You could make a case either way, but if the data is going to be used for analysis and you often want to see multiple columns from that data at the same time, go with the wide table. Make sure you know your databases column quantity and row size limits. Make sure you get the datatypes right. If many of the columns are null, SQL Server allows you to optimize the table for that. You could also consider using a NOSQL (Not Only SQL) solution for analysis of this type of data.
If this data is going to be less for analysis, you might want to normalize it as stated in your question.
You have pretty much 4 choices:
NoSQL - definition Every record is stored as a set of Key/Value pairs. It is very flexible and fast. Not all of the report writers out there support this style of storage. There are many example database implementations of NoSQL. The one that seems to be most popular right now, is MongoDB.
EAV - definition This is where you turn either the whole table or a portion (in another table) on its side. This is a good choice if you already have a relational database in-house that you can't move away from easily. The custom info table example you gave is a good example of an EAV table.
Standard tables with XML columns - Think of this as NoSQL meets relational tables. The data stored in an XML column can be any format that XML supports, including multiple correlated sub-data. For the columns that you know are going to be "regular" columns, they can be built as the appropriate type of column to store the data (LastName, Address, City, State, etc.).
Standard tables with lots of extra columns - You have a relational database, you can't use either XML or EAV, and NoSQL is not an option. Add lots of extra columns of each type. I would guess 30 or more varchar, 30 or more integer, 15 or more numerics. And once you use a column for a value, don't re-use it. And don't delete the column either.
Out of all of these solutions, my own opinion is that you will find either the NoSQL or the EAV approach to be the most successful with the least amount of refactoring your code and your schema.
You will have a situation where you collect data one year, not the next, and then collect it again afterward. Trying to get the older data updated with the correct information is problematic and expensive. Storage is neither.
Best Answer
One idea:
Use "day of year", but offset