Scalability
For every trade message, there exist 99 other non-trading interactions (100:1); platform compresses, terminates, multiplexes all non-trade messages, while forwarding only the trades to broker's side.
Additionally mobile use can be unpredictable, as the news breaks; this cushions broker side api against flash trading traffic using infinite resources of computing cloud.

 Device CPU/Mem Limits
Mobile devices have limited resources, such as cpu and memory, and thus cpu intensive processing such as technical analysis cannot be run easily, however platform "lends" the CPU from cloud to make the actual device act as thin client, this ensure snappy real-time feel to app, while still maintaining rich resource heavy feature set.

 Just-in-time Push or Always-on Drain
Battery Limitation apply on alerts unlike on a desktop; if a device is left with streaming, the battery on most blackberries would be completely drained out within a hour, however using a split thin/client-server model lets server (or in this case cloud) to keep tabs on alert levels/algorithms set by clients, only to wake up the app just-in-time, when required rather than run always. This also include advanced technical alerts such as a alerts on crossing moving averages, PSAR reversals, etc.

 Cross-Device Consistency
Each mobile device is different in architecture, offering a consistent application requires middleware abstraction for interactions to a simplified common interface. Same feature sets can be offered on BlackberryOS(MicroJava), iphone(ObjectiveC), Symbian(C++), Palm(WebOS) & Android(Java+)

 Manageability
Dynamic provisioning of features and permission on the go are possible; example, same application binary can provide level-2 view for power traders while level-1 for normal users during normal course, while during certain time/promotional periods enable special features to entice upgrades. Also another use is for providing customer support; most issues can be easily diagnosed with a look at the management console.

 Every changing Trade Catalysts
Analysis & Signal sources are plenty but they keep rotating on what's in fashion this week, while setting up a static stand-alone application cannot be so; a middleware abstracts such "Value Add Drivers" across all applications & devices in one go to keep a fresh supply of analysis, recommendations, signals & news sources.

 Use it or just adore it
One other reason of "stand-alone" mobile trading apps do not pick up mainstream traction are all the facts, usability issues, limits above, making simple tasks awkward, which on a desktop are taken for granted(cpu/memory/screen size/battery). This point can be validated by looking at the case of blackberry's best practice of having a middleware with BES middleware; BES provides much of the seamless capabilities similar to above, which makes the device very elegant & usable in contrast to other standalone email enabled phones. This is crucial for "active trading" to be sustained beyond just being a novelty prototype that looks nice, but is dumb.