Simplified AB-Test Example

"Write programs that do one thing and do it well." — Doug McIlroy, Unix philosophy

Make user-driven decisions with a simplified A/B test system. The client side is self contained without external dependencies. The client uses cryptographically random A/B test variant and track the results to a microservice backend.

• Live AB-Test Example
• Ad-Hoc-Statistics

##Setup Example

###Client

open app/index.html

###Service

cd services
subkit update abtest
subkit start abtest

##How-To Testing different versions of text, backend rendered HTML templates, different backend APIs, layouts, images, or colors in the browser has never been more simple.

add variants

Register the a/b test variants. e.g.

AB.test(['red','green','blue'], 1);

frequently you want to run the a/b tests

• 1 will always run the a/b test
• 0.5 will run the test 50% of the time
• 0.2 will run the test two out of ten times
• 0.001 will run the test one thousandth of the time

track shown and chose

Track impression

AB.track({shown:cssClass});

Track visit

AB.track({chose:cssClass});

##Conversion rate statistics How you analyze the actionable data is up to you. In this simplified example the client uses one microservice for a AB-Test created with Subkit to track the impressions and visits for different feature variants.

An ad-hoc Event-Log-Projection is used to analyse and calculate the conversion rates for different feature css-colored-button variants.

Ad-Hoc-Statistics

task.eventsource  
  .from(['AB-css-colored-button'])  
  .run({
    $init: function(){
      return {
        visit: {
          count: 0
        },
        impression: {
          count: 0
        },
        conversion: {}
      };      
    },
    $any: function(state, e){
      if(e.shown){
        state.conversion[e.shown] = 0;
        state.impression.count += 1;
        if(!state.impression[e.shown]) state.impression[e.shown] = 0;
        state.impression[e.shown] += 1;
      }
      
      if(e.chose){
        state.conversion[e.chose] = 0;
        state.visit.count += 1;
        if(!state.visit[e.chose]) state.visit[e.chose] = 0;
        state.visit[e.chose] += 1;
      }      

      return state;
    },
    $completed: function(state){
      for(var idx in state.conversion){
        state.conversion[idx] = state.visit[idx] / state.impression[idx] * 100 || 0;
      } 
      return state;
    }
  }, task.done);
  ```