# Inbound campaigns > How incoming calls are handled: the queue, the agent and what gets recorded ## A look at the module An **inbound campaign** is how Audara wraps up an incoming call operation: a group of agents answering calls that line up in a queue while they wait their turn. Everything that defines that operation lives in the campaign: on what criteria the calls are handed out, what the customer hears while waiting, what the agent has to record when hanging up and what gets analysed afterwards. The module is in **Setup > Voice Campaigns**, and its sibling, **Voice Groups**, is right underneath. > **Note** > Audara says **campaign**, not queue. The queue is the line of calls waiting; the campaign is all the configuration around that line. You will see the word "queue" in the name of some fields, always meaning the line. A campaign is not used on its own. It connects to the rest of the product on three sides: - **Where the calls come in**: An **inbound route** or an **[IVR](https://docs.audara.io/en/ivr/)** option points to the campaign. Without that, the campaign exists but never rings. - **Who answers it**: Agents and supervisors are not assigned to the campaign: they are assigned to a **group** of campaigns. So every campaign somebody is going to answer has to be in a group. - **What the agent works with**: The **CRM**, the script, the typing and the integration URL you configure here are what the agent gets in the panel when the call comes in. A second thing lives on the same screen, and it is not telephone work: **On-site**, for recording walk-in service. It is a separate entity with its own reduced form, and it has its section at the end of this article. ## The campaign list The first screen is the list of the campaigns of the account, one per row. ![The campaign list, with the search box and the create button at the top](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-lista.jpg) *The campaign list, with the search box and the create button at the top* The icon on the left tells you the type of each row without having to open it: the **blue telephone** is a normal inbound campaign, the **purple icon** is a campaign reserved for a predictive dialer and the **green handshake** is on-site service. - **Dialer**: The predictive dialer that has taken this campaign, if there is one. - **CRM**: The CRM that opens for the agent during the call. - **Survey**: **Yes** when the campaign has a post-call survey configured. - **Client**: The client name given to the campaign. It is a free label, useful when one installation serves several brands. - **Groups**: How many groups it is in. If the number is not zero, you can press it. - **Creation Date and Modification Date**: When it was created and when its configuration was last touched. - **Status**: **Active** or **Inactive**. An inactive campaign still exists and still shows in the list, but it stops receiving calls. The search box at the top filters by name, and the **+** button in the header opens the form for a new campaign. ### The menu of each row ![The actions menu of a campaign](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-acciones.jpg) *The actions menu of a campaign* - **Edit**: Opens the full form of the campaign. It is the same as pressing its name. - **View logs**: Shows the change history of that campaign: who touched it, when and what changed. - **Deactivate**: Leaves it **Inactive**. This is what you want when an operation ends but you need to keep its reports. - **Delete**: Erases it. > **Important** > Audara stops you if you try to **rename or delete** a campaign that is in use, and "in use" means two different things: that at that moment there is an **agent or a supervisor with an open session** in it, or that the campaign **belongs to some group**. The second one is the surprise, because it does not depend on the time of day: to change the name of a campaign that is in a group, you first have to take it out of the group. If the warning talks to you about a "group in use" while you are editing a campaign, this is the case. ### The groups of a campaign The number in the **Groups** column opens the list of the groups that campaign belongs to. ![The groups of a campaign: three, here](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-asignaciones.jpg) *The groups of a campaign: three, here* It is the window to look at before touching the name of a campaign that has been in production for a while, and also the quick answer to "why does this agent not see this campaign". ## Creating a campaign The **+** button opens **New Inbound Campaign**, a form of nine tabs that is saved in one go with the **Save** button in the header. The tabs are not steps: you can move between them in any order, and nothing is saved until you press Save. ![The new form opens on General, with Save turned off](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-crear.jpg) *The new form opens on General, with Save turned off* Save starts off and turns on when the form meets the minimum. If it does not turn on, it is for one of three reasons: a mandatory field is missing, you do not have write permission on the module, or there is a badly built typing. > **Important** > The third reason is the one that costs you time, because it **leaves no message on screen** and you can be standing on General seeing nothing strange. Save stays off if some **active** typing has a name shorter than three characters, ended up with no tags, or repeats the name of another active typing. If Save does not respond and General looks complete, check the **Typing** tab. There are two asterisks that do not tell the truth, and it is worth keeping them in mind: - **1.** **Autofill** carries an asterisk and can be left empty. - **2.** **Time in Queue**, in the FlowOut block, does **not** carry an asterisk and is in fact mandatory, with a minimum of 30 seconds. Editing a campaign opens the same form. There Save starts off for a different reason: it turns on when you change something. ## General: how the calls are handed out The **General** tab is the longest one and the only indispensable one. It defines the behaviour of the queue: how the agents are picked, how long each one rings and what happens when there is nobody. At the very top there are two switches that change the whole form: - **Predictive dialer**: Reserves the campaign for a predictive dialer. Only campaigns with this option on can be assigned to a dialer. - **On-site**: Turns what you are creating into walk-in service. The form shrinks to three tabs. It is explained at the end of the article. ![The second half of General Parameters](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-general-2.jpg) *The second half of General Parameters* - **Name***: The one that identifies the campaign all over Audara. Three characters minimum and it cannot be repeated. - **Client name**: A free label to group campaigns by brand or by end client. The **+** button next to it adds a new one to the list. - **Strategy***: On what criteria the agent who is going to ring is picked. It is explained separately, further down. - **Queue Size**: How many calls are accepted waiting at the same time. Empty means no cap. - **Music On Hold***: The music the customer hears while waiting. The options come from the **Audios** module. - **Announcement Message**: A short audio that **the agent** hears, not the customer, right before the call is handed over. It is there so they know which campaign to greet with when they handle several. - **Ring Time (sec)***: How long it rings on one extension before moving on to look for another. Out of the box, 15. - **Retry Time (sec)***: How long Audara waits before sending the call to the next extension. Out of the box, 5. - **WrapUp Time (sec)***: How long the extension rests after hanging up before another call from this campaign can come in. Out of the box, 1. - **Join Empty**: Whether a call can enter the queue when there is no agent connected. - **Leave Empty**: Whether the calls that are already waiting are pulled out of the queue when it runs out of agents. - **Autofill***: Hands out the calls in parallel instead of one by one, so a call further back can move ahead when there are free agents. Setting it to **No** can affect the distribution strategy, and the screen itself warns you. - **Simultaneous Calls***: How many calls from this campaign one extension can have at the same time. Out of the box, 1. It only applies to SIP channels. - **Transfers**: Enables the transfer button in the agent panel for the calls of this campaign. - **Pause Recording**: Lets the agent pause and resume the recording. This is what gets used when the customer has to read out sensitive data. - **Campaign Priority**: The higher the number, the more priority the calls of this campaign have over another one's when they compete for the same agent. Careful, this is different from the **Priority** tab, which hands out among agents. - **Context (Queue Out)**: Lets the customer leave the queue by dialling a digit during the wait. - **Exit to callback**: The digit exit, instead of leading somewhere else, schedules a call back. Turning it on enables the two CallBack fields in the block below. - **Allow individual queue logout**: Lets the agent leave this campaign without leaving the whole group. ### The distribution strategies The strategy decides who the call rings for. There are eight, and out of the box it is **rrmemory**: - **rrmemory**: By turns, remembering where the previous round left off. It is the most even distribution and it is the one that comes set. - **ringall**: Rings on every free extension at once and keeps the first one that answers. It distributes badly, but it answers fast. - **leastrecent**: To the agent who has gone longest without a call from this campaign. - **fewestcalls**: To the agent who has handled the fewest calls from this campaign during the shift. - **random**: At random. - **linear**: Always in the same order, the order in which the agents entered the campaign. It gets used when there is a real hierarchy of service. - **wrandom**: At random, but weighing the penalties of each agent. - **rrordered**: By turns respecting the order of the list, without the memory of rrmemory. > **Note** > The help that shows next to the field is out of date: it describes an option called "Roundrobin" that is no longer in the list and it does not mention **fewestcalls** or **rrordered**. The list above is the good one. ## FlowOut and CallBack: what happens if nobody answers Underneath the general parameters is **FlowOut**, which answers a single question: what to do with the call that has been waiting too long. ![FlowOut and CallBack, one after the other](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-flowout.jpg) *FlowOut and CallBack, one after the other* - **Time in Queue (sec)**: How long a call holds in the line before going off to the destination. The **minimum is 30 seconds**, and it comes with **9000** set out of the box, which is two and a half hours: in practice that is "never". If you want the FlowOut to be good for something, this is the number to bring down. - **Destination**: Where the call goes when the time is up: an extension, a voicemail, an IVR (and its option), a conference room or another campaign. > **In practice** > The pair **Time in Queue** plus **Destination** is the safety net of the operation: it is what keeps somebody from waiting into the void when the team disconnected or went to lunch. It is worth setting a realistic time, in the order of minutes, and a destination that actually resolves, like a voicemail somebody does check. The **CallBack** block below is for calling back instead of losing the call: - **Create Manual Callback**: Lets the agent schedule a call back themselves from the panel, using a dynamic Preview dialer with Callback. - **Queue Exit Callback Dialer and Queue Exit Callback Scheduling**: The automatic call back: which dialer makes it and how far ahead it is scheduled (10, 30 or 60 minutes). > **Note** > Those last two fields show greyed out until you turn on **Exit to callback** in the general block above. They enable themselves the moment you do. ## Extra Functions: CRM, AI and surveys The last block of General is what the agent is going to have at hand while working. ![Extra Functions, at the end of the General tab](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-extras.jpg) *Extra Functions, at the end of the General tab* - **CRM**: Which CRM this campaign works with. It is the one that opens for the agent with the record of whoever is calling, and the one that decides which base the incoming number is searched against. - **AI Agent Copilot**: The AI Agent the agent can talk to while working, to look something up without dropping the call. It has to be **published**: if you point at one that is not yet, the agent sees a notice that the copilot is unavailable instead of an answer. - **Knowledge Base**: The base the copilot of this campaign can consult. - **Post-Call Survey**: The survey-type IVR that plays when a call from this campaign ends. - **Survey by typification**: Limits the survey to the calls that ended with certain typings. It only enables once you have picked a survey above. > **Note** > The copilot no longer has a type: it is always an **AI Agent**, and with that the type selector and the OpenAI assistant disappeared from this screen. A campaign that came configured with an assistant keeps it and goes on working the same, but the moment you pick an AI Agent here it is left with the new copilot, and from this screen there is no going back. > **Important** > The AI integration is no longer picked here. The copilot brings its own, the one configured on the AI Agent you point at, and the one that analyses the calls is now picked in the **Speech Analytics** tab, next to the models that use it. ## Script: what the agent says The **Script** tab is a rich text editor where you write what the agent has to say. What you put here shows in their panel when a call from this campaign comes in. ![The script, with its formatting bar and the available variable underneath](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-script.jpg) *The script, with its formatting bar and the available variable underneath* You have headings, bold, italics, underline, lists, quotes, code blocks and text colour. One variable can be used, **{{agentname}}**, which is replaced by the name of the agent who is handling the call. The screen reminds you of it under the editor. > **In practice** > The script gets read out of the corner of the eye while talking, so what works is short sentences, in the order they are said, with whatever is mandatory in bold. Long paragraphs do not get read. And it is the best place for the legal recording notice, because it stays in sight on every call without depending on the agent learning it by heart. ## Announcements: what the customer hears while waiting The **Announcements** tab has two independent blocks that play over the music on hold. ![Periodic announcements at the top, informative ones below](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-anuncios.jpg) *Periodic announcements at the top, informative ones below* ### Periodic Announcements A message of yours that repeats every so often: the promotion of the month, the opening hours, the reminder that the paperwork can be done on the web. - **Repetition Time**: Every how many seconds it plays. - **Audio**: Which one plays. You can put **several separated by commas**: they play in order and when they finish it goes back to the first one. > **Note** > The **Audio** field starts locked. It unlocks the moment you write a repetition time and, when you do, Audara puts the **queue-periodic-announce** audio in there by itself. It is a starting value: change it for yours. ### Informative Announcements The second block is different: it is not a message of yours, it is Audara telling the customer **what position they are in and how much longer**. It is built by gluing little pieces of audio together, and that is why there are eight files and not one. ![With the frequency set, the selectors and the eight audios enable](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-anuncios-2.jpg) *With the frequency set, the selectors and the eight audios enable* - **Announcement Frequency**: Every how many seconds the position and the estimated time are announced. **This is the field in charge**: while it is empty, the whole block below stays greyed out. - **Minimum Announcement Frequency**: The minimum time that has to pass between one announcement and the next. - **Hold Time**: Whether the approximate wait time is included or not. - **Position in Queue**: Whether they are told what position they are in or not. - **Round Time**: What multiple the announced time is rounded to. The valid values are 0, 5, 10, 20 and 30. The **Default Audios** button fills the eight fields with the audios Audara ships out of the box, and it is the fastest way to leave the campaign working. ![The eight pieces the sentence is built from](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-audios-defecto.jpg) *The eight pieces the sentence is built from* Each field is a piece of the final sentence. The names are in English because they are the names of the audio files, but what they say is this: - **You are next**: "You are now first in line". - **There are**: "There are". - **Calls waiting**: "Calls waiting". - **Hold time**: "The approximate wait time is". - **Minutes and Seconds**: "Minutes" and "seconds". - **Thank you**: "Thank you for your patience". - **Report hold**: "Wait time". > **In practice** > If you are going to record the announcements in your own voice, record them as **eight separate files** and respect what each one says. Audara chains them in real time with the number in the middle, so a file that says the whole sentence sounds wrong. And record them all with the same voice and the same volume, which is what shows the most when it does not match. ## Typing: what the agent records Typing a call is leaving a record of what it was about. The **Typing** tab builds the form the agent gets when hanging up. ![The catalogue of tags at the top, the four typings below](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-tipificacion.jpg) *The catalogue of tags at the top, the four typings below* It is built in two stages: first you create the **tags**, which are the vocabulary, and then you spread them over up to **four typings**, which are the questions the agent is asked. ### The tags The **+** button in the top block creates a new tag. The list is shared by the whole account, so a tag you create here can be reused in another campaign. ![A new tag: name, abbreviation and colour](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-etiqueta.jpg) *A new tag: name, abbreviation and colour* - **Name***: What the agent is going to read. Three characters minimum. - **Description**: A clarification of when it is used. It is informative. - **Abbreviation***: Two or three letters. It is what shows inside the coloured circle in the lists and the reports. - **Color***: The colour of the circle. > **Note** > The Save button of this window does not turn on until you **press the little Colour square and pick one**, even though it already shows a colour. Until you do, the preview at the side does not appear either. The three-dot menu of each row lets you edit it or delete it. ### The four typings Below are the four boxes the agent can fill in. Each one is a switch: turned off it does not exist for the agent. ![A typing turned on: name, tags and whether it is required](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-tipificacion-1.jpg) *A typing turned on: name, tags and whether it is required* - **Name***: The title of the question exactly as the agent sees it. Three characters minimum. - **Tags***: The options they are going to choose from. The agent picks **one only** per typing. - **Required**: Forces them to fill it in before becoming available again. It only applies if the group has ACW without a time limit. The fourth one is the **open typing** and it is different: it carries no tags, it is a free text field for the agent to write a comment. > **Important** > This tab is the one that turns **Save** off in silence. A typing that is on with a name that is too short, with no tags, or with the same name as another one that is on, leaves the button dead with no message at all. If Save does not respond, start by looking here. ### Typings that depend on another From the second one onwards **Depends on** appears, which chains one question to the previous one. ![A typing that only appears depending on what was picked in another](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-dependencia.jpg) *A typing that only appears depending on what was picked in another* - **Depends on**: Which of the previous ones it depends on. With **-- No --** it is independent and always shows. - **Dependent typification tags**: Which answers of the parent it appears with. If you leave it empty, it appears with any of them. > **In practice** > Chaining is what turns four loose lists into a form that makes sense. The typical one is **Reason** in the first and the detail in the second, hanging only off the reasons that need detail. The agent sees two fields instead of a list of forty options, and the report comes out with the hierarchy already in it. ### Conversion At the end of the tab you mark which answer counts as a sale or an achievement. ![The conversion is picked from the tags of the active typings](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-conversion.jpg) *The conversion is picked from the tags of the active typings* - **Conversion**: The tag that counts as a conversion. The dropdown shows them grouped by the typing they belong to. - **Conversion threshold (%)**: The percentage from which the agent sees their indicator in green. Below that they see it in red. - **Daily conversion goal**: How many conversions are expected per day. The agent who gets there wins a medal in their panel. ## Alarms: what warns the supervisor The **Alarms** tab picks which warnings light up for the supervisor in their panel when this campaign goes out of the ordinary. ![The available alarms, each with its comparator and its value](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-alarmas.jpg) *The available alarms, each with its comparator and its value* - **Available Agents**: How many free agents there are in the campaign. - **Calls in Queue**: How many calls are waiting. - **Paused Agents**: How many agents are on a break. - **Registered Agents**: How many agents have an open session in the campaign. Each row is ticked with its checkbox and given a comparator and a number. Ticking the box **fills in by itself** the comparator and a starting value, so almost always you only have to adjust the number. ![An alarm that is ticked appears below, under Selected Alarms](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-alarmas-2.jpg) *An alarm that is ticked appears below, under Selected Alarms* Whatever stays ticked is listed below, under **Selected Alarms**. Those are the only ones that are going to fire in the supervision panel for this campaign. > **Note** > The **Selected Alarms** chips show the internal name of the alarm while the row above shows the translated name. *Hold Calls* is the one for **Calls in Queue**. > **In practice** > An alarm that fires all day stops being an alarm. Before setting the number, look at a week of the campaign report and pick a value that today gets crossed two or three times a day, not every half hour. And turn few of them on: the supervisor watches one screen, not four. ## Priority: which agents answer first The **Priority** tab lets you say that certain agents get the calls of this campaign before the others. ![On a new campaign the tab only shows its requirement](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-prioridad-aviso.jpg) *On a new campaign the tab only shows its requirement* On a campaign you just created there is nothing to configure, and the tab tells you why: **the campaign has to be assigned to two or more agents**. As agents are assigned by group, the order is create the campaign, put it in a group, assign agents to that group and come back. With agents already assigned, the tab lists each one with a number. **The lower the number, the higher the priority**, and **--** is the highest possible priority. The higher the number, the later the calls reach them. > **In practice** > It is the clean way to build a second level: the experts on the high priority and the rest of the team one step below, only getting calls when the first ones are busy. Careful not to leave a single person on the high priority, because they end up handling the whole campaign. ## Integration: opening a URL for the agent The **Integration** tab builds a web address that opens for the agent with the data of the call, to connect the campaign with an outside system without developing anything. ![The base URL at the top and the call variables below](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-integracion.jpg) *The base URL at the top and the call variables below* - **Auto Open**: Turned on, the URL opens by itself in a new tab when the agent answers. Turned off, the agent opens it by hand from their panel. - **URL**: The base address, with no parameters. - **Assign variables**: What each piece of call data is called on the side of the other system. There are six variables available: **{UNIQUEID}** (the identifier of the call), **{QUEUEUNIQUEID}** (the one for its pass through the queue), **{CALLERID}** (the number of whoever is calling), **{CAMPAIGN}** (the campaign), **{AGENT}** (the agent who answered) and **{WSRESULT}** (the result of an IVR WebService, if there was one). Next to each one you write the name of the parameter your system expects. The ones you leave empty are not sent. The last row, the one with **Value** and **Parameter**, is for a fixed piece of data of yours that does not come from the call. ![The Final URL builds itself as you type](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-integracion-2.jpg) *The Final URL builds itself as you type* Below, **Final URL** shows you the complete address exactly as it is going to open. It is the way to check it came out right without having to wait for the next call. ## Speech Analytics: analysing the recorded calls The **Speech Analytics** tab starts with a single switch. Turned off, there is nothing else to see. ![Turned off, the tab is a switch and nothing more](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-speech-off.jpg) *Turned off, the tab is a switch and nothing more* That switch does something before any analysis: it **keeps the recordings separated by channel**, one track for the agent and another for the customer. Without that there is no way to know who said what, and neither the transcript nor the analysis nor the emotions are worth anything. ![Turned on, the first thing it asks for is who is going to transcribe and analyse](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-speech.jpg) *Turned on, the first thing it asks for is who is going to transcribe and analyse* ### AI integration (LLM) These three fields decide who does the work, and they are picked together because they condition each other. - **AI integration**: The integration that transcribes and analyses the calls of this campaign. It can be **ChatGPT** or **Gemini**, and it is what decides which models you see in the two fields below. If you leave it empty nothing is transcribed and nothing is analysed. - **LLM model for transcription**: The model that turns the audio into text. - **LLM model for analysis**: The model that reads that transcript and answers the QA rules, the emotion and the summary. It is not the same list as the one above: transcribing and analysing are different capabilities, even if the provider sometimes gives them the same name. > **Important** > The two model lists are cut down to the ones of the provider you picked above. If you change integration, the two models are cleared and you have to pick them again: it is on purpose, because a model from one provider does not run on another one's key. And while you have not picked an integration, a yellow strip warns you that nothing with AI in it is going to work. The two model fields carry a **Default** option, which means "whichever the integration has". It is useful when you run many campaigns against the same integration: you change the model once, over there, and not campaign by campaign. ### Automatic transcription - **Transcribe automatically**: Whether the calls of this campaign transcribe themselves when they end. - **Automatic transcription percentage**: What share of the calls gets transcribed. Transcribing everything costs money; a sample is usually enough for quality. - **Target typification**: Limits the analysis to the calls that ended with certain typings. **Empty means all of them**, not none. ### Normalization ![Normalization cleans the text before analysing it](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-normalizacion.jpg) *Normalization cleans the text before analysing it* - **Normalize Transcript**: Runs an automatic correction over the text: it unifies numbers, dates and abbreviations, and leaves it readable. - **Normalization Context**: What your operation is about: the industry, the type of service, the tone. With that the model disambiguates better. - **Glossary**: Your own terms, acronyms and product names. Whatever you put here is respected exactly as written. > **In practice** > The glossary is what pays off most in this block, and almost nobody fills it in. Product names, internal acronyms and codes are exactly what the transcript gets wrong, and they are exactly the words the QA rules go looking for afterwards. Five minutes here are worth more than changing model. ### QA Rules A QA rule is a question asked of every analysed call, and the automatic score of the agent comes out of the answers. ![The catalogue of rules. The icon says which type each one is](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-qa.jpg) *The catalogue of rules. The icon says which type each one is* The **+** button creates a rule, and the first thing it asks is what type it is going to be. ![The two types of rule: look for words or ask the AI](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-qa-tipo.jpg) *The two types of rule: look for words or ask the AI* - **Matches, the chain**: Looks for literal words in the transcript. It is exact, cheap and does not depend on any model. - **Analysis, the magnifying glass**: Asks the AI a question about the call. It understands the meaning, not the words. > **Note** > The icons read backwards from what you would expect: **the magnifying glass is the AI rule** and **the chain is the word one**. It is the same in the list and in the creation window. ![A matches rule: the words and who they are looked for in](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-qa-coincidencias.jpg) *A matches rule: the words and who they are looked for in* ![An analysis rule: a question instead of words](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-qa-analisis.jpg) *An analysis rule: a question instead of words* - **Name***: The one it shows under in the quality reports. - **Matches***: Only on the chain ones. The words to look for, one by one. It is enough for any of them to appear. - **Rule Prompt***: Only on the magnifying glass ones. The question the AI answers about each analysis, written so it is answered with yes or no. Do not confuse it with the **AI Analysis Prompt** at the end of the tab: that one asks for a summary of the whole call, this one resolves a single rule. - **Analyze on**: Only on the chain ones: whether the words are looked for in what the **Agent** said, the **Customer** or **Both**. - **Adds towards score**: Whether this rule adds or subtracts in the total mark of the agent. Turned off, the rule is evaluated all the same but only stays as information. - **Invert positivity**: For the rules where a "yes" is bad. Without this, "the agent was rude" would be painted blue when it is met. - **Positive result tags and Negative result tags**: Typings that get put on the call by themselves according to the result of the rule. The rules you create stay in the catalogue, and below, under **Selected rules**, you pick which ones apply to this campaign. > **In practice** > For whatever is verifiable word by word, like saying the name of the company or reading the recording notice, use **Matches**: it is exact and it costs nothing. Leave **Analysis** for what cannot be written as a list of words, like whether the agent was kind or whether they understood the problem. A prompt that is answered with yes or no works much better than an open one. ### AI summary and emotions ![The summary of the call and the emotion analysis close the tab](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-speech-ia.jpg) *The summary of the call and the emotion analysis close the tab* - **AI Analysis Prompt**: What you want the AI to tell you about each call: a summary, the commitments that were made, the real reason. It is saved with the call. - **Word limit per response**: How long that answer can be. Out of the box, 100. - **Emotion Analysis**: Classifies the call as Positive, Neutral or Negative. - **Emotion analysis tolerance**: **Low** tends to classify as neutral or positive, **High** demands clear signals before going to an extreme and **Normal** sits in the middle. ## Real Time ASR: transcribing while people talk What the previous tab does happens once the call has ended. **Real Time ASR** is the same thing but live: the call is transcribed as it happens, so the supervisor can read it without having to go in and listen. ![The live transcription and its emotion analysis](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-asr.jpg) *The live transcription and its emotion analysis* - **Real Time ASR**: Turns on the real-time transcription for this campaign. - **ASR Model**: Which integration does the transcribing. If the one you pick is disabled, the screen tells you. - **Real-time Emotion Analysis**: Measures the tone of the conversation as it goes along. If you also have Speech Analytics on, one last analysis of the whole thing is done when the call ends. - **Emotion analysis frequency (sec)**: How often it is measured again. Out of the box, 60. - **Emotion analysis tolerance**: The same scale as the other tab: Low, Normal or High. > **Important** > The two transcriptions do not add up: **with Real Time ASR turned on, the automatic transcription configuration of Speech Analytics is ignored**. The Speech Analytics tab warns you. It makes sense, because the call was already transcribed live, but it is worth knowing before spending half an hour adjusting a percentage that is not going to be used. ## On-site The **On-site** switch at the start of General creates something else: the record of a walk-in service, with no telephone and no chat. The agent starts it from their panel and records the typing and the CRM contact just as if it had been a call. ![With the switch on, the form shrinks to three tabs](https://docs.audara.io/colas-inbound/imagenes/colas-inbound-presencial.jpg) *With the switch on, the form shrinks to three tabs* The title changes to **New On-site** and the form is left with three tabs: **General**, **Script** and **Typing**. Everything to do with telephony disappears, because here there is no queue and no ringing and no music on hold, and so do the predictive dialer switch and the client name. What is left in General is the **Name** and an **Extra Functions** block cut down to three fields: CRM, AI Agent Copilot and knowledge base. The surveys are not there, because a post-call survey plays over the phone and here there is no call. Script and Typing work exactly the same as in a campaign. > **Note** > There is no separate button to create on-site service: it is created from the same **+** of the campaign list, turning the switch on. In the list it appears mixed in with the campaigns, with the handshake icon. ## Voice Groups A **group** puts several campaigns together so they can be assigned in one go. Agents and supervisors are not assigned to loose campaigns: they are assigned to groups, and the permissions to see reports come from there too. That is why every campaign somebody is going to answer has to be in a group. Groups live in **Setup > Voice Groups**, and they have their own article: [Inbound campaign groups](https://docs.audara.io/en/grupos-inbound/). That one has the form field by field, how agents and supervisors get assigned, and the locks on renaming and deleting a group.