What to expect if I compare GA to DataMilk
While GA can provide session count breakdowns, it does not clean the data for outliers and is not designed for measuring financial uplift.
In contrast, DataMilk has its own proprietary bot recognition technology and cleans the data for weird sessions or orders, and also provides confidence intervals and evaluations of the significance of the results.
You can use GA to validate:
- Overall session counts (not expected to match exactly)
- Breakdown session counts by original and optimized during periods of stable original vs optimized traffic ratio.
- DataMilk’s effect on conversion rate during periods of stable original vs optimized traffic ratio.
However, GA will not provide confidence intervals. GA does not clean the data for outlier purchases and bots for example. Furthermore, GA is not statistically designed to correctly measure financial uplift.
Note: When viewing GA and comparing to DataMilk, you have to choose to view the period before or the period after smart traffic ratios were adjusted (if ever). We can’t view the whole time period when there is a traffic ratio change in the middle.
What’s the difference between DataMilk’s data vs GA?
- GA expires sessions at midnight, DataMilk does not
- DataMilk developed its own proprietary bot recognition technology
DataMilk may clean the data for other weird sessions or orders, for example:
Strong outliers in average order value.
- See this article for detailed overview about outlier detection & removal
- DataMilk only attributes orderIds once for the sessions and also discards previously seen pixels.
- Strong outliers in average order value.
- DataMilk balances the sessions (learn more here: https://medium.com/datamilk/why-we-need-to-balance-visits-when-calculating-conversion-rate-uplift-c9b9739291a9)
- We calculate confidence intervals and display evaluations of the significance of the results.
- You may be filtering data by IP address in GA, DataMilk is not (If you want DataMilk to filter your data by IP address, please email your Customer Success Manager your request.)
We throw away all orders where any of the following values are ‘undefined’
- orderId (which must be unique for each order)
- currency code
Here is an article on how to generate a custom DataMilk report in GA