Deployment to Customers (WIP)
- Ondřej Tesař
Analytics
Concept
The schema above shows how we get from the customer's unique Transaction and Quote data structures to a uniform Plasma structure.
Please note that:
The "Extractor" and "Distiller" steps happen on the Customer's partition
There are two versions of the extractor: one for Quotes and one for Transactions. They operate independently and have their own configurations.
The extractor takes only the data we need from the customer's data and also takes care of the required translations and lookups. For example, using a part of the customer's customer address to look up the region (sub-continent), which is the only regional information Plasma needs.
After the Extractor step, the Standardized DM has the required information, at transaction/quote level.
The "Distiller" step compiles monthly metrics from the Standardized DM, with additional keys for SellingFrom (region), SellingTo (region) and Product/Service.
As there are 11 regions, the maximum number of records is 242 per month (11x11x2). In reality this number is much lower as not every possible region permutation will occur. Again, there are separate distillers for Transactions and Quotes
After the distiller step, the Metrics are stored in a Data Mart on the customer's partition, one for Transaction Metrics and one for Quote Metrics.
A "Harvester" Process that runs on our central Plasma partition checks for new Metrics data regularly and if new data is encountered it will download it. Depending on the partition the data is coming from, it ads a unique key, which is kept in the Harvester config. This way it is not possible to know which customer the data is from but it is still possible to maintain consistency over time.
The Plasma Metrics are handed over to Bain & Co. over SFTP monthly. They load this data into their benchmarking platform. Bain analysts have access to this platform, but not to the raw data. They can query the benchmarking platform and if there are at least 5 entities that have data, the platform will return the KPI.
A standard set of KPI's is exported each month by the benchmarking platform and made available to Pricefx. We then load this data into a Data Mart of all participating partitions.
Deployment
Eventually, the components for Plasma (Extractor, Distiller and all related tables, Data Marts and Calculation Flows) will be deployed through Platform Manager using a package.
Until the package is fully tested, the necessary components will be deployed using Pricefx Studio, from a Bitbucket repository.
Configuration
The extractor needs to be configured so every required field in the Transaction and/or Quote Standardized Transaction DM is available.
The Extractor is configured through a Pricing Parameter table: full documentation <here>