When planning an Front-End pool first you should consider the number of servers required to support the required user capacity and also the level of concurrency, i.e. will 100% of users be online at the same time or is usage split across timezones and therefore concurrency is lower.
Next you should consider Multiple Point of Presence (MPOP) will users just be using the Lync Client, or will they use an IP Phone or the Lync Mobile client.
Next consider resiliency or n+1, i.e. if you need 6 servers for capacity then you should have 7 server for resiliency i.e. one server is offline for maintenance/patching or a server has failed.
With Lync Server 2013 pool pairing then you need to consider the capacity of the paired pool, so if pool A needs 7 servers for capacity (including n+1) and pool B (paired pool) required 5 servers for capacity (including n+1) then both server are required to be 12 servers in order to support 100% concurrency if one of the pool fails (i.e. DR is invoked).
With a stretch pool between two geographic locations then you need to consider how many servers are in each location in order to support capacity so if you need 6 servers for capacity then you will need 6 servers in each location, therefore the pool is at least 12 servers.
Finally you need to consider number of servers that must be running for pool to be functional [1] due to fabric manager (or an majority node set quorum), i.e. you need required a pool of 9 servers, then 5 server must need online for the pool to be functional, this is again important for stretched pools.
There are the basics, but there are other factors too, such as the usage model for example the level of conferencing and the percentage of media bypass if collocating the mediation server plus also if the server are virtualized.
Reference
[1] Topologies and Components for Front End Servers, Instant Messaging, and Presence http://technet.microsoft.com/en-us/library/gg412996(v=ocs.15)
