You need to know your % Ultraviolet transmission (%UVT) first of all to see how much UV power you need to penetrate the water.  Low %UVT will require lots of power making it uneconomical.  65% UVT is the cut off usually.  %UVT can be drastically affected by industrial discharges.  Printed circuit boards process uses "UV photo resist" to make the boards.  The waste chemicals may look completely clear in normal light, but they can totally block UV light.  I had a sample that looked like tap water, but I had to dilute 1:100 to get a reading above zero.  Buy a meter and start monitoring %UVT, you can have seasonal changes.

Cleaning is another issue; a pilot is the only way to test how often cleaning is needed.

Bulb orientation is another choice.  If the bulbs are parallel to flow your choices for flow pacing are to turn entire modules on or off or dim them with more complex ballasts.  Lamps perpendicular to the flow can turn on and off individually or in rows as the flow changes which is much more economical than switching on lamps a module at a time.  Perpendicular lamps don't have electrical connections underwater like the parallel ones do.

 Still another choice in low or medium pressure lamps.  Low pressure are more efficient, better if you have high power costs and lots of room and manpower available for manual cleaning, medium is much more compact and powerful, better for higher flows and typically have a mechanical wiper system to reduce manpower costs.

Contact your permit agency field guy and ask him what he has seen work well in your state.  Support is important too, a "fly by night" firm can throw some lamps together and call it a system, don't buy the cheapest stuff you can find, require a minimum number of installations and call them to see how the stuff works long term.

You might be able to tell my bias is to perpendicular, medium pressure with an automatic wiper system for 5 MGD or more.  Perpendicular, low pressure for smaller flows or high power costs.