Raw Data Access – Using Piwik PRO for Advanced Attribution Modeling

Unlike other analytics platforms, such as Google Analytics (GA) which are hosted in the service provider’s Cloud, Piwik can be hosted on-premises or in the private cloud in a dedicated database or on dedicated hardware. Cloud-based solutions generally have API limitations around data export. And if you’re an analyst who’s using analytics as a data collection platform for modeling purposes, you may run into issues with GA’s API limits.

Since Piwik PRO gives you the option to host on your own servers one of the many benefits is having raw access to the database and many of our clients use Piwik PRO for exactly for this. By having raw data access Piwik PRO allows you to export as much data as you need, where it’s data on 1 million events, or 100 million events.

Web Analytics Data Is Just One Piece of the Puzzle

In today’s world, data is everywhere and the main challenge is to be able to properly collect all of the data and – more importantly – to connect the data from multiple sources.

Let’s assume you’re a major retailer with an e-commerce store and retail storefronts. Here are some of the data sources which you could have:

• Direct mail data – offers sent to customers’ physical mailboxes.
• Email marketing data – offers sent to customers’ email boxes.
• Web behavioral data – all of the clicks, page views, downloads, purchases, etc., which occurred on a website.
• Ad impression data – impression views and clicks on digital advertising (e.g. banner ad, or Facebook ad).
• TV advertising data – TV commercial(s) viewed by the customer.

For a major retailer outlined above, a typical customer’s buying (conversion) path could have looked something like this:

1. Receives a piece of direct mail promoting the annual sale with special discounts.
2. Views a TV advertisement promoting the annual sale.
3. Goes to the company’s website to check out the sale.
4. Receives and opens an email marketing offer promoting the annual sale.
5. Views a retargeting ad promoting the annual sale.
6. Goes to the physical storefront and makes a purchase.

If you’re a marketer, being able to combine all of these marketing touchpoints and properly attribute the impact of each one to the conversion is essential. However, this is a big challenge most marketers face. Such analysis is very complex from both a data collection standpoint, as well as finding a common identifier to connect all of the data in order to understand every touchpoint that leads to a conversion.

This is why many organizations have data analysts (or data scientists) who work alongside marketers to help with data collection efforts and develop custom attribution models to measure the impact of the marketing channel mix. Having access to the data, such as raw web behavioral data in Piwik PRO, makes this task a lot easier.

Sample Use Case – TV Attribution Using Piwik PRO

One of the common use cases for Piwik PRO is doing offline to online attribution. Particularly measuring the effectiveness of TV advertising campaigns towards driving online visits and conversions. By taking a certain time slot after an airing and the geolocation (windowing) and measuring the impulse response curve, you are able to calculate the lift in website visitor traffic and attribute it to that specific airing.

As a marketer or analyst, you would take the raw website behavioral data from Piwik’s MySQL database and combine it with TV airings data using a tool such as Tableau to create the necessary reports. Some of our clients have also found ways to automate this process by building a custom front-end on top of Piwik PRO, which combines data from TV airings and Piwik PRO and there are even some advanced TV attribution platforms that have been built on top of Piwik PRO.

However, geolocation data is not always perfect and it tends to be a lot more effective for measuring the impact of local TV advertising campaigns, whereas for a national TV advertising campaign for a popular website with millions of visits per day, it can be very challenging to determine exactly how many of the visits were a result of the TV ad.