Typicaly, early game use a cargo hull, fill it with sensor components and maybe one engine. As joeball said park them at key locations.

Mid to late game I have what I call my Galactic Scanners, which usually start with a scan range of 15- but I've gotten them up to 72 (only need 1 or 2 of these).

Another tactic I would suggest, rather than hit just their forward planets is to take more strategic path and hit planets which will allow you to culture flip planets (those that have the most culture). Then protect those more than the others that way even if they do take some back you will simply culture flip them back to you.

A note here:

Due to the mathematics of space and cost when you're building a Sensor Boat, you should favor Engines over Sensors in early designs.  That is, so long as you're restricted to using Interstellar Sensors, always favor having more Engines than IS.

In general, for the early game, the most cost/effective Sensor Boat (using a standard cargo hull with no miniaturization) is a 2 Interstellar Sensor, 3 Hyperdrive+, 2 Environmental Life Support ship. The next one should be replacing the HyperDrive+ with Ion Drives.  It's really not very useful to have more than 2 Interstellar Sensors on a ship, due to the way Sensor Radius scales.

Once you have access to SensorArrays, however, the benefit swaps back to them over drives.  Personally, I find the 2 Ion Drives, 2 Environmental Support, and 3 Sensor Arrays work very well for the "AWACS" roll of border guards and fleet accompanyment. 

Just Remember, you need to keep those Sensors Boats moving. They work best when running racetrack patterns at full speed.

---

As to Survey Modules, yeah, I'm with Philocrthetes:  build them for anomaly collection and raiding.  To start, I make medium hulls, just the Survey Module, Shields (to fight pirates and Graveyards), two of one kind of weapon, 2-3 engines, and as much Life Support as I can get.   When I eventually have enough tech to build better, I still stick with Medium Hulls, Survey Module, 2-3 Sensor Arrays, 3 engines, 3 Life Supports, and sufficient weapons and defenses to take down first the Heavily Defended anomalies, and then enough to kill starbases.  I tend to use them in the same way the Kriegsmarine used the Deutschland-class "pocket battleship" -  singletons able to kill lightly defended stuff with ease, and fast enough (and far-seeing enough) to flee or dodge anything that really threatens them.

(yes, a patrol command would be really nice)

The key is this:

Range and Speed generally increase linearly with the addition of engines and life support.

Sensor radius requires the Arithmetic Sum to increase, and you're only adding SensorPower in fixed increments.  It's similar to a logrithmic progression. I.e. Sensor Range maps to SensorPower like this:  1:1, 2:3, 3:6, 4:10, 5:15, 6:21, 7:28, 8:36, 9:45, 10:55...  More generically, to have a SensorRange of X, you need a minimum SensorPower of (X)(X+1)/2

Even worse, the NEW area swept out by one's sensors is not a function of Radius Squared, rather it's 2x radius. That is, a ship starts out with seeing a number of hexes equal to 1 + 6 * (Arithmetic sum of SensorRange), but, from there out, only sweeps out a number of new hexes equal to ShipSpeed * (1 + 2*SensorRange). 

In short, it means that after adding about 2 sensors (of whatever power), you're ALWAYS a better bet putting more engines than more sensors, if the goal is to maximize new area seen each turn.  % bonuses to sensors and such modify this math, but it's a good rule of thumb.   Of course, that's not the only possible goal for a SensorBoat (early warning is another), so you might certainly see the benefit of a lower-speed, but big-radius ship.  But in pure math terms, engines > sensors  (and especially since Engines cost significantly less than Sensors from a Manufacturing/Maintenance standpoint).