Please login

Use the login button in the left sidebar to start analysing your Google Analytics data

What does this app do?

Was that success or just random noise?

This webapp uses Bayesian time-series to model your Google Analytics traffic under normal circumstances.

You can then pick a date when you know 'something' happened (the event) and compare the model's counterfactual trend to what actually happened.

This method attempts to give you an idea of what effect your chosen event had on your website metrics.

Read an introduction to the app on

You can draw data using any Google Analytics segment and from the selected metrics to let the model help answer a wide variety of questions, such as:

What effect did those SEO changes make to my SEO traffic?

Did my direct traffic increase during my first TV commerical run?

Has my website relaunch increased the number of new visitors significantly?

...and so on.

Select Dates

Choose date range of data. Include lead time before and after event date for sensible results.

Select Test Segment

This is the data that you want to see if the event effected. You can create more segments in the normal GA interface. Refresh this page afterwards to load them here.

Control Segment

Select a segment of data that will be used as control, unaffected by the event. This makes the model much more robust.

Click-Drag to Zoom. Double-Click to Reset. Shift-Drag to Pan.

When did the event to be tested happen?

This date is what will be tested to see if it had a statistically significant impact. Example: Test SEO impact of title tag changes: Segment = Organic Traffic; Control = Referral Traffic; Metric = Sessions; Event Date = When <title> tag changes went live.

Does the Data Have Seasonality?

If the data includes seasonal trends, this can affect your results. Selecting a seasonality will let the model attempt to adjust for natural seasonal variation when judging your effect. If date range is over 1 year you can select annual/monthly seasonality, if under 1 year you can select weekly/monthly seasonality.

How did the expected trend with no effect present compare to the observed?

How did the observed trend with the event perform verses the expected trend without it?

CausalImpact 1.0.3, Brodersen et al., Annals of Applied Statistics (in press). Your Google Analytics data is only used for the statistics carried out above and all data is forgotten when your browser closes or you refresh the page.
Copyright (c) 2020 IIH Nordic A/S - v 0.1 If you have special circumstances or needs to build on top of this free demo, or want to work with IIH Nordic for digital marketing/cloud data science needs, please get in touch for a quote. Please review IIH Nordic's Terms of Use and Privacy Policy