Позволю себе процитировать отрывок из книги - список того, чего нет в BISM http://www.amazon.com/gp/product/0735658188/?tag=se04-20]
Marco Russo & COHere is a list of the most important functionality missing in Tabular:

Writeback , the ability for an end user to write values back to a Multidimensional database.
This can be very important for financial applications in which users enter budget figures, for
example.

Translations , in which the metadata of a Multidimensional model can appear in different
languages for users with different locales on their desktops. There is no way of implementing
this in Tabular.

Dimension security on measures , in which access to a single measure can be granted or
denied.

Cell security , by which access to individual cells can be granted or denied. Again, there is no
way of implementing this in Tabular, but it is only very rarely used in Multidimensional.
16 Microsoft SQL Server 2012 Analysis Services: The BISM Tabular Model

Ragged hierarchies , a commonly used technique for avoiding the use of a parent/child
hierarchy. In a Multidimensional model, a user hierarchy can be made to look something like a
parent/child hierarchy by hiding members if certain conditions are met; for example, if a mem-ber has the same name as its parent. This is known as creating a ragged hierarchy. Nothing
equivalent is available in the Tabular model.

Role-playing dimensions , designed and processed once, then appear many times in the
same model with different names and different relationships to measure groups; in the
Multidimensional model, this is known as using role-playing dimensions. Something similar
is possible in the Tabular model, by which multiple relationships can be created between two
tables (see Chapter 3, “Loading Data Inside Tabular,” for more details on this), and although
this is extremely useful functionality, it does not do exactly the same thing as a role-playing
dimension. In Tabular, if you want to see the same table in two places in the model simultane-ously, you must load it twice, and this can increase processing times and make maintenance
more difficult.

Scoped assignmentsand unary operators , advanced calculation functionality, is present in
MDX in the Multidimensional model but is not possible or at least not easy to re-create in DAX
in the Tabular model. These types of calculation are often used in financial applications, so
this and the lack of writeback and true parent/child hierarchy support mean that the Tabular
model is not suited for this class of application.
The following functionality can be said to be only partially supported in Tabular:

Parent/child hierarchy supportin Multidimensional is a special type of hierarchy built from
a dimension table with a self-join on it by which each row in the table represents one member
in the hierarchy and has a link to another row that represents the member’s parent in the hier-archy. Parent/child hierarchies have many limitations in Multidimensional and can cause query
performance problems. Nevertheless, they are very useful for modeling hierarchies such as
company organization structures because the developer does not need to know the maximum
depth of the hierarchy at design time. The Tabular model implements similar functionality by
using DAX functions such as PATH(see Chapter 9 for details), but, crucially, the developer must
decide what the maximum depth of the hierarchy will be at design time.

Support for many-to-many relationshipsin the Multidimensional model is one of its most
important features, and it is frequently used. (For some applications, see the white paper at
http://www.sqlbi.com/articles/many2many/.) It is possible to re-create this functionality in
Tabular by using DAX, as described in Chapter 12, “Using Advanced Tabular Relationships,” but
even though query performance is likely to be just as good if not better than Multidimensional
when using this approach, it adds a lot of complexity to the DAX expressions used in mea-sures. If a model contains a large number of many-to-many relationships or chained many-to-many relationships, this added complexity can mean that maintenance of the DAX used in
measures is extremely difficult.

Drillthrough , by which the user can click a cell to see all the detail-level data that is
aggregated to return that value. Drillthrough is supported in both models but, in the
Multidimensional model, it is possible to specify which columns from dimensions and mea-sure groups are returned from a drillthrough. In the Tabular model, no interface exists in
SQL Server data tools for doing this and, by default, a drillthrough returns every column
from the underlying table. It is possible, though, to edit the XMLA definition of your model
manually to do this, as described in the blog post at http://sqlblog.com/blogs/marco_russo/
archive/2011/08/18/drillthrough-for-bism-tabular-and-attribute-keys-in-ssas-denali.aspx. A
user interface to automate this editing process is also available in the BIDS Helper add-in
( http://bidshelper.codeplex.com/).

Лично мне больше всего будет не хватать Translations и M2M