Confidence.js is a light-weight JavaScript library to help you make sense of your A/B test results.
Include confidence.js on your page.
<script src="path/to/confidence.js"></script>var myConfidence = new Confidence();Confidence helps you compare the variants in your A/B test. Variants in Confidence.js look like this:
variant = {
id: 'A', // short identifier
name: 'Variant A', // descriptive identifier
conversionCount: 50, // number of events that successfully converted
eventCount: 300 // total number of events tracked
}Adds a variant to your A/B test. You can add and compare as many variants as you'd like.
Parameters:
-
variant: the variant object you'd like to add to this A/B test
// first, create some variants
variantA = {
id: 'A',
name: 'Alluring Alligators',
conversionCount: 1500,
eventCount: 3000
}
variantB = {
id: 'B',
name: 'Belligerent Bumblebees',
conversionCount: 2500,
eventCount: 3000
}
// then add the variants to your A/B test
myConfidence.addVariant(variantA);
myConfidence.addVariant(variantB);Evaluates the variants in your A/B test and determines which is the winning variant, if there is one.
Returns an object containing:
-
hasWinner:trueif a winner could be calculated,falseotherwise -
hasEnoughData:trueif there is enough data to calculate a statistically significant result,falseotherwise -
winnerID: the ID of the winning variant, ornullif there isn't one -
winnerName: the name of the winning variant ornullif there isn't one -
confidenceInterval: the confidence interval, ornullif there is no winner. - ex:
{ min: 0.154, max: 0.187 } -
readable: human readable result. - ex:
There is not enough data to determine a winner.
Case 1: There is not enough data to determine a result.
// create some variants
variantC = {
id: 'C',
name: 'Cranky Capybaras',
conversionCount: 5,
eventCount: 50
};
variantD = {
id: 'D',
name: 'Diligent Ducklings',
conversionCount: 60,
eventCount: 200
};
variantE = {
id: 'E',
name: 'Effervescent Elephants',
conversionCount: 30,
eventCount: 40
};
// add the variants to your A/B test
myConfidence.addVariant(variantC);
myConfidence.addVariant(variantD);
myConfidence.addVariant(variantE);
// evaluate the variants to get the result
result = myConfidence.getResult();
/*
{
hasWinner: false,
hasEnoughData: false,
winnerID: null,
winnerName: null,
confidenceInterval: null,
readable: 'There is not enough data to determine
a conclusive result.'
}
*/Case 2: There is enough data, but there is no clear winner.
// create some variants
variantF = {
id: 'F',
name: 'Freaky Flamingos',
conversionCount: 1501,
eventCount: 3000
};
variantG = {
id: 'G',
name: 'Gregarious Gorillas',
conversionCount: 1500,
eventCount: 3000
};
// add the variants to your A/B test
myConfidence.addVariant(variantF);
myConfidence.addVariant(variantG);
// evaluate the variants to get the result
result = myConfidence.getResult();
/*
{
hasWinner: false,
hasEnoughData: true,
winnerID: null,
winnerName: null,
confidenceInterval: null,
readable: 'We have enough data to say we cannot
predict a winner with 95% certainty.'
}
*/Case 3: There is enough data and there is a clear winner.
// create some variants
variantH = {
id: 'H',
name: 'Hungry Hippopotami',
conversionCount: 2500,
eventCount: 3000
};
variantI = {
id: 'I',
name: 'Irritable Iguanas',
conversionCount: 1500,
eventCount: 3000
};
// add the variants to your A/B test
myConfidence.addVariant(variantH);
myConfidence.addVariant(variantI);
// evaluate the variants to get the result
result = myConfidence.getResult();
/*
{
hasWinner: true,
hasEnoughData: true,
winnerID: 'H',
winnerName: 'Hungry Hippopotami',
confidenceInterval: { min: 82, max: 84.67 },
readable: 'In a hypothetical experiment that
is repeated infinite times, the average
rate of the "Hungry Hippopotami" variant
will fall between 82% and 84.67%, 95%
of the time'
}
*/npm install
npm test
- Variant
nameparameter optional - requires changes to
addVariant, anderrors.js - add "not provided" default name if left blank
- add
removeVariantfunction - zscore table lookup to provide more accurate results if 95% confidence is not available
Found a bug? Create an issue here on GitHub!
For general questions, tweet me @jessicaraygun.
Developed and maintained by Jessica Thomas, Data Scientist @ sendwithus.com