Acquainto
← All guides

Guide

The conversation — follow-ups, branches, and the check at the end

Your questions are the skeleton. What a respondent actually gets asked depends on what they say, and four things shape it: follow-ups, branches, whether a question is required, and the check at the end.

Follow-ups on a vague answer

After each answer Acquainto reads it against the question and everything already said, and asks one clarifying question if the answer is too vague to be useful, or contradicts something earlier.

A follow-up may only narrow the question that was already asked. It reuses that question’s own wording and the respondent’s own words. It may not introduce a category, an example, or a sub-topic your question did not already name — if it cannot narrow without inventing something, it accepts the answer instead. An intake that asks for things nobody configured is worse than one that takes a vague answer.

One per question. If a question has already been clarified once and the reply still reads as vague, the answer stands. A question that needs a third phrasing is usually a question that needs rewriting.

A follow-up does not add a question. It lands as more detail on the same question, and the field it arrives under in your data is the same one. A “Follow-up” note marks it in the conversation, so the respondent knows why they’re seeing the same question again.

Before: a concrete answer to "What would make onboarding easier for your team?" and the conversation moves on to the next question. After: the vague answer "Better onboarding, I guess." to the same question, marked with a "Follow-up" note and answered with "When you say better onboarding — what would make it easier for your team?" — the same question narrowed, not a new one.

The exception: a question that asks for something specific

Six of the question types ask for an answer in a particular shape: email address, phone number, date, money amount, number and web address. Respondents answer them by typing, exactly as they would an open question — there is no special box and no picker. What changes is what happens to the answer.

The check is a plain format check rather than a judgment, so it is instant and costs nothing. It is also more forgiving than it looks: “sure, it’s ana@example.com” is a valid answer to an email question, and a budget question takes 1500 as readily as $1,500.

Then the answer is tidied into one shape before it reaches you. 3 March 2026 and 3/3/26 both arrive as 2026-03-03. $1,500.50 arrives as the number 1500.5, with the currency alongside it, so a column of them adds up. acme.com arrives as https://acme.com.

Two things it will not do:

On a required question, an answer it cannot read is asked for again — every time, with no limit. On an optional question, it is accepted as written: somebody answering “Your email, if you’d like updates” with “I’d rather not” has answered, and asking again would be badgering them for something you said was optional.

What a choice option stores

You type the wording respondents read. We store a value alongside it, derived from that wording in lowercase — “Blocking my work” is stored as blocking my work. That stored value is what your webhook, API and exports carry, and what a branch matches on.

The wording is translated for a respondent reading in another language. The stored value never is — it is the same key in every language, so a branch and everything you have built downstream behave identically whoever answers.

It is fixed the moment the option is created. Rewording the option later changes what respondents read and not what you receive, on purpose: a branch, and whatever you have already built downstream, are both keyed to the stored value, and silently moving it under them is the kind of break that shows up as a branch that quietly stops firing rather than as an error.

If your own system needs a particular key — a CRM that expects P1, a pipeline keyed on codes you do not control — tick Set my own stored value for each option on the question. Every option then needs a value, and they have to be different from each other.

After a tap, respondents read your wording exactly

Most questions are reworded slightly as they are asked, so each one refers back to what the respondent just said and the intake reads as a conversation rather than a list. A question that follows a tapped option is not. The respondent gets your wording, character for character, and the same is true after a skipped question.

There is nothing to gain from rewording there — the only thing available to refer back to is the label they just tapped, which is still on their screen — and dropping it takes seconds off those turns. The practical consequence for you: on an eflow that is mostly options, what you type in the question box is what respondents read. Write it to stand on its own.

If the respondent is reading in another language, “your wording exactly” means a translation of your wording — the same question, in their language, rather than a reworded one. We translate an eflow’s questions and option labels once per language, keep the result, and reuse it for everyone after that. Edit a question and only that question is translated again.

Branches

Branches come off a single-select question, and only after you tick Allow branching from this question’s answer on it. Each option then gets an Add branch for ”…” control: a branch is a set of questions asked only of respondents who chose that option. Branches can nest.

No other question type can open a branch — not multi-select, not a rating scale, and not any of the typed or open text questions.

Someone who did not take a branch never sees its questions, and never sees a gap where they would have been.

This matters for your data: a branch question’s field is absent from the responses of everyone who went the other way. That is normal, not an error. The Response fields table on the Integrations tab marks each of these with the path that reaches it — Only asked after: ….

Two finished conversations from one eflow. The first chose "Yes, something went wrong", was told "Switching to: Setup problems", was asked what went wrong, and has three progress segments. The second chose "No, it went fine" and went straight to the next question, with two segments and no gap where the branch would have been.

Required, and skipping

Each question has a Required checkbox in the editor.

Before, with Required unticked: a "Skip question" link sits in the footer, to the left of Privacy and divided from it by a thin rule. After, with Required ticked: the same question and the same footer, with Privacy alone and no rule before it.

The check before it sends

Every conversation ends with a review screen, headed Quick check before it goes. It lists every answer with a Change control beside each one, and the line under it says: Change anything that looks wrong — sent to [your organization] only when you confirm.

The conversation itself collapses behind a Show conversation link at the top of it, so the list starts at the top of the card and the send button is in view without scrolling. Nothing is hidden — one press puts the whole transcript back.

Five things about it are worth knowing:

Two states of one review screen, headed "Quick check before it goes", with the conversation collapsed above it behind a "Show conversation - 4 answers" link and the line "Change anything that looks wrong — sent to Northwind Traders only when you confirm." below the heading. At rest: four answers — a name tagged "Edited", a single choice reading "Product quality", a multiple choice reading "Returns portal, Order tracking", and a rating reading 6 — each with a Change control carrying a pencil icon, above a "Looks right, send it" button and the line "Answers can't be changed after you send them." With Change pressed on the rating: that row has become the 0-10 rating row it was answered on, 6 still selected, "Not at all likely" under 0 and "Extremely likely" under 10, with Save and Cancel below it. Every other row is unchanged.

Using Acquainto already? The same guides are in the product under Help & support, with the ones that only make sense signed in.