Open-end survey responses, coded into your categories — not silently guessed.
Send your codebook and your open-end responses (CSV/XLSX). Get back every response coded into your categories, with anything ambiguous flagged for your review instead of forced into a category it doesn't clearly fit.
Delivered by email. Anything ambiguous is flagged, never silently guessed.
We flag, we don't guess
Generic-purpose classifiers and junior coders alike tend to force a category on anything close enough, which quietly corrupts a codebook's real signal. We don't do that.
reviewer_note column instead of
guessed at.
This isn't a promise — it's how the pipeline is built and tested (0% false-confidence across 41 synthetic verbatims, including an adversarial sample).
Scope note: English-language survey responses only. The coder is keyword-based, so non-English verbatims can't be matched — they abstain safely rather than guess wrong, but a largely non-English batch will abstain heavily. Ask first if you're not sure your batch qualifies.
How it works
- Pick a tier below. Starter pays through Stripe immediately; Project starts with a quick email to scope your codebook and volume.
- Email your codebook (fixed categories + definitions) and your response export to
hello@tsaielectro.com. - If this is a new codebook we haven't coded before, we hand-author the matching keyword ruleset first (the one-time Codebook Setup step below) — a repeat batch on a codebook we've already set up skips straight to coding.
- You get back one spreadsheet: a
category,confidence(MATCH/ABSTAIN), andreviewer_notecolumn per response, plus a one-paragraph summary of what matched and what didn't.
Pricing
Codebook Setup first order only
$60 one-time
Hand-authoring the keyword ruleset for your specific codebook. Charged once per codebook — repeat batches on the same codebook don't pay this again.
Order Codebook SetupStarter
$75
Up to 500 responses, one codebook, one pass. New codebooks also need the Setup tier above.
Order StarterProject
Custom quote
500–5,000 responses, one codebook, one included review round. Quote includes Setup if the codebook is new.
Request a quoteWhat it's for — and what it isn't
Good fit
- Market-research firms and insights teams coding open-ends against a fixed codebook
- Survey platforms/agencies that need a fast first pass before analyst review
- Batches large enough that manual coding is the real bottleneck
Not this (we'll decline & refund)
- Building a codebook from scratch (we code against yours, not invent one)
- Open-ended qualitative analysis / thematic research write-ups
- Non-survey free text (reviews, support tickets) — ask, may be a future fit
FAQ
What format should the codebook and responses be in?
CSV or XLSX for both. A codebook is just your fixed category list with a short definition for each — if you already use one for human coders, that's exactly what we need.
Why is there a separate Setup fee?
The coder doesn't auto-generalize to an arbitrary codebook — a person has to hand-author the matching keyword ruleset for your specific categories before the first batch can run. The $60 Setup fee prices that real, one-time labor honestly instead of hiding it in the per-batch rate. It's charged once per codebook; repeat batches on the same codebook go straight to Starter/Project pricing.
What happens to responses that don't fit any category?
They're marked ABSTAIN with a reviewer note explaining why, not forced into the closest category.
What's your turnaround?
Codebook Setup (new codebooks only): within a couple of days of receiving it. Starter (up to 500 responses): within a few days of receiving your files, once any needed Setup is done. Project: scoped per volume when you request a quote.
Is this a general-purpose AI classifier?
No — it's built specifically for closed-codebook survey coding with an explicit abstain discipline, not a generic text-classification API.
Who is behind this?
Engineer Tsai Studio — a one-person studio building small, automated document and data tools. Built in public on @Tsaielectro.