One objective is to have enough more security (or apparent security) than your competitors so that the thieves go elsewhere first.
One problem with many systems is they are ex post facto-- you see the DVR record after the fact, the items are gone, and probably it helps to recover the stolen items but the best case next to prevention is to catch the thieves in the act.
Real time response has the problem of falsing-- enough false alarms and no one will respond.
Something that has not yet caught on widely is to have alarms filtered by humans-- they can even be outsourced somewhere cheap (yeah, no one likes that until they get the bill for first world drones to stare at video streams). With IP cams and Internet access the filtering humans can be anywhere, the level1 escalates to a first worlder if the action seems suspicious.
If the signal cuts off, the local hires local security-- should not happen that much, so the cost can be contained.
Once you stop the outsiders, then you get to deal with the insiders-- those are the ones that can really clean your clock.