01Brief
Four questions before a single survey question.
What decision is this for, what happens on each outcome, who counts as your market, and when it is due. The agent will not build a questionnaire around a vague wish.
Ask your market.Read it like a researcher.
Verbalab plans the study, writes the questionnaire, runs the field and reads the results. You bring the question.
01Method
A study, start to finish.
02Field
The questionnaire answers back.
Branching runs on the server, never in the browser. Screeners cross-check declared answers against what the sample already knows, and inconsistent answers are sent back for correction instead of being recorded.
03Read
Findings, not a spreadsheet.
Answers land in an append-only ledger with their full history. The agent reads the base sizes, the screen-outs and the open ends, and writes what they mean for the decision you named in the brief.
02System
You, the agent, your market.
A study is three parties passing one signal around. Findings feed the next brief — that is what makes it a loop, not a pipeline.
You
Bring the decision. The brief is four questions.
brief ↓
The agent
Designs the questionnaire, runs the field, reads the results.
questionnaire ↓ answers ↑
Your market
Real people answering one question per screen.
record · read · findings ↺
03Findings
Clean data is the product.
Contradictions are prevented, not deleted.
What the sample already knows about a respondent gates what they can be asked and what they can claim. Bad rows never get written.
Lists that follow the answer.
Name two reasons and the next screen asks about the ones you did not name. Option lists derive from earlier answers on the server, and the removed options are refused on the wire too.
Inconsistent answers get a second ask.
A total that does not match its parts comes back to the screen with a sentence, not into the data.
Built for tracking.
Questions carry a semantic identity across versions, so wave two is comparable with wave one by construction.
Respondents finish what they start.
Close the tab, come back, land on the same question with your answers intact. Plain language, one question per screen.
04Engine
The engine underneath.
Two decades of field-tested survey methodology, rebuilt as software. These are the guarantees the data rests on.
- 01append-only ledger
- Answers are never updated in place. A change of mind supersedes; the history stays readable.
- 02server-side branching
- Skip and show logic runs on the server. A client cannot answer a question it was never asked.
- 03signed sample context
- What your CRM knows about a respondent arrives cryptographically signed, or not at all.
- 04publish gate
- A questionnaire with an unresolvable rule, a dangling list or a cycle does not go to field.
- 05screen verdicts
- Every screen-out carries a disposition code. Incidence is a query, not an estimate.
- 06quality flags
- Speeders, flatliners and contradictions are flagged for review. The engine marks; it does not silently delete.
05Start