CODEWITHSUNDEEP
google-analyticsgoogle-experiments
andi
andi

Reputation: 6522

In Google Experiments, how can I use Conversions per Unique Visitor as the Conversion Rate

We're running A/B Testing Experiments through Google Analytics. In one experiment, we defined a "conversion" as being when a user signs up for an account on our site. Each unique visitor can only convert once, since then they'll already have an account and can't sign up again. However, Google Analytics calculates the Conversion Rate as being Conversions per Experiment Session, so that if a user who already converted returns to our site, it will count as another session without a conversion.

I believe that this is messing up our results, because registered users are more likely to return to our site, so they would be adding sessions but not conversions to the total count, thus lowering the Conversion Rate for their variation. I would like Google to choose an Experiment Winner based on a Conversion Rate defined as Conversions per Unique Visitor. Is this possible?

Here's the experiment data (with a 50/50 split):

Original:
{Experiment Sessions: 652, Conversions: 53, Conversion Rate: 8.13%, Compare to Original: 0%, Probability of Outperforming Original: 0.0%}
Variation 1:
{Experiment Sessions: 492, Conversions: 54, Conversion Rate: 10.98%, Compare to Original: 35.02%, Probability of Outperforming Original: 94.6%}

In this case, the Original had about the same number of conversions, and assumingly about the same number of unique visitors, but quite a bit more sessions. I assume this to mean that people who saw the original version were more likely to return to the site in another session after account creation. However, Variation 1 is approaching 95% confidence because of the higher Conversion Rate, and is likely to be the winner. This just doesn't seem right to me. Is there a way to get better results here?

Upvotes: 0

Views: 97

Answers (1)

Lucia
Lucia

Reputation: 4767

I believe the best approach here would be to use a custom analytic's event and trigger it according to your rules. Then on the experiment you choose this event as the conversion goal for it.

Upvotes: 0

Related Questions