without them. They appear only where results are shown — the Company criteria card of the dashboard and the criteria report, the call panel and its PDF report, and the call list filter — and not in the criteria settings or in alert rules; they are never asked of your real conversations, do not count toward the limit, and removing the demo data removes them.
Demo conversations also carry answers from your organization’s built-in criteria that are switched on, so the call panel and the criteria report show them as they would for a real call. The built-in criteria themselves belong to your organization, not to the demo: removing the demo data removes these answers but keeps the criteria. Demo data loaded before this change has no such answers.
| Type | What the answer is |
|---|---|
| Yes / no | Whether something happened. You choose what is looked for: that it happened, or that it did not happen (useful for “the rep never named the price”). |
| Score 0–100 | A score for how well something was done. |
| Choice from a list | One of your values — up to 10 active ones. Not found in the conversation is always added on its own and means the topic did not come up; if it came up but none of your values fits, the model answers that it is not sure: the Criteria report counts it as No answer, and the call does not match the Not found in the conversation filter. Calls processed before this rule keep their earlier answer until they are processed again. Mark a value Needs a quote when it must be backed by the conversation. |
An answer that points at something in the conversation comes with a verbatim quote. The answer type cannot be changed after the criterion is created — create a new criterion instead.
Open Analysis Setup → Criteria in the main menu. Only the organization Owner creates and changes criteria; an Admin sees the list read-only.

| Field | Limit | Tip |
|---|---|---|
| Name | 60 characters | Short — it is shown in filters and reports. |
| Question to the conversation | 300 characters | One question, one thing to check. |
| What counts — and what does not | 800 characters | Optional. Exclusions are the main lever of accuracy: say what must not count. |

Instead of a blank form, press Start from a template at the top of the new-criterion form and pick one of the ready-made wordings. They are grouped as Sales, Requests and service, and Rep work quality, and each shows its name and question:
- Yes / no: offered an additional product, asked about the budget, discovered the need with questions, stated the delivery or completion time, offered to place the order, named the next step in the case, the customer mentioned a competitor.
- One of several values: reason for contact — new purchase or order, product or service inquiry, status of an existing order, complaint or claim, technical support, return or exchange, another reason.
A template fills in the name, the question, the “what counts — and what does not” text, the answer type and the list values, with the texts in your organization’s language (Settings → Organization → Language). If you have already typed something, the form asks before replacing it; the scope you chose stays. A template is only a starting point and is not linked to the criterion afterwards: the saved criterion is an ordinary criterion of your own, you can rewrite any part of it, and it counts toward your plan’s criteria limit. Press Check on 20 calls to see how the wording behaves before you save it.

If no template fits, press Describe in your own words at the top of the new-criterion form and, in the “What do you want to find out?” field, say it in one or two sentences, up to 500 characters, for example: “I want to see calls where the customer complained about delivery”. Press Draft a criterion: the AI model suggests a name, a question, the “what counts — and what does not” text, an answer type and, for a list, its values — in the language of your description. The draft fills the form the same way a template does: if you have already typed something, the form asks before replacing it, and the scope you selected stays. Right after that the form runs the check on 20 calls by itself. If the description does not make a single check of a conversation, the form asks you to describe it differently. The draft is a starting point: read it through and adjust it before saving.

A criterion can be limited by conversation category, by project and by channel (calls, chats). Nothing selected means everywhere. The criterion is not asked of a conversation outside its scope. An answer saved earlier stays visible even when the conversation is no longer in scope — for example, after the scope was narrowed or the conversation was re-categorized.
The scope of a built-in criterion can only be narrowed within what Aelo set for it — for example, calls only. In the form such a dimension offers only the allowed values, and nothing selected means the whole allowed range (see Built-in Criteria).

Before saving, press Check on 20 calls: the current wording — not the saved one — is put to the last 20 analyzed conversations in the criterion’s scope, and you see each answer — with a quote when the answer points to a moment in the conversation. Nothing is saved, and the check is not charged to your balance. Checks are limited per organization — up to 20 a day on paid plans and 3 a day on the free plan (Pay as you go); when the limit is reached, try again later. Demo conversations are not part of the check: the wording cannot be tried on demo data, it needs your own analyzed conversations. The check takes conversations that are not archived, up to 30 minutes long and with more than 800 characters of text; until the criterion’s scope has some, it answers: “There are no analyzed conversations in this scope yet — nothing to check on.”

In the check results, press The answer is wrong next to an answer you disagree with and, if you like, add a short reason (up to 200 characters). Mark up to 10 answers and press Suggest a fix: the AI model reads the wording you checked, the marked answers, your reasons and fragments of those conversations around the quotes, and suggests new wording — with one sentence on what changed and the fields before and after. Only the question, the “what counts — and what does not” text and the labels of list values change; the name, answer type and the values themselves stay the same. Accept puts the edit into the form and runs the check again; Reject leaves the wording as it was. A conversation the check skipped cannot be marked. If the text of a marked conversation cannot be reliably cleared of personal data, no fix is suggested: the affected conversations are highlighted in the results, repeating will not help — remove the mark from them or mark other conversations.
Drafts and edits are not charged to your balance, and nothing is saved until you save the criterion. Their number per organization is limited, separately for each: up to 30 a day on paid plans and up to 5 a day on the free plan.
- Call panel — a Criteria group in the analysis panel. Each criterion shows its answer, or one of three states: Not found in the call, Not applicable (the model judged that the question does not apply to this conversation), Not evaluated (the criterion is active and covers the conversation, but there is no answer — for example, the evaluation failed or the criterion was paused at the time). A paused or archived criterion with no answer is not shown; as a rule, neither is a criterion created after the conversation was processed. A quote plays that moment of the recording. A built-in criterion appears here only when it has an answer from this processing of the conversation; the matching older row of the Compliance or Speech card (recording disclaimer, TCPA, personal data, profanity, mandatory phrases) is then not shown — never both. Without such an answer the older row stays as before, and a built-in you paused, with no answer, has no row anywhere. When all rows of the Compliance card have moved to built-in criteria, the card disappears. For a check paused by Aelo the older row stays, with the note “Paused by Aelo until re-measured”. New analyses do not write the older rows at all: they exist only for conversations analyzed before these checks moved to built-in criteria.
- PDF report — the same answers in the exported report, by the same rule.
- Call list filter — the Criteria section of the filters: up to three conditions, all of which must match — triggered or did not trigger, a score range, or list values (archived values and “not found in the conversation” included). The filter stays within the calls you can already see. A built-in criterion can be a condition too, but it matches only calls whose answer comes from the current analysis of the call; older calls without such an answer do not match.
- Dashboard → Quality tab → Company criteria card — for the selected period, every criterion with its counts: Evaluated, Not applicable, No answer (the model could not answer with confidence) and Not evaluated (analyzed calls created after the criterion appeared and within its current scope that have no result for it, the time it was paused included; not counted while the criterion is paused or archived); a paused or archived criterion and a change of wording in the period are marked. This count includes only calls created after the criterion appeared, so more calls can show Not evaluated in their panels. On the card Not evaluated is an upper bound (“at most N calls”): some calls may still be processing, the criterion may have been paused, or its step may not have run; the definition in brackets is the one the criteria report gives. Open in the report opens the criteria report (Analytics → Criteria report) at that criterion, with the card’s period and projects (the card has no such link for a period longer than 93 days). The report also gives the date the criterion took effect and the dates its wording changed; open a criterion to see it by day or week, the distribution of its answers and a table by rep. An answer in the distribution, a number in the table by rep and, for yes/no and score criteria, a bar of the trend open the calls behind them. A row is one person: a CRM id and the Aelo account of the same rep are one row, and the same id in two CRM portals is two rows (marked with their portal when the names match). The number in a row combined from several ids, or one whose id also belongs to another row, is plain text, not a link. A supervisor sees their own projects.
- Dashboard → Quality tab → Standard checks card — counted from built-in criteria: recording disclosed, TCPA and personal data (for personal data, “passed” means no violation); for calls with no answer from a built-in, the older analysis field is used. A check paused by Aelo is marked under the card.
- Alerts — a criterion can be a field of an alert rule: triggered, its score or its list value (see Alerts). For built-in criteria, the gallery templates Recording Disclaimer Missing (on Disclosed call recording), Rep used profanity and Rep used a forbidden word exist only on the criterion; the earlier Profanity detected template (any speaker) has been removed. A rule on a paused criterion, built-in or your own, does not fire, and its card says so.



Criteria are answered while a conversation is processed. Re-analyze in the call panel does not ask them again, so it adds no answers for criteria created or reworded later; neither does Reprocess of a conversation whose processing completed. Reprocess of a conversation whose processing or delivery failed runs the processing again, criteria included, once analysis runs for it (Automatic mode or a manual start).
- Name and the labels of list values can be changed at any time (for a built-in criterion the name and wording are fixed).
- Changing the question, what counts, what we look for, or the set of list values is a change of wording: answers before and after it are not strictly comparable. The Company criteria card marks the change and shows the date of the latest one; the criteria report shows the earlier ones and the date the criterion took effect. Calls before and after a change are counted together.
- Pause stops the criterion from being asked of new conversations; Resume brings it back. If you switch off a criterion of your own that enabled alert rules read, the row names those rules — they go silent while the criterion is paused — and offers Pause anyway or Keep active; nothing is paused until you confirm. Resuming asks nothing, and pausing a built-in criterion asks nothing either. A paused criterion of your own still counts toward the plan limit; built-in criteria never count.
- Archive retires a criterion for good. Its past answers stay in the call panel and remain available in the filter, and it no longer counts toward the plan limit. Built-in criteria cannot be archived — pause them instead.
- A list value is never deleted, only archived, so past answers keep their meaning.
Alongside your own criteria, Aelo asks up to eight built-in ones — questions that Aelo worded itself — of the conversations they apply to; which of them apply depends on the organization and the conversation, and ones paused by Aelo are not asked (see Pauses). They are visible to every organization by default: open Analysis Setup → Criteria, and they are in a separate Built-in section below your own criteria.
| Criterion | What it checks | Where it applies |
|---|---|---|
| Named the company | Did the employee name the company they are speaking on behalf of? | Any channel |
| Agreed on the next contact | Did the employee and the customer agree on a concrete next contact — a call, a message, sending materials or passing on a contact — no matter which of them proposed it? | Any channel |
| Disclosed call recording | Did the employee or the system say that the conversation is being recorded or may be monitored? | Calls only; starts paused for a new organization |
| Explained the satisfaction guarantee | Did the employee explain a satisfaction guarantee to the customer — a refund or exchange if the product or service does not suit them? | Any channel; only when the customer asks about a return, fears the item will not suit, or the rep offers a return on their own for that case |
| Rep used profanity | Did the employee use profanity or insulting language? | Any channel |
| Rep used a forbidden word | Did the employee use a word or phrase from the forbidden list? | Any channel; paused while the list has no words |
| Met TCPA requirements | Did the employee meet TCPA requirements on this outbound call — identify themselves, the company and the purpose of the call, and honor a request not to be called again? | Outbound calls about a purchase, a product inquiry or scheduling; only for organizations with a US jurisdiction |
| Broke personal data rules | Did the employee break the rules for handling the customer’s personal data? | Calls only; needs a concrete jurisdiction |
Which of them an organization gets depends on its jurisdiction: Met TCPA requirements exists only for organizations with a US jurisdiction, Broke personal data rules needs a concrete jurisdiction, and the wording of both names it. All of them are yes / no questions, and the answer is Yes when the event in the name happened — so Yes is good for “Named the company” and bad for “Broke personal data rules”.
Broke personal data rules counts when the rep asks for the full card number, CVV, PIN, a password or an SMS code; when, after a sign that the person who answered is not the customer (says they ordered nothing, that the number or order is not theirs, that they do not know this person), the rep still names the customer’s data — full name, address or pickup point, phone — before finding out who they are talking to; or when the rep reads the customer’s data to a third party. An ordinary call about an order to a person who accepts the order as theirs — greeting by name, reading back the order, amount, city, pickup point — is not a violation.
Explained the satisfaction guarantee applies only when the customer asks directly about a return or exchange, says “I’ll return it if needed”, directly voices fear or doubt that the item will not suit them (“what if the size is wrong?”), or the rep offers a return or exchange on their own if it doesn’t fit. Choosing a size or fit, measurements, asking “will it fit?”, changing the size before shipping, questions about features, compatibility, price or availability, turning down or postponing a purchase over price or to think it over, service, delivery, order status or payment — Not applicable, not “not explained”. So on most conversations this criterion shows Not applicable. A short confirmation from the rep to the customer’s question about a return or exchange (“yes, of course”) counts as an explanation; a time limit or conditions are not required.
- The wording is Aelo’s and is fixed: the name, the question and the “what counts” text cannot be changed. It is shown in your organization’s language (Ukrainian, Russian, English, Spanish or Portuguese; any other language gets English).
- The Owner can pause and resume a built-in criterion and narrow where it applies — projects, categories, channels — within the limits Aelo set for it. In the form, a limited dimension offers only the allowed values, and nothing selected means the whole allowed range.
- The Owner cannot widen the scope beyond those limits, change the direction of conversations, or archive a built-in criterion. Check on 20 calls is not available for it either.
The criterion Rep used a forbidden word takes its words from a list that the Owner keeps in the criterion’s own panel (Analysis Setup → Criteria → Built-in). The list holds up to 100 words or phrases of up to 64 characters each; duplicates, ignoring case, are dropped. Any form of a listed word counts — the model takes inflected forms into account (“cheap” also covers “cheaper” and “cheapest”) — and only the rep’s lines count, not the customer’s. Where the criterion applies — projects, channels, categories — is set by its scope, like any other criterion: “only project P” is the criterion’s scope, and there is no separate list in project settings. There is one such criterion per organization, so there is one list. Editing the list creates a new version of the wording: answers before and after an edit are not compared as one.
A paused built-in criterion shows the reason in words:
- Paused by Aelo: until the quality of this check is re-measured. Aelo pauses a built-in criterion this way until its check is re-measured; no criterion is paused this way at the moment. New conversations are not evaluated by it. You cannot resume it — the switch is not shown.
- Paused by Aelo: the wording does not fit your organization’s jurisdiction. It appears when the jurisdiction changed after the criterion was added. You cannot resume it either.
- Paused because this check was switched off in the analysis profile. You can resume it here. The analysis profile no longer has these switches; the note stays on criteria that were paused that way earlier.
- Paused: the word list is empty. Add words and the criterion switches on. It applies to Rep used a forbidden word: while the list has no words there is nothing to check, and the criterion cannot be switched on. The first save of a non-empty list lifts this pause, and clearing the list sets it again. A pause you set yourself is not touched by the list.
When you pause a built-in criterion yourself, new conversations simply do not get that check — the analysis does not make it in its place, because it no longer makes these checks at all. The attention flag never gives a paused check as the reason.
The analysis profile has no switches for Mandatory Phrases, Recording Disclaimer and Profanity Detection any more, and the main analysis no longer makes these checks or the compliance checks (personal data, TCPA): they are done only by built-in criteria. A new organization gets them from Aelo’s defaults: Named the company, Agreed on the next contact and Explained the satisfaction guarantee apply to Product Inquiry, Purchase, Scheduling and Follow-up conversations; Rep used profanity applies to all conversations; Disclosed call recording starts paused — you switch it on. Organizations that already have these criteria keep their scope and pauses. Conversations analyzed before this change keep the checks they stored.
Built-in criteria are outside the plan’s limit and inside the per-minute price — there is no separate charge.
| Plan | Criteria |
|---|---|
| Pay as you go | 5 |
| Starter | 5 |
| Pro | 10 |
| Business | 15 |
| Scale | 20 |
| Enterprise | 30 |
The limit counts your own criteria that are not archived; built-in criteria never count toward it. Moving to a smaller plan does not switch off criteria you already have — it only stops new ones from being created until you are under the new limit. The Pro trial does not change the limit. A free organization (Pay as you go) can have at most 5 criteria of its own during the trial too; the trial opens Pro features, not Pro limits.
In the embedded Bitrix24 app, organizations that have criteria turned on see the answers in the call panel and the PDF report of a call opened on the Records tab, just as in the web app. The Company criteria card of the Quality tab appears there too, as a report only: criteria are created and edited in the Aelo web app. The card’s Open in the report link, like the full criteria report, opens the Aelo web app in a new window; the first time, Aelo asks you to sign in (see Opening the full web app). Built-in criteria follow the same rules there as in the web app: their answers appear in the call panel, the PDF report and the dashboards. The CRM report gets a Criteria block with one line per answered built-in criterion; it replaces the older recording-disclaimer, mandatory-phrases and profanity lines for those checks, and the customer’s profanity is no longer sent to the CRM. The answers of “Disclosed call recording”, “Met TCPA requirements” and “Broke personal data rules” also fill the card’s Recording Disclaimer, TCPA Compliance and Data Privacy fields when you map them (see Bitrix24).
The criterion’s question and guidance are sent to the AI model together with the PII-redacted transcript, the same way the analysis itself is. Quotes are taken from the redacted transcript. Answers are deleted together with the conversation — including when it is removed by the retention period. See Privacy and PII.
The description you write for a draft and your reasons for marked answers are sent to the AI model the same way as a criterion’s wording: phone numbers, email addresses, card numbers and IBANs in them are masked. For an edit, fragments of the marked conversations are sent too — from the same redacted transcript. Neither the draft nor the edit is stored until you save the criterion.
Built-in criteria are answered on Aelo’s account, in Aelo’s wording: that wording goes to the AI model together with the PII-redacted transcript the same way as the wording of your own criteria, and their answers are deleted together with the conversation.