Our Methodology

How we work the numbers out

My grandad never bought a bag of compost more than he needed. He would pace the bed, scratch a number on the back of a packet, and that was that. No second trip, no leftovers going hard at the back of the shed.

This page is that habit, written down. Every calculator on this site produces a number, and every number comes from somewhere. Below is where from, how we work it out, and the places where we are honest that we are estimating rather than measuring.

If you find something here that is wrong, tell us. There is a contact page and we read everything that comes through it.

Frost dates: we calculate these ourselves

Most UK frost tools quote a figure. We build ours from the raw observations.

The source is HadUK-Grid, the Met Office gridded climate dataset, version 1.3.2 at 5 km resolution, downloaded from the CEDA archive. It is published under the Open Government Licence v3.0, which permits commercial use. The dataset DOI is 10.5285/bbca3267dc7d4219af484976734c9527.

We pull daily minimum temperature for every year from 1991 to 2020. That is 360 monthly files. 1991 to 2020 is the standard climate normal period used by the World Meteorological Organization, and it is the baseline the Met Office publishes against, so it is a figure you can check rather than one we picked.

For each postcode district we find the nearest land cell in the grid, read 30 years of daily minimums, and work out the last spring frost and first autumn frost for every one of those years. That gives us 30 answers per district, not one.

Why we give you odds instead of a date

A last frost date is not a date. It is a spread. Some years it is early March, some years it is May, and the difference decides whether your tomatoes live.

So we publish three figures for each district:

• Typical — the middle of the 30 years. Half the years were earlier, half were later.
• Safe 9 years in 10 — wait until this date and you would have been clear in 27 of the last 30 years.
• The worst we saw — the latest spring frost in the whole 30-year record.

A gardener needs to know the risk they are taking. One date pretends there is no risk.

Air frost and ground frost are different things

Air frost is when the air temperature drops to 0 °C or below. That is what nearly every frost tool publishes.

Ground frost happens on clear, still nights when heat radiates away from the soil surface while the air a metre up stays above freezing. Ground frost is what kills tender seedlings. A tool that only tells you about air frost will tell you it is safe to plant out on a night that kills your plants.

We publish both.

What we are estimating, and where we are not

This is the part most sites leave out, so here it is plainly.

HadUK-Grid contains observed monthly counts of ground frost days. Those are real Met Office measurements and we use them as published.

It does not contain daily ground frost records, so there is no way to read an observed ground frost date out of it. Our ground frost dates are derived: we count a ground frost on any night the minimum air temperature falls to +2 °C or below, because the ground surface typically sits about 2 °C colder than the air a metre above it on a clear, still night. So the threshold is applied to the air reading we have, not to a ground reading we do not. The observed monthly counts let us check that the derived dates are sensible, but they remain an estimate and we label them as one.

We use the 5 km grid rather than the 1 km grid. Same dataset, same licence, same observations, larger squares. The 1 km daily files run to around 26 GB for a 30-year baseline, which is more than the machine that builds this has room for. We would rather tell you the grid is 5 km than claim a precision we have not paid for.

Where a postcode district has no land cell within 12 km, we publish nothing for it rather than stretch a reading across the sea.

Years with no frost at all are kept in the record, not dropped. Cornwall genuinely has springs with no air frost after January. Throwing those years away would push every figure later and make the whole country look colder than it is.

Materials: volume, density, and the bag it actually comes in

The soil, compost, mulch, gravel and concrete calculators all do the same three things.

1. Work out the volume. Length by width by depth, or the equivalent for a circle or a triangle. This is geometry and it is not up for debate.
2. Convert it by density. A cubic metre of topsoil weighs about 1.4 tonnes, though it runs from roughly 1.2 tonnes dry and sandy up to 1.7 tonnes damp or clay-rich. A cubic metre of fresh concrete weighs about 2.4 tonnes. Different materials, different numbers, and getting this wrong is how people order half of what they need.
3. Round it up to what the shop sells. You cannot buy 1.3 bags. We round up to real pack sizes — 25 litre and 40 litre bags, 400 litre and 850 litre bulk bags, and tonne bags — because the number you need is the number you have to order.

That third step is the whole point of the site and almost nobody else does it.

Where the material figures come from

Specifications come from British Standards. Each one is named by its number on the pages that use it:

• BS 3882:2015 — the specification for topsoil quality: texture, pH, stones and contaminants. Used across the soil and raised bed calculators. It does not set a bulk density, so the weight figure on those pages is a sourced supplier guide rather than a standard, and the page says which.
• BS 8500 and BS EN 206 — concrete mix design, fresh concrete density, and the 90 minute window between batching and placing.
• BS 594987 — asphalt laying, used by the tarmac calculator.

Bag sizes and pack weights come from what UK retailers actually stock, checked against live product listings rather than assumed.

Growing advice: sources, not opinions

Sowing windows, seed rates, feeding rates and plant spacing are cited to the body that publishes them. In practice that is mostly the Royal Horticultural Society, which is referenced on around 40 pages across the site.

Where a question is regulatory rather than horticultural — fence heights, outbuildings, hosepipe restrictions, waste rules — we cite GOV.UK and the Planning Portal directly. We also use the Woodland Trust, the Environment Agency and DEFRA where they are the right authority.

The RHS reorganised their advice URLs in early 2026 and a lot of sites are still linking into dead pages. We keep a migration table and re-check every cited link; a link that does not resolve fails our own compliance check before the page can go live.

One thing worth saying straight. I am not a gardener. I am ex-Army and I build the numbers. Nothing on this site is me telling you what worked in my garden. It is the published guidance, cited, with the arithmetic done properly on top.

What we do not know

No site like this is ever finished, and pretending otherwise is how you end up trusting the wrong number. The honest limits:

• Climate figures describe 1991 to 2020, not next week. They are a 30-year average, not a forecast. A warming climate means recent years trend milder than the middle of that baseline.
• Your garden is not the grid square. A frost pocket at the bottom of a slope, a south-facing wall, or a city centre can shift the real answer by a week or more in either direction. We give you the district; you know your own plot.
• Ground frost dates are derived, not measured. Covered above, and worth repeating here.
• Prices move. Costs shown are checked against UK retailers and dated, but they go out of date between checks. Treat them as a budgeting guide, not a quote.
• Soil and material densities vary. Wet topsoil is heavier than dry. Screened is different to as-dug. We use the standard figure and tell you which one.

How often we check

The frost dataset is rebuilt from source when the Met Office publishes a new HadUK-Grid version, and the build is deterministic, so running it again on the same inputs gives the same answer.

Cited links are re-checked when a page is rewritten, and any page carrying a link that no longer resolves cannot pass our own publishing checks. Prices are re-checked on a rolling basis and carry the date they were last looked at.

Every calculator change goes through a review before it reaches the site, including a check that the arithmetic on the page agrees with the arithmetic in the code.

Tell us when we are wrong

If a number here does not match what you measured, or a source has moved, or something simply does not add up, we want to know. That is not politeness. A calculator that is quietly wrong is worse than no calculator, because you will buy from it.

The contact page comes straight to us.