How the two differ in kind
Ask what the floor is missing. If it is HEIGHT — the structural deck is well below the finished level, there are underfloor pipes that need cover, there is an acoustic requirement that wants mass — the answer is a screed, laid at tens of millimetres and levelled to a datum. If it is FLATNESS — the deck is at the right height but wanders by a few millimetres over a couple of metres, enough to telegraph through a thin floor covering — the answer is a compound that flows out and finds its own level.
The two fail in opposite directions when swapped. Compound poured deep is expensive per unit volume, shrinks and cracks beyond its stated maximum thickness per pass, and is being asked to do a structural job it was not formulated for. A screed laid thin does not achieve its strength, curls at edges and debonds, because a sand-cement screed below its minimum thickness has nothing to hold it together.
The constraint that catches most projects is not either of those. It is DRYING. A traditional sand-cement screed dries at roughly a day per millimetre for the first fifty millimetres and considerably slower beyond that, in good conditions — which means a seventy-five millimetre screed is a two-to-three month item on the programme, not a two-week one. Laying a moisture-sensitive covering over a screed that has not dried is the classic failure, and it is not detectable by touch or by how long ago it was poured.
The factors that actually differ
Show
| Sand-cement or anhydrite screed | Self-levelling compound | |
|---|---|---|
| What it supplies | Depth. Cover over pipes, a build-up to a datum, mass for acoustic separation. | Flatness. A few millimetres of correction over a substrate already at the right level. |
| Typical thickness | Tens of millimetres, with a minimum below which it does not work — bonded, unbonded and floating screeds each have their own. | A few millimetres, with a maximum per pass. Deep-fill products extend it, at a price. |
| Drying time before a covering | The dominant programme item. Roughly a day per millimetre for the first fifty in good conditions, slower after — and slower again if the building is not being ventilated and heated. | Hours to a day or two, which is the reason it exists on refurbishment programmes. |
| Substrate preparation | Depends on whether it is bonded, unbonded or floating — and a bonded screed's failure is almost always a preparation failure. | Unforgiving. It must be primed, sound and clean, and compound over a dusty or contaminated substrate delaminates as a sheet. |
| Underfloor heating | Gives cover over the pipes and thermal mass. Slow to respond, and run continuously with weather compensation. | Used over low-profile systems, where the point is fast response — much less mass, much less output. |
| Flatness achieved | Whatever the screeder achieves against the datum, which is a workmanship result and specified as a tolerance class. | Flat by physics. The material finds its own level, which is exactly what it is for. |
| Cost shape | Cheap per unit volume, expensive in programme. The material is sand and cement; the cost is time. | Expensive per unit volume, cheap in programme. The material is engineered; the cost is bags. |
| What goes wrong | Laid too thin, dried too fast at the surface, curling at edges, or covered before it is dry. | Poured deeper than the product allows, applied over an unprimed or contaminated substrate, or used to correct a level problem it cannot reach. |
Which one, and when
Choose sand-cement or anhydrite screed when…
- There is a real height to make up — a structural deck well below the finished floor level.
- Underfloor heating pipes need cover and the system is intended to run with thermal mass.
- There is an acoustic requirement that wants mass or a resilient layer beneath a floating layer.
- The programme has room for drying, or the screed is going in early enough that drying is not on the critical path.
Choose self-levelling compound when…
- The substrate is at the right height and simply is not flat enough for the covering going on it.
- The programme has no room: a refurbishment where the floor has to be ready in days.
- The build-up cannot grow — thresholds, existing skirtings, a stair riser that would become non-compliant.
- The floor covering is thin and unforgiving, where every undulation in the substrate will show through.
Now run your own numbers
This page holds no prices on purpose — a national average is wrong for almost every real project. Quantify both options with your dimensions and your local quotes.
Frequently asked questions
- How do I know when a screed is dry enough to cover?
- By measuring, and by one of the recognised methods rather than by judgement. A hygrometer box or an in-situ relative humidity probe left in place for the specified period gives a reading that the flooring manufacturer's limit can be compared against, and that reading is the only acceptable evidence. Surfaces feel dry long before the body of a screed is, and the day-per-millimetre rule is a planning guide for a well-ventilated, heated building rather than a test. This matters because the consequence falls on the floor covering rather than the screed: adhesive fails, resilient sheet blisters, timber cups, and the remedy is lifting a finished floor. A surface damp-proof membrane is the recognised route when the programme will not wait, and it is a deliberate specification rather than a rescue.
- Can I build up thirty millimetres with self-levelling compound?
- Not with an ordinary one. Standard compounds have a maximum thickness per pass in the low tens of millimetres at most, and pouring deeper produces shrinkage cracking, weak material and an expensive mistake. Deep-fill and aggregate-loaded products exist and reach further, and they cost considerably more per unit volume than a screed would — so at that depth the question is genuinely whether a screed is the right answer instead. The other route, where the height is awkward but the programme is tight, is a proprietary fast-drying screed: it costs more than sand and cement and dries in days rather than months, which is a legitimate trade when the programme is the binding constraint.
- Anhydrite or sand-cement?
- Anhydrite — calcium sulphate — screeds flow, so they are laid faster, self-level to a good tolerance, can go thinner over underfloor heating and dry faster than sand-cement. In return they bring three constraints that catch people. They form a surface LAITANCE that must be removed by sanding before anything is bonded to them, and skipping that is a common debonding failure. They are not compatible with cement-based adhesives without the correct primer, because the sulphate reacts with cement. And they must not get wet in service, which rules them out of bathrooms, wet rooms and anywhere with a real risk of flooding. Sand-cement is slower, more forgiving of misuse, and indifferent to water.
- What is laitance and why does it matter?
- A weak, fine layer that forms at the surface of a screed as the finest particles and water rise during placing and curing. It looks like the screed and is not: it has very little strength, so anything bonded to it is bonded to a layer that will shear off. On anhydrite screeds it is pronounced enough that removal by mechanical sanding is a specified operation rather than good practice. On sand-cement screeds it is less severe but still present where the mix was wet or over-trowelled. The test is simple and worth doing: scratch the surface firmly with a coin or a screwdriver, and if it powders off you are looking at laitance rather than screed.
- Can I turn the underfloor heating on to dry the screed faster?
- Only as a controlled commissioning cycle, and only after the screed has cured for the period the screed manufacturer specifies — commonly a week or more before any heat at all. The cycle then starts at a low flow temperature and steps up over days, holds, and steps back down, with the whole sequence recorded. Turning a system straight on to speed the programme cracks the screed, because it heats a material that is still gaining strength and creates thermal stress in something restrained by its own bond. The commissioning record matters afterwards too: flooring manufacturers commonly require evidence that the cycle was carried out before they will warrant a covering laid over it.
- Which one goes under which?
- Screed first, compound second, and the compound is often not needed at all. A screed is laid to a datum and achieves a tolerance class; where that tolerance is not tight enough for the covering — a thin resilient sheet, a large-format tile, a poured resin — a thin compound over the cured screed brings it to the flatness the covering needs. Doing it the other way round makes no sense: compound is not a substrate for a screed. And if the specification calls for both as a matter of course, it is worth asking whether the screed tolerance was specified tightly enough in the first place, because achieving flatness in the screed is cheaper than achieving it twice.