> ## Documentation Index > Fetch the complete documentation index at: https://docs.voxworks.ai/llms.txt > Use this file to discover all available pages before exploring further. # Survey Variables and Piping > Carry values through a survey: the Variables card, the Set variable action, piping earlier answers, variables and contact fields into prompts, the readiness checks on piped tokens, and every custom variable a survey call writes. ## Three kinds of value A survey can speak three kinds of value inside its prompts, and writes a set of custom variables to every call: | Token | What it is | Example | | ---------------------------------- | ------------------------------------------------------------------------------------- | ------------------------------------- | | `{{survey..result}}` | The recorded answer to an earlier question. | `{{survey.q3.result}}` → *Labor* | | `{{survey.var.}}` | A **survey variable** — a value the interview carries, set by a rule. | `{{survey.var.first_pref}}` → *Labor* | | `{{contact.}}` | A field from the contact being called: `first_name`, `last_name`, `phone` or `email`. | `{{contact.first_name}}` → *Sam* | All of them are spoken as text. *** ## Piping into prompts In a question's **Voice prompt**, type `/` to open the picker: *"Type / to insert an earlier answer, a variable or a contact field."* It offers the survey's variables, every earlier question's result, and the four contact fields, and inserts the token for you. For example, a follow-up that refers back to a first preference: > You said `{{survey.q7.result}}` was your first choice. Which would be your second? Give any question you pipe a lower-case key made of letters, numbers and underscores — for example `q7` or `first_pref` rather than `Q7`. On the call, results are stored under the lower-case form of the key, so a lower-case key keeps the token and the stored answer identical. Tokens work in the **Voice prompt**, in **Alternative phrasings**, and in battery items' longer forms. ### What a piped answer sounds like `{{survey..result}}` speaks the recorded answer: | Question type | Spoken as | Example | | --------------------- | ---------------------------------- | ------------------------------ | | Single choice, Rating | The chosen option's label | *Very satisfied* | | Number | The digits | *3941* | | Long text | The respondent's words | *the parking near the station* | | First name | The name | *Sam* | | Phone number | The number in international format | *+61412345678* | A question that was skipped, not reached, or asked later has no answer yet, so the token speaks as nothing — a sentence with a hole in it. Anything that is not one of the three forms above is not substituted at all and is **spoken out loud as the raw token**. [Readiness](/surveys/readiness) checks for both. ### Readiness checks on piped tokens Each warning names the question and where the token is — its **prompt**, **alternative phrasing (n)** or **item “(key)” longer form (n)** — then one of: | Warning | Fix | | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | *"pipes `{{…}}`, and “x” is not a variable of this survey — the token speaks as an empty string."* | Add the variable on the **Variables** card, or correct the key. | | *"pipes `{{…}}`, and no rule anchored on an earlier question sets “x” — the token speaks as an empty string."* | Add a **Set variable** rule on a question asked before this one. | | *"pipes `{{…}}`, but every rule that sets “x” is anchored on a question that can be skipped — the token speaks as an empty string on a call that never asked it."* | Set the variable somewhere every respondent passes, or reword the prompt so an empty value still makes sense. | | *"pipes `{{…}}`, which is none of `{{survey..result}}`, `{{survey.var.}}` or `{{contact.}}` — nothing substitutes it, so the raw token is spoken."* | Use one of the three supported forms. | | *"pipes `{{…}}`, and “x” is not a question of this survey — the token speaks as an empty string."* | Correct the question key. | | *"pipes `{{…}}`, which is this question itself — the token speaks as an empty string."* / *"…which is asked later — …"* | Pipe only questions asked before this one. | | *"pipes `{{…}}`, but “x” can be skipped — the token speaks as an empty string on a call that never asked it."* | The piped question is hidden or jumped over for some respondents. Reword the prompt so it still makes sense with an empty value, or pipe a question every respondent is asked. | A contact field that is empty on the contact record also speaks as nothing, so write prompts that still read naturally without it. *** ## Survey variables A survey variable holds a value that is not simply one question's answer — a first preference carried into several later prompts, a region derived from a postcode, or a flag set by a screener. ### The Variables card On the [Logic tab](/surveys/logic), with nothing selected, the **Variables** card reads: *"A value the interview carries rather than one question — set it from a rule with “Set variable”, and speak it in a later prompt with `{{survey.var.}}`."* Click **Add variable**: | Field | Rules | | ------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | Key | *"Variable keys start with a letter and use letters, numbers or underscores."* Up to 64 characters. Must be unique: *"This survey already has a variable called (key)."* | | Label | *What it holds, in plain words.* | | **Data type** | text, boolean, integer, number, date, time, datetime, choice or json. Default text. | Every survey variable is stored and spoken as text on a call, whatever its **Data type** — the data type describes the value for your own records. The confirmations are *Variable added*, *Variable updated* and *Variable deleted*. A variable no rule writes is tagged *no rule sets it*. ### Setting a variable Use a [routing rule](/surveys/routing-rules) with the **Set variable** action. It writes either: * **The answer to this question** — the anchor question's recorded answer, or * **A fixed piece of text** — whatever you type (*The text this rule stores in the variable*). Every variable starts each call empty. If several rules set the same variable, the one that runs last wins. A rule with a condition writes its value when the condition holds and **clears the variable** when it does not, so two conditional rules on the same variable do not combine — the later one decides the value on every call. Give each conditional value its own variable. A variable that a rule still sets cannot be deleted: *"(rules) still sets this variable. Change or delete that rule first."* **Example — carrying a first preference.** A rule on `q7` (first preference), trigger **When an answer is submitted**, no condition → **Set variable** `first_pref` to *The answer to this question*. A later prompt can then say *"Apart from `{{survey.var.first_pref}}`, which would you choose?"*. *** ## Custom variables a survey call writes Every survey call writes its results to the call's custom variables under `survey.*`. All values are text unless noted. ### Per question `` is the question key in lower case, with anything other than letters and numbers turned into underscores — a question keyed `Q3` writes `survey.q3.result`. | Variable | Use | Type | Example | | ------------------------------------ | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------ | ------------------------------------------------------------------------- | | `survey..result` | The recorded answer: the option label, digits, the respondent's words, a name or a phone number. Empty if none. For a battery, a JSON map of item key to rating, where `NA` means the item was excluded and never asked and an empty value means it was asked but not rated. | Text (JSON text for a battery) | `Very satisfied`; `3941`; `{"your_own_life":"Positive","education":"NA"}` | | `survey..status` | How the question ended: `captured`, `skipped`, `invalid` or `abandoned` (busy or stop during the question). Empty if never reached. | Text | `captured` | | `survey..termination_requested` | Whether this answer ended the interview. | Text, `true` or `false` | `false` | | `survey..screened` | Number questions only: the answer was recorded but failed the question's **Screen out** list. | Text, `true` or `false` | `true` | | `survey..object` | The full question record — the answer plus details such as whether it was confirmed or spelled. | JSON | `{…}` | ### Per call | Variable | Use | Type | Example | | ------------------------------------------------------- | -------------------------------------------------------------- | ----------------------- | -------------------------- | | `survey.temp.outcome` | The outcome code the call ended with. Starts as `in_progress`. | Text | `completed` | | `survey.temp.termination_reason` | A readable reason when the call ended early. | Text | `Declined at the greeting` | | `survey.var.` | Each survey variable's final value. | Text | `NSW` | | `survey.temp.call.is_mobile`, `.is_inbound`, `.is_test` | The call facts rules can read. | Text, `true` or `false` | `true` | ### Gatekeeper variables Written by the [Gatekeeper](/surveys/gatekeeper) on every call: | Variable | Use | Type | Example | | --------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---- | -------------------- | | `survey.gatekeeper.result` | What happened at the front door: `agree`, `decline`, `not_now`, `wrong_person`, `voicemail`, `screener_no_connect` or `no_response`. | Text | `agree` | | `survey.gatekeeper.status` | The detailed disposition, for example `connected`, `declined`, `callback`, `wrong_person`, or a voicemail status such as `voicemail_sms` (see [Voicemail and SMS](/surveys/voicemail#how-voicemail-is-recorded)). | Text | `voicemail_sms` | | `survey.gatekeeper.reason` | The disposition class. | Text | `callback` | | `survey.gatekeeper.answerer` | Who answered: `unknown`, `target`, `screener` or `machine`. | Text | `target` | | `survey.gatekeeper.answerer_name` | A name the person who answered gave for themselves. | Text | `Sam` | | `survey.gatekeeper.entry_class` | How the pickup sounded: `live`, `named`, `unclear`, `screener`, `machine` or `silent`. Empty on inbound calls. | Text | `live` | | `survey.gatekeeper.callback_time` | A callback time in the respondent's own words. | Text | `after six tomorrow` | | `survey.gatekeeper.callback_slot` | The same time trimmed to a day and time of day. | Text | `tomorrow morning` | The gatekeeper's own result is not shown in Results or in the CSV export; calls that end at the gatekeeper appear in the **Contact funnel** (see [Results](/surveys/results)). ### Contact fields | Variable | Use | Type | Example | | -------------------- | ------------------------------------------------------------------------------------------------------------ | ---- | -------------- | | `contact.first_name` | Written by a [First name](/surveys/question-types/first-name) question when it captures a name. | Text | `Sam` | | `contact.phone` | Written by a [Phone number](/surveys/question-types/phone-number) question, in international (E.164) format. | Text | `+61412345678` | These are written to the call. To update the contact record in [Contacts](/contacts/overview), use the recorded answers from Results or the export.