GratisAIlearning · field guide · local-first · no account
response-rate index card

Keep Survey Percentage Points Separate from Percent Change — Free

A response rate rising from 30 to 40 percent moves ten percentage points, not ten percent. Give each survey wave its own denominator before discussing a change.

The proof surface

Reading-comprehension trends compare learning outcomes. This examines survey completion fractions against a chosen baseline without making population-inference claims.

Inputsurvey wave | invitations delivered | completed responses; reference response rate percent
Rare deviceReading-comprehension trends compare learning outcomes. This examines survey completion fractions against a chosen baseline without making population-inference claims.
Output artifactpercentage-point difference per row with transparent workings
Cost$0 · local-first · one free run · samples labelled
Sample, not your facts: Illustrative sample: Spring | 200 | 100; Autumn | 100 | 30. With reference response rate percent set to 40, pooled response-rate point difference is 3.33 percentage-point difference.

Why the flat version breaks

Changing the denominator between waves

A rate using delivered invitations is not directly the same measure as one using a mailing-list size.

Reporting points as relative growth

Ten points added to forty percent is a 25 percent relative increase.

Calling more responses representative

Response count alone does not establish a representative sample.

How to work the response-rate index card

Count delivered invitations

Use a documented denominator for each wave. Decide how bounced invitations and duplicate invitations are treated before comparing the rates.

Count completed responses only

Record whole counts no larger than invitations. Mixing partial responses into one wave but not another makes the arithmetic consistent and the comparison misleading.

Subtract the reference in percentage points

Compute 100 × completions ÷ invitations minus the reference. A 50 percent response rate against 40 gives +10 points; a 30 percent rate gives −10.

Report waves individually

The summary pools completions and invitations, then subtracts the reference; larger invitation groups carry more weight. Negative rows are below your reference. Bias, nonresponse and representative sampling require separate analysis.

What the response-rate index card replaces

QuestionBeforeVisible working
Changing the denominator between wavesA rate using delivered invitations is not directly the same measure as one using a mailing-list size.Count completed responses only
Reporting points as relative growthTen points added to forty percent is a 25 percent relative increase.Subtract the reference in percentage points
Calling more responses representativeResponse count alone does not establish a representative sample.Report waves individually

A local arithmetic aid replaces hand calculation, not expert review. No paid AI service is needed.

Run it on the samples, right here

FIRST-LOAD

Research prototype — checkout unavailable. Input is processed locally and a draft is saved automatically in this browser profile when storage is available. Reset to sample clears that draft; Pro history has its own clear button. A state link encodes your inputs in its URL: share only non-sensitive rows. Browser history, clipboard and anyone receiving the link may retain it. Optional AI use below leaves this device; it is not required.

Descriptive arithmetic, not statistical inference or research ethics advice. Use anonymous aggregate counts; no respondent data is needed.

Data note: Everything runs in this browser tab on the response-rate index card: your survey wave | invitations delivered | completed responses stays on this device, nothing is uploaded, and the reading is rebuilt only when you press run.

Go deeper: the companion app files the same reading as a index card with a stamped header

The article demo above runs without limits. The companion app keeps a local history, exports the rows as CSV, prints the response-rate index card reading, and holds your drafts on this device — one complete free run; the proposed $4 one-time filing layer is not currently for sale.

The response-rate index card reading is complete for free. The optional $4 layer adds print, row CSV and the last five local reading summaries; it does not add hidden answers. Checkout is not configured yet; the article demo remains unlimited.

Open the keep survey percentage points separate from percent change companion

Boundary

Descriptive arithmetic, not statistical inference or research ethics advice. Use anonymous aggregate counts; no respondent data is needed.

What this is built on

Before: A rate using delivered invitations is not directly the same measure as one using a mailing-list size. After: the response-rate index card shows the working beside each named row so the reader can change the assumption and inspect the consequence.

Optional AI formatting, not calculation

For this response-rate index card, use the free AI Studio interface only if available to you, with no paid API key. Manual entry completes the same workflow for free. Supply only fictional or non-sensitive notes. Review its output against the source; never paste unresolved questions into the numeric rows.

Format my non-sensitive notes for a response-rate index card. Return plain rows only: survey wave | invitations delivered | completed responses. Preserve supplied quantities exactly. Do not guess missing values; list questions separately. Do not calculate or add advice. I will check every row before pasting into the local tool.

Accepted schema: survey wave | invitations delivered | completed responses. No AI response is executed as code.