CEDX Surveys · Marketing
Surveys treats response quality as a measurement problem: every question carries its drop rate, every response carries a bias score, and the NPS is computed only from rows that passed.
Fourteen completes with identical Likert selections is not signal. The straight-line cluster alert catches it before it reaches a deck.
Runs on demo data. The chrome itself reads demo seed · no live collectors.
What it is
320 rows with channel, status, time, score and bias. A 57-second complete on a research survey scores bias 17.5; a six-minute response that contradicts itself gets disqualified at 16.3. Spam and disqualified are statuses in the table, not deletions.
The question bank shows type, survey, required, drop and salvage per question. Required ranking questions drop hardest — 34% abandon the packaging-ranker — and the salvage score says which are worth rewriting.
The ring reads 78 — quality plus completion — and the alerts are specific: a straight-line cluster on onboarding CSAT, a 34% drop-off on a ranking question placed after long text, an orphan survey owner after a rotation.
Product tour
Not a mockup and not a concept deck. This is what opens at /app/surveys.
01 — Overview
Starts and completes trend together until they do not — the annotations mark a pricing-pulse spike, an in-app collector launch, a straight-line cluster, and an exit survey shortened by three questions.
02 — Responses
Filter to complete-only or bias ≥ 15 and the table complies — 229 completes of 320, 48 above the bias threshold. The NPS mix is computed from scored rows, so the 39 means something.
03 — Questions
When a required ranking question drops 34% after a long text field, the respondent is not the problem. The bank shows it — drop 34%, salvage 50.6 — and the rewrite starts from evidence.
Who runs it
Roles, not references. We have no named customers yet, so nobody in these photographs is quoted, credited or claimed as one.
Designs the instruments and reads the bias column before the scores — a 39 NPS on clean rows beats a 45 on dirty ones.
97 scored rows · NPS 39
Works the salvage queue: the questions that lose a third of respondents get rewritten, split or dropped.
27 hot salvage steps
Collects on QR and in-person channels — the same ledger scores a kiosk response and an email one.
channels · QR to in-appThe shape of it
Every figure below is legible in the captures above. Nothing here is a projection of your estate — it is the state of the demo data.
How it runs
Questions come from a bank that remembers drop rates, so known-losing formats get flagged at design time.
QR, link, in-app, SMS, email and embed all land in one response ledger.
Bias scoring catches straight-liners and speeders before the aggregate does the lying.
The salvage queue ranks the questions worth rewriting; the NPS mix only counts rows that passed.
One record
A response is attached to a person the estate already knows — so scores, journeys and support all read the same truth.
The capture sibling: short forms and long surveys share spam scoring and the drop model.
CEDX FeedbackThe always-on listening posts the survey program feeds and calibrates against.
CEDX JourneysDetractor responses can open a save path; promoters open the advocacy one.
CEDX DeskCSAT after a ticket lands next to the ticket, not in a separate silo.
Limits
Finding this out on the third call is worse for you than reading it here, and worse for us.
Start
Open it right now
Your instruments, your respondents
Surveys with the rest of it
Questions
Yes. Every screenshot is a capture of the running build and you can open the same build at /app/surveys. The chrome reads demo seed · no live collectors.
Straight-lining — identical selections down a Likert matrix — and speeders, among other patterns. The demo flags 48 responses above threshold and keeps them visible for review.
Because an aggregate that includes flagged responses is a worse number, not a bigger one. The 39 comes from 97 rows that passed screening, and the screen says so.
Yes. The question bank carries drop and salvage per question — required ranking questions after long text drop hardest, 34% in the demo, and that is where the rewrite starts.
No certification has been issued. What we can evidence about hosting, encryption, tenant isolation and production access is written up on the security page.
Live build, demo data, no card. Then ask which of your questions would land on the salvage queue.