Upload suitable datasets only to receive accurate results from "Magellan".
What type of data must your dataset contain?
First, your dataset must include a large amount of operational data that relates to the
business area that you want to analyze.
Second, ensure that it contains as many records and fields as possible. When you include as
many records as possible, "Magellan" provides you with excellent analyses. The
application is designed to work with thousands of fields and millions of records. At
a basic minimum, the application requires 100 records, or 10 times the number of
fields in the dataset. Therefore, if the dataset has 1000 fields it needs 10000
records to be accurate. The above figures are a rule of thumb and will vary from
dataset to dataset. After which, you can narrow the scope of the business question
in the User Interface if you wish to focus on a particular area.
In general the more suitable data you give to "Magellan", the more accurate the
analysis will be for you. This is true as long as the information provides business
context.
What are examples of suitable data?
The following types are examples of strong operational data that "Magellan" requires to provide
accurate results:
- In Telecommunications most business questions are about the subscriber, such as the
following: How much they spend on different products? Why they churn? In
this scenario the best type of data is subscriber records.
- In Retail Banking most business questions are about the retail client, such as the
following: Which are the most profitable types of customers? What types of
products do they have? In this scenario the best type of data is client
records.
- In B2B Sales most business questions are about individual accounts, such as the following:
What product mix do they have? Who are the most profitable? Who spends the
most? Who is likely to buy this product? In this scenario the best type of
record is an account record.
How do you ensure that the data leads to a high Insight Quality rating?
"Magellan" Insight Quality provides a star rating (1-5) for the accuracy
and robustness of the model (for use on different datasets). To measure Insight Quality,
"Magellan" runs a series of statistical checks over the uploaded data. After which, it
also returns repeatable pattern in your data. The Insight Quality is high when there is
a sufficiently large number of columns and attributes in the dataset.