There is no single best fill material. I say that as a person whose job is to approve or reject shipments of it for a living.
When I moved into quality and brand compliance at a natural stone supplier, I assumed the hard part was holding size tolerance. Match the sieve range, match the spec, ship it. Roughly 1,200 lot reviews later, I know that's about a third of the story. The material that gets rejected isn't usually the wrong size. It's the wrong material for what the buyer was actually trying to solve.
So before I give anyone a recommendation, I sort the request into a scenario. Four of them. Most buyers land in one of the first three in under a minute.
- It has to be looked at. Flat glass marbles, glass sand, polished river pebbles — display media, surface finish, tabletop.
- It has to hold ground. Outdoor patio gravel, large garden pebbles — drainage, weed suppression, foot traffic.
- It has to grow something. Perlite in compost and potting mixes — aeration and root health.
- It's a trial. A sample box, a first small order, one floor of one hotel. Not a material — a stage.
I said four and then cheated on the last one. Sorry. It's on the list because it changes the specification more than any of the others, and it's the bucket this industry handles worst.
Scenario 1: The material has to look right
If your final judge is a guest, a stylist, or a hotel inspector, you're in this bucket. The spec that matters isn't "8–12 mm polished pebble." It's the same 8–12 mm polished pebble in eleven months, when you reorder.
Two things kill this category, and neither is the one most people watch for.
Dust, not chips. Unwashed stone looks identical to washed stone in the bag. It does not look identical in a vase, or on a lit surface, or under a camera. Fines settle, water goes cloudy, and under warm hospitality lighting the whole installation reads gray. Ask your supplier for the percentage of material passing the No. 200 sieve (ASTM C117). That number predicts cloudiness far better than any photo will. "Washed" with no figure attached is basically a marketing term.
Here's the thing: I used to think washing was an upsell. Then we ran an unwashed batch into 400 hotel vases in 2021 and every single one had to be emptied and redone on site. That's a mistake you only need to make once.
Color drift, batch to batch. Natural stone moves. Quarry faces change. A shade shift you can't detect between two bags becomes glaring across twelve display tables. The fix is boring: retain a golden sample from the approved lot, keep the lot number on file, and require sign-off against it. A supplier who can't produce a retained sample on a reorder isn't managing consistency — they're hoping.
A quick note on glass versus stone
Flat glass marbles and glass sand sit around 5.5 on the Mohs scale. Marble and calcite sit around 3. Granite runs 6–7. Practically, glass media holds its surface longer under handling, but it's harder on bare feet and bare hands at the edges. Glass sand is usually recycled and more angular than people expect. If it's going anywhere people will kneel, walk, or reach into, ask about edge condition, not just mesh size.
Scenario 2: The material has to hold ground
Outdoor patio gravel and large garden pebbles. Now the question isn't "does it look right" — it's "will it stay where I put it and drain the way I promised."
The most common specification mistake I see is rounded river pebbles on a slope or a path. Rounded stone rolls. Crushed, angular stone locks. If the install has any pitch to it, or foot traffic, or a dog, shape matters more than color.
Second: fines content again, but with the opposite failure mode. In décor, fines make things cloudy. In ground cover, fines fill the void space between stones and turn a draining layer into a slick, compacted crust. Ask for gradation data and a washability figure. The same stone at 5% fines and at 15% fines behaves like two different products.
Depth, from the specs I review: base prep is its own conversation, but decorative cover over a compacted base with a weed barrier generally lands in the 2–4 inch range. Under 2 inches and you're seeing the barrier through it. Much past 4 and you're just paying to bury your own drainage.
And if any of that stone is fixed rather than loose — caps, step treads, a seating edge — check the slip rating for wet conditions. ANSI A326.3 sets the minimum dynamic coefficient of friction at 0.42 for level interior wet surfaces, and plenty of outdoor projects borrow the same logic. Polished, wet, and walked on is a bad combination no matter how it photographs.
Scenario 3: The material has to grow something
Adding perlite to compost is where online advice gets genuinely bad, mostly because it gets stated as a ratio instead of a reason.
What perlite actually is: volcanic glass ore heated to roughly 1,600°F (871°C), which pops it like popcorn to 4–20 times its original volume. What you buy is mostly air sealed inside a thin glass shell. It's near-neutral — published pH ranges cluster around 6.6–7.5 depending on the source rock — and it doesn't break down in the pot. That combination is why it aerates.
Extension guidance for general container mixes typically lands around 20–30% perlite by volume. Succulents, cuttings, and anything you're actively trying to dry out run higher. Treat that range as a starting point, not a law.
Here's the counterintuitive part. More perlite is not automatically better. Perlite floats. In a mix that's already coarse or already bark-heavy, adding more can give you a segregated profile where water channels down one side of the pot and roots on the other side stay dry. And in most of the waterlogged containers people blame on "not enough perlite," the real cause was compaction and watering frequency. Adding perlite to a mix you're overwatering just means you're overwatering something lighter.
One safety point that gets skipped constantly: perlite dust. Perlite ore can contain crystalline silica, and NIOSH's recommended exposure limit for respirable crystalline silica is 0.05 mg/m³ as an 8-hour TWA. Dampen the bag before you open it, mix outdoors or with real ventilation, and wear at least an N95. It's a five-second habit with an actual payoff.
For comparison's sake, vermiculite holds water where perlite sheds it. Neither is better. They do different jobs, and plenty of good mixes use both.
Scenario 4: You're ordering a trial, and someone is about to treat you like a nuisance
This is where I get opinionated.
Small orders get handled badly in this industry. A 5 kg sample box gets pulled off whatever is open on the floor, labeled "assorted," and shipped with no lot number. The buyer commits to a real order based on it. The real order doesn't match. I've watched that exact sequence cost a supplier a client they'd spent a year courting.
When I set up our verification protocol at Mikasa in 2022, I pushed for one rule that people argued with me about: same inspection on a sample box as on a pallet. Retained reference sample, wash result, sieve data, lot number on the label. It costs about twenty minutes per shipment.
Look — I'm not going to pretend a 25 kg order prices like a 25 ton order. Volume pricing is real economics, not discrimination, and anyone who tells you otherwise is selling you something. What shouldn't scale with order size is the service.
A landscape designer who ordered three sample boxes from us in early 2023 now specifies us for roughly $40,000 a year in project work. She didn't get there because we were the low bid. She got there because the third box matched the first one, and she noticed.
If you're a small buyer, ask for four things: a retained reference sample, a wash or fines figure, the lot number, and written confirmation of what happens if the delivery doesn't match the sample. If a supplier can't answer all four on a small order, that's information too.
How to tell which scenario you're actually in
You probably know within a minute. Three questions get you there.
What happens if it fails? If the failure is visible — cloudy water, mismatched color, an installation that reads cheap on camera — you're in Scenario 1. If the failure is structural — pooling water, weeds pushing through fabric, stone migrating off a slope — Scenario 2. If the failure is a dead plant, Scenario 3.
Who's the judge? A guest, a client, or a photograph means consistency is the spec. A drainage bed, a path, or a truck tire means gradation and shape are the spec. A root system means air and water movement are the spec. Those three criteria barely overlap, which is exactly why one "best material" answer doesn't exist.
Will you reorder? If yes, lock the lot and pay for consistency. If no, optimize for the one delivery in front of you and stop paying for consistency you won't need.
And if you're not sure yet what you need, don't buy the ton. Buy the sample. A trial order is a legitimate way to make this decision, and any supplier worth keeping should treat it that way.
Pick the scenario first. The material basically picks itself after that.
