A relationship is the mechanism by which separate tables are related to each other. You can think of a relationship as a VLOOKUP, in which you relate the data in one data range to the data in another data range using an index or a unique identifier. In databases, relationships do the same thing, but without the hassle of writing formulas.
Relationships are important because most of the data you work with fits into a multidimensional hierarchy of sorts. For example, you may have a table showing customers who buy products. These customers require invoices that have invoice numbers. Those invoices have multiple lines of transactions listing what they bought. A hierarchy exists there.
Now, in the one-dimensional spreadsheet world, this data typically would be stored in a flat table, like the one shown here.
Because customers have more than one invoice, the customer information (in this example, CustomerID and CustomerName) has to be repeated. This causes a problem when that data needs to be updated.
For example, imagine that the name of the company Aaron Fitz Electrical changes to Fitz and Sons Electrical. Looking at the table, you see that multiple rows contain the old name. You would have to ensure that every row containing the old company name is updated to reflect the change. Any rows you miss will not correctly map back to the right customer.
Wouldn't it be more logical and efficient to record the name and information of the customer only one time? Then, rather than have to write the same customer information repeatedly, you could simply have some form of customer reference number.
This is the idea behind relationships. You can separate customers from invoices, placing each in their own tables. Then you can use a unique identifier (such as CustomerID) to relate them together.
The following figure illustrates how this data would look in a relational database. The data would be split into three separate tables: Customers, InvoiceHeader, and InvoiceDetails. Each table would then be related using unique identifiers (CustomerID and InvoiceNumber, in this case).
The Customers table would contain a unique record for each customer. That way, if you need to change a customer's name, you would need to make the change in only that record. Of course, in real life, the Customers table would include other attributes, such as customer address, customer phone number, and customer start date. Any of these other attributes could also be easily stored and managed in the Customers table.
The most common relationship type is a one-to-many relationship. That is, for each record in one table, one record can be matched to many records in a separate table. For example, an invoice header table is related to an invoice detail table. The invoice header table has a unique identifier: Invoice Number. The invoice detail will use the Invoice Number for every record representing a detail of that particular invoice.
Another kind of relationship type is the one-to-one relationship: For each record in one table, one and only one matching record is in a different table. Data from different tables in a one-to-one relationship can technically be combined into a single table.
Finally, in a many-to-many relationship, records in both tables can have any number of matching records in the other table. For instance, a database at a bank may have a table of the various types of loans (home loan, car loan, and so on) and a table of customers. A customer can have many types of loans. Meanwhile, each type of loan can be granted to many customers.