WhatsApp Surveys

Fly runs surveys over WhatsApp as well as Messenger. The survey itself is the same survey — same Typeform form, same shortcode, same export. What is different is how a respondent gets into it, and what you are allowed to send them afterwards.

Read Channels first for the question-by-question differences. This page is about running a study on WhatsApp: getting a number, getting people into the survey, and staying inside WhatsApp's messaging rules.

Getting a number

A WhatsApp survey runs on a WhatsApp Business number connected to your Fly account. Unlike a Facebook page, you cannot currently connect one yourself: tell us the study and we will set the number up and attach it to your account.

Two consequences worth knowing before you plan a study:

• One number belongs to one Fly account. Every survey you create is reachable on your number, and no two researchers can share one. A researcher with many Facebook pages still needs only one WhatsApp number.
• A study cannot change channel once recruitment starts. The channel an ad opens is fixed when its ad set is created and cannot be changed afterwards, so decide between Messenger and WhatsApp before you launch.

The Connect → WhatsApp Business Account button on the Accounts screen is a preview of self-service connection and is not the supported route yet.

How someone starts a WhatsApp survey

On Messenger, a link carries the survey's shortcode invisibly. WhatsApp has no equivalent. The only thing that can carry a shortcode into a WhatsApp conversation is the text of the respondent's own first message.

Everything below is a way of getting the right text into that message.

The entry reference

The text Fly is looking for is:

form.<shortcode>

optionally followed by extra key/value pairs, exactly as on Messenger:

form.hpvintro.creative.3b.gender.men

which reaches your survey as the hidden fields creative and gender.

Rules, all of which matter:

• It must be the entire message. form.hpvintro starts the survey; hi, form.hpvintro does not. A leading start is allowed (start form.hpvintro). Surrounding spaces are fine.
• Only letters, digits, _ and - in the shortcode and in each key and value. A space or a / will not be recognised.
• Case is preserved for the shortcode and every value; only the word form itself is case-insensitive.
• Being the whole message is what stops a respondent's own mid-survey answer from accidentally restarting the survey. It is deliberately strict.

Ads created by Virtual Lab may instead carry a short opaque reference that looks like r.AQhtbmNod2Vla6QffA6T — this is the same thing in a compact form and is handled automatically. See what an ad's ref carries.

1. A click-to-WhatsApp ad

This is the production path, and the one place a WhatsApp study usually goes wrong.

When someone clicks a click-to-WhatsApp ad, WhatsApp opens with a message already typed into their compose box. That prefilled text comes from the ad, and it is the only carrier of the shortcode — the referral information WhatsApp attaches to the click contains nothing that names your survey.

Two more things about ad clicks:

• The respondent can edit the prefilled text before sending it, and some do. A study will always have a few arrivals that lost their reference this way.
• Click-to-WhatsApp works on mobile only. Someone clicking your ad on WhatsApp Web or Desktop arrives without the ad context.

2. A wa.me link

For a link you can put in an email, a poster or a WhatsApp message:

https://wa.me/<number>?text=form.<shortcode>

<number> is your WhatsApp number in international form with no +, spaces or dashes. Tapping the link opens a chat with the message prefilled; the respondent presses send.

• Only text= survives. Any other query parameter you add — ref=, id=, anything — is silently dropped before it reaches us. There is nowhere else to put a reference.
• Percent-encode & and # if a value ever contains one, or the message is silently cut short at that character.
• On a desktop browser the link shows an interstitial page and needs a second click.
• The respondent sees and can edit the text. If you are carrying targeting values in the reference, they can read them.

A QR code that opens a chat is the same mechanism with a different wrapper, and carries no extra information.

3. Typing it

Anyone can start a survey by sending form.<shortcode> to the number by hand. This is how you test, and it is why shortcodes must stay typeable — a hand-typed space is a literal space and will not match.

What a WhatsApp number does not do

A message to the number that is not an entry reference and does not belong to a conversation in progress gets no reply at all. Someone who writes "hello" to the number is ignored.

That is deliberate — a WhatsApp business number is not a broadcast channel — but it does mean a respondent who mistypes the reference gets silence rather than help. Expect a small number of arrivals to be lost this way, and make the entry reference easy to send correctly rather than easy to type from memory.

Keeping out people who did not click your ad

This is structural, not a bug, and it is specific to WhatsApp.

On Messenger the entry reference is invisible: it travels inside the link or the ad, and a respondent never sees a string they could pass on. On WhatsApp the entry reference is the respondent's own message. It sits in their compose box before they press send, where they can read it, screenshot it and forward it.

Four properties combine badly on a survey that pays:

• Possession is the whole check. Fly starts a survey for anyone who sends a valid reference. There is no second test.
• A forwarded arrival is invisible to you. It answers the same questions and reaches the same ending, so nothing in your responses, your completion count or the Monitor tab distinguishes it from a respondent you paid to recruit.
• Your budget limits who sees the ad; nothing limits who receives a forward. Ad delivery is metered and costs money per click. Forwarding a message to a group is free, instant, and repeats without bound.
• The respondent learns it pays only after they have the reference. The incentive is announced inside the survey — by which point the text that starts it is already sitting in their message history, ready to share.

So on a paid WhatsApp study, treat the entry reference as public. What separates a real respondent from a forwarded one is not the reference — it is ad_id.

Why ad_id is the right test

ad_id is set by Fly from what the messaging platform reports about the click. It cannot be set from a link or a message: if a reference contains ad_id.something, that value is discarded rather than used. See Hidden Fields. Somebody who was forwarded your reference has no way to manufacture one.

On WhatsApp it is also reliably present — a click-to-WhatsApp referral arrives attached to the first message of the conversation, so an ad click carries an ad_id and a forwarded reference does not.

Example 1 — the simple gate

Add a logic jump on your first question, so nobody reaches an incentive question, and point it at an "ineligible" ending you have already written:

On consent_1: if ad_id is empty → jump to the ending ty_ineligible.

That is the whole rule. Two things about it are worth knowing before you rely on it:

• There is no is_empty operator. The operators available are is / equal, is_not / not_equal, contains, not_contains, the four numeric comparisons, and, or and always. Compare ad_id against an empty value with is — a respondent who arrived without an ad has ad_id set to an empty string, so this matches.
• The jump fires after the question is answered, as all logic jumps do. A blocked respondent sees your first question and answers it before being sent to the ending. Put the rule on the first question and this costs one tap; put it after the phone-number question and it costs you the incentive.

Example 2 — with an escape hatch for testing

The rule above locks you out too. You cannot test with a wa.me link, because a link carries no ad_id — only a real ad click does.

The fix is a second hidden field that only testers use. Pick any name; testing is the obvious one. Turn away a respondent only when both are empty:

On consent_1: if ad_id is empty and testing is empty → jump to the ending ty_ineligible.

Testers then start the survey with the extra key/value pair on the end:

https://wa.me/<number>?text=form.mytestcode.testing.1

which arrives as the hidden field testing and holds the gate open for that one conversation. Everyone else — including everyone who was forwarded your plain reference — still has both fields empty and is turned away.

Declare both ad_id and testing as hidden fields on the form so they are available to the logic.

Testing a WhatsApp survey

Before an ad exists, and before any money is spent:

1. Create your form and give it a shortcode as usual.
2. From your own phone, open https://wa.me/<number>?text=form.<shortcode>, or just send form.<shortcode> to the number.
3. Answer the survey.

That exercises the whole path — WhatsApp, Fly and back — and is the only way to find the channel-specific failures on the Channels list before your respondents do. Watch the Monitor tab while you do it.

If you added the ad_id gate above, step 2 will send you straight to your ineligible ending — a wa.me link carries no ad_id, which is the entire point of the gate. Add your tester field to the reference instead: https://wa.me/<number>?text=form.<shortcode>.testing.1. See Example 2.

Fly does not let anyone take a survey twice. To test the same form again, use the reset reference your administrator gave you, exactly as on Messenger. See Testing.

To test the ad itself, there is no substitute for clicking a real ad on a real phone and reading the compose box before pressing send. That text is the whole configuration.

The 24-hour rule

WhatsApp only allows a business to send an ordinary message within 24 hours of the respondent's last message. Outside that window, only a pre-approved template will send. There is no exception and no fallback: an ordinary message sent outside the window does not arrive.

Messenger now has the same rule — Meta retired its opt-in alternatives in early 2026 — so this is not a WhatsApp tax you can design around by choosing the other channel. What is different here is that a mistake is silent: on Messenger the failed send shows up as an error in the Monitor tab, on WhatsApp the message simply never arrives.

Everything in a Fly survey that reaches out to a quiet respondent is affected:

So a multi-wave WhatsApp survey looks like this:

[wave 1 questions]
       ↓
[wait — 3 days]
       ↓
[utility_message]  ← an approved template, not a statement
       ↓
[wave 2 questions]  ← the respondent's reply reopens the 24-hour window

The respondent's tap or reply on that template message reopens the window, and the rest of the wave proceeds normally.

Templates

Create templates in the dashboard's Message Templates tab, and use them with a utility_message question. The setup is the same on both channels; what WhatsApp changes:

• Review is real. Messenger usually approves in seconds; a WhatsApp template can take minutes to hours, and can be rejected for wording Messenger accepts. Create and approve your templates before the study starts, not during it.
• Sample values are required for every {{1}}, {{2}} placeholder.
• Buttons stay visible in the chat and their labels are locked at approval — your question's choices must match them exactly, in the same order.
• A template is registered per messaging account. If a study runs on both a Facebook page and a WhatsApp number, register the same template name on both — the survey names a template, not a channel. The Duplicate action in the template list copies one to another account so the shapes stay in agreement.
• Approved templates cannot be edited. Delete and recreate to change wording. A name freed by deleting an approved template stays reserved for 30 days; a rejected one can be reused immediately.

How many messages you can send

Only messages sent outside the 24-hour window count against WhatsApp's quota — that is, template messages. Survey starts and the whole back-and-forth of an active conversation are free.

A new number starts at 250 unique recipients per 24 hours outside the window, rising to 2,000 and beyond as the number builds a delivery history at good quality. In practice:

• A study whose respondents answer promptly consumes almost none of it.
• A study that leans on long waits and follow-up nudges can exhaust 250 per day at a few hundred respondents.

If your design sends a template to every respondent on a schedule, tell us the expected volume before you start, so the number is on a high enough tier.

There is also a hard rate of one message every 6 seconds to the same person, which matters only if a survey tries to send several messages in a burst.

What is not supported on WhatsApp

• Receiving photos, files or voice notes from respondents. See Upload.
• Shared locations and contact cards. They arrive and are ignored.
• Notify questions and one-time notifications. These never existed on WhatsApp, and Meta retired them on Messenger in early 2026. Use a template on both.
• Handoff to another app. WhatsApp has no thread control.
• More than 10 choices on a question, and a link button label longer than 20 characters — both are send failures. See Channels.