Hi
What CNC machine is it?
What motors? steps per rev?
what drivers? set to what microsteps?
belt drive or leadscrew? what pitch?
if belt, what size, and what size pulleys, number of teeth?
what are the current GRBL settings? command '$$' in bCNC terminal to get the settings.

lots of questions I know, but with these numbers we can calculate the correct settings for GRBL, and then it will cut nearly the correct size.
After that you can calibrate for fine tuning.
Use 'terminal' in bCNC for changing GRBL parameters.

Thing is that Sketchucam outputs Gcode correctly, using either inch or metric mode depending on how Sketchup is set. Since it uses the actual numbers Sketchup is using, it cannot get the size wrong unless you did the drawing the wrong size.
So, I must conclude that the machine is not calibrated correctly.
This means that GRBL has been told to use a certain size of 'steps per mm' that is incorrect for your machine.

So, first calculate the steps/mm
motor: 200 steps per rev
driver microstepping = 8
belt pitch = 2mm
pulley: 20 teeth

so, one rev of the motor will move the belt (20 * 2) mm
one rev of the motor takes (200 * 8) steps
therefore:
(200* 8)steps / (2 * 20)mm = 40 steps per mm

That value would be set for GRBL config items $100 and $101 for X and Y. With belts you still need to calibrate after this, with quality ballscrews, it should be spot on (-:

Maybe you can try the calibration procedure at (you may need to load the page twice to get to the correct post of mine)
http://openbuilds.com/threads/openbuilds-ox-cnc-machine.134/page-57#post-8636
and let us know what you find?
(Using that procedure does not require answers to any of the the questions about the hardware , just measure what it is doing and recalculate the steps/mm from that.)

Fundamentally, if you command the CNC to move an inch, assuming X is now at 0, then the Gcode command
G20 G0 X1
should move the tool exactly 1 inch to the right. If it does not move 1 inch, then the calibration is off.
You need to calibrate each axis separately, you cannot assume that what works for X also works for Y (difference in belt tension), and definitely not for Z since it is usually a leadscrew when X and Y are belts.

(added detail...)

For anyone reading this that has a completely unknown hardware setup and no clue where to start, just set the steps/mm to '1'.
This means that for every 1mm you command it to move, it will move the motor 1 step.
now you can command a move of 200mm, and if the motor is not using microsteps and is a 200 step per rev motor (1.8degree per step) it will go around once. if it goes half way round, it might be a 0.9 degree per step motor (400 step per rev), or it may be in half step mode = 2x microstepping.

You can also 'jog' until the motor does one revolution, and then read off the 'distance' covered which will give you a clue to the steps per rev.

However, you don't need to know, you only need to follow the process I linked, command a long enough move to measure, do the math, adjust the steps/mm, and do it again until satisfied with the accuracy.
but be careful! after calibrating the long moves you needed at first will be too long for the machine.
say I have a 200 step/rev motor at 16x microstepping, and a 8mm pitch leadscrew. (but I don't know that).
I set steps/mm to 1 and command a 200mm move.
G0 X0
G0 X200
I observe a roughly 1/16th rotation of the leadscrew, a distance of 0.5mm, so I can obviously move more to get 1 rotation.
I can jog till I get 1 rotation, or command
G0 X3200
which is 16 times 200.
Now it should have moved 8mm if my guess at 1/16th of a rotation was correct.
do 'Go X0' and then
set the steps/mm for the X to 400 (for the case above), now a
G0 X10
should move X 10mm.

required reading:
https://github.com/grbl/grbl/wiki/Configuring-Grbl-v0.9

 

hi
looks like one rev on a leadscrew will move 8mm. (2mm pitch * 4 start = 8mm)

so, if you want to mill really small features, like thin tracks on PCB's, set to 400 steps/mm
set drivers to do 16x microstepping
and set GRBL parameters
$100=400 (x, step/mm)
$101=400 (y, step/mm)
$102=400 (z, step/mm)
theoretical resolution is 0.0025mm per step. this will vary, stepper motors are not exact.
(this will limit your max speed!)

if you just want to mill within 0.1mm, ie normal stuff (-:
set the drivers to 4x microstepping
and set GRBL to 100 steps/mm
$100=100 (x, step/mm)
$101=100 (y, step/mm)
$102=100 (z, step/mm)
theoretical resolution is 0.01mm, and speed should be quite good.

the fewer microsteps you use, the more accurate it is, and the more power it has. more power = more speed.

for each axis
start with a low max feed rate, say 500mm/min, and adjust the acceleration up until it stalls, then use 75% of that figure.
now adjust the feed rate up until it stalls, and then use 75 to 80% of that figure.
now run an actual job and check for lost steps, if it loses any steps during the cut, drop the acceleration a bit.
and so on.....

standard test cuts are
square - to check for size, and diagonals must be equal, if not frame is not square.
circle - to check for backlash, any kinks indicate backlash.

you must know the exact diameter of the bit, and check that the slot cut is that size, before judging the size of a cut square.
if the slots are wider than the bit, then you have runout in the spindle.

 

Hiya
sure, just link to it here is probably best, then people can see the entire context. and that is after all the purpose of the web, link to existing information instead of endlessly repeating it.

To link to a post, look for the permalink, just left of the 'Like' link, there is a #2 (or whatever number) that is the link to the post in quetion.