Are you going to contract this out, or do you already have some instrumentation available. That is, what do you already have the capability to run - Agilent arrays, Affy arrays, Taqman assays (if you have a scanner for it, Taqman 384-well microfluidic cards are a good way to repeatedly collect data on a discrete set of genes of interest) or something else?
Do you need/want to monitor whole genome expression over time, or just a subset of genes.
If you want whole genome data repeatedly over time, is there an array platform already available, or would you need to get a custom array?
It's hard to define "cost effective" for you without knowing exactly what data you need to collect, what you are already capable of collecting in-house, and what is available to you if you plan on outsourcing this - cost effective is a largely relative term.
In my opinion - Transcriptome analysis by SAGE is the most cost-effective way and also the most reliable. Microarray techniques have inbuilt difficulties in validation. We have observed that SAGE results validate much better than Microarray. However, if you are interested in Rare transcripts, then SAGE might have an inherent disadvantage - which can only be overcome by combining it with next gen seq and performing that at a very Deep level. (~1million reads).