Lecture_2 Flashcards
Omnichannel data collection helps
Marketer know what your customer want before they do
Which is No.1 enemy of profitable use of ML?
Poor data quality
What are the requirements of historical data if a good predictive model is required?
Exceptionally broad （广泛的） and high quality standards.
The requirements for historical data
Data must be right (correct, properly labeled, deduped)
Data must be unbiased ( ML用于法庭案例会有种族bias)
Acceptable data range
97-100
How to improve data quality?
- Clarify project objectives and only choose the right data.
- Execute data quality fundamentals.
- Maintain an audit trail as preparing the training data.
- Charge a specific individual with responsibility for data qualty.
- Obtain independent rigorous quality assurance.
How to collect identity data?
- POS system/online transaction database.
- Clienteling（shopping associate built a profile for customer）
- Social network profile.
- Third-party data source.
One thing that need to notice when you get data from third-party sources
The data information from a third party should be treated differently from your primary data.
How to collect quantitative data
Transaction database
Web Analytics Tool
3rd Party pixels
In-store tracking
How web analytics works?
Create account and add Javascript code to your website.
Every time a user visit a page, it will collect the information from the user’s browser.
Then the data will be transfered to the Google server, which will generate a report for you.
Web analytics tools allow us to track every action the customer is performing on a website.
True. Including browser behavior, search data, purchase history.
The challenge of web data
If a customer uses different browser to visit a site, the cookie in that site will consider this unique customer as different visitor.
What is difference between cookie and 3rd party pixel?
3rd party pixel such as Facebook or Google pixel will collect data from different websites if the pixel is deployed on these websites.
The definition of customer data silos
Company will use multiple tools to store different customer attributes.
Marketing teams don’t operate the aggregated database, so that data fragmentation exists and it leads to loss money and effectiveness of campaigns.
The definition of identity graph.
The database that stores all identifiers that correlate with individual customers. So that it can create a unified customer view and break the data silos.
