Procedure for running a prep TLC

1. Cut the plate in half by scoring with a pencil, so there is no contact of silica across the middle of the plate. This is a good general practice as a whole plate is typically not necessary when dealing with small to moderate amounts of sample (5-25mg). This is a good practice both for economical reasons and to avoid line broadening which obscures the boundaries of the bands which are later scraped. For larger amounts of sample, the entire plate may be used.
2. Gently mark a pencil line roughly 1-1.5 inches from one side of the plate (careful not to scrape the silica gel). This is the "origin" line.
3. Prepare a relatively concentrated solution of your crude sample (~1-2 mL) in a fairly low-boiling solvent (i.e. DCM or Et₂O). Warning: too large a volume will require multiple applications which broadens the band.
4. Using a short pipet, carefully deposit a thin line of sample across pencil line, avoiding the edges (stay roughly an inch away from each edge of the plate). It is very important to do this slowly and uniformly, without contacting the pipet too much against the silica as this will scrape it. The main difficulty is trying not to accidentally plunger too much at one place. Intentionally drawing the solution into the pipet such that there are periodic air spaces up the stem can make application easier, as well as sucking up a small amount of solution at a time into the pipet. Sometimes using a syringe and needle helps in controlling the dispensing, but the needle may scrape the silica too much.
5. Apply remaining solution repeatedly on the original line of sample, drying in between applications. Not drying fully after an application will cause the solution to diffuse outward, broadening the band. Also try to spot in areas that didn't get much sample to make the line of uniform sample loading.
6. Obtain a prep TLC chamber. This is a large, heavy, square glass chamber. It is useful at this point to narrow the sample band by running a polar solvent (typically Et ₂O or EtOAc) up the plate to the upper edge of the band, and then evaporating the solvent (this should be repeated as needed, typically 3 times). Then in a dry chamber prepare about 100 mL of eluent. The ideal eluent should give your compound an Rf of around 0.1 on regular TLC. This will allow multiple elutions if the first one doesn't separate well enough.
7. Place the plate in the chamber and seal the top with the lid or aluminum foil. A typical run takes 40 mins to 1 hour.
8. Remove the plate and visualize under UV. If your desired band(s) are not separated enough, repeat run another elution.
9. If you need to stain the plate to see where the band(s) is, you can pipet a thin line of the stain down an edge of the plate, and heat. This will reveal approximately where the bands are, going across the full width of the plate.
10. Another staining option is berberine (an alkaloid), which is prepared as a 0.1% solution in EtOH and should be sprayed on the plate, and allows for visualization under UV light. Because it is a salt, berberine irreversibly binds to the silica gel, allowing it to be easily removed.
11. Scrape off each band of interest with a spatula or razor blade. Some find it easier to do this while the plate is still wet from eluting (you need to work fast), since dry silica dust tends to blow or move around more easily. There are a couple of different methods to retrieve your sample at this point: placing the silica gel in a pipet with a cotton plug, and washing with polar solvent is often sufficient; another option is to sonicate the silica gel in a polar solvent or one which causes the compound to elute fast on TLC, then filter through a frit and concentrate the purified compound.

For an outstanding practical perspective on Prep TLC see “Giving Preparative Thin-Layer Chromatography Some Tender Loving Care” by Hayward, Mader and Trant (Department of Chemistry and Biochemistry, University of Windsor).