How to Improve Garden Soil: Depth, Water, Then Food

The garden that prompted this guide sits on a subdivided smallholding portion near Simondium, in the Drakenstein Valley, with about 40 square metres of beds wrapped around the north and west sides of a four-year-old house. The owner had replanted those beds twice in two seasons and lost most of both, and had spent somewhere around R4,000 on plants and bagged compost doing it. The working theory was that the soil needed feeding.

We put a spade in. It went 140mm and stopped dead.

Soil is fixed in one order and only one: depth first, then water, then food. Almost everybody starts at food, because food is the only one of the three you can buy in a bag and carry home in a car, and it is the one that fixes the least. Compost is the answer to exactly one of the three problems, and on the beds we are called to rescue it is usually not the one the garden has.

There is a second half to that, and it is the part that changes how you spend money. On our hot, fast-draining, winter-wet soils, whatever organic matter you dig in is broken down and gone within about three years. The 50 to 75mm layer you keep on the surface ends up doing more work than anything you bury. So this article is about diagnosing a bed before you spend on it, and about where the material actually goes once you do.

Why good soil still fails here

Bad soil is not a diagnosis. In seven years of Winelands gardens there are three real failures, and they want three completely different fixes: the soil is too shallow for what you planted in it, the soil will not hold water where the roots are, or the soil is genuinely hungry. Only the third one is solved by buying something.

One consistency note, because we say the opposite elsewhere on this site and both are true. Under paving the whole job is to compact the ground so it carries load. In a bed the whole job is to loosen it so roots can get through it. We compact what carries and we loosen what grows, and the two are never the same piece of ground.

Know what is under your garden

Four kinds of ground, four different jobs, and knowing which you have saves you buying the wrong thing.

• ✓The decomposed granite valley floor around Paarl, Pearl Valley and Stellenbosch. Fast draining, coarse, and permanently low in organic matter because it mineralises fast under a hot dry summer and washes through under a wet winter. Forgiving to dig, unforgiving to leave bare. This is the ground everything below is calibrated to.
• ✓The richer valley soils around Simondium and the warm, fertile Wellington side. Genuinely good soil, and the point worth making is that good soil fails on depth and on water exactly as readily as poor soil does. That is why the worked example in this article is where it is.
• ✓Franschhoek, higher and wetter with more winter rain and the occasional frost. The risk that catches people there is winter waterlogging rather than summer drying, so beds want raising and draining more than they want feeding.
• ✓The new-build platform, which is not a soil at all and gets its own section further down.

If you want to know how fast your ground drains, our guide to water-wise and indigenous planting has the hole-and-water test, and we will not repeat it here.

Step 1: Diagnose the bed before you buy anything

Three tests, ten minutes, and the only equipment is a spade and a watering can.

1. 1The spade test, for depth. Push a spade in vertically with normal foot pressure and note where it stops. If it stops shallower than about 200mm, you have a depth problem. Then dig one hole right the way through, because a pan, a builders platform, buried rubble and natural clay all feel similar underfoot and all want different answers.
2. 2The wet-and-dig test, for water. This is the one nobody publishes and the one that changes the most minds. Water the bed the way you normally water it, wait an hour, then dig a small hole and look at where the wet stops. Wet in the top 40mm with dry soil underneath means the water is not going in at all. Water still sitting on the surface after an hour means the opposite problem.
3. 3The handful test, for structure and life. Take soil from about 200mm down, wet it slightly, squeeze it and open your hand. Good soil breaks into crumbs. Sand runs through your fingers. Clay holds a smooth ribbon and takes a fingerprint. Look for worms, fine roots and any smell at all, because dead soil smells of nothing.

On testing: we do not send domestic garden soil to a laboratory, and we do not chase a pH number. On an ordinary bed the result arrives weeks later, costs more than the compost would have, and tells you about chemistry when the problem is almost always physics. A spade and an hour tells you more. Turf is a different matter with its own thresholds, and our lawn rescue plan covers that diagnosis separately.

Step 2: Fix the depth first, because nothing works above a pan

How deep the soil needs to be workable depends on what is going into it.

• ✓Shrubs and perennials want soil a spade can enter to about 300mm.
• ✓A hedge or a small tree wants about 400mm.
• ✓Annuals and groundcovers will manage on 200mm, and will need watering all summer to do it.

Then how to get there. We fork, and we do not rotavate, and the reason is specific rather than dogmatic. A rotavator works to a fixed depth and polishes a new smooth pan at the bottom of its own cut, usually around 180mm. On ground that already has a pan at 140mm you have simply replaced one problem with a tidier one. A fork lifted and rocked rather than turned breaks the pan without inverting the profile, and keeping the profile the right way up matters, because the biology lives in the top 100mm.

What the spade is hitting, and what each one wants:

• ✓Compaction from foot traffic, mowers or machinery. Forks out, and stays out if you keep the bed covered.
• ✓A builders platform. Deliberate, dense, and usually wants a pick or a mattock through it before a fork will do anything.
• ✓Buried rubble. Comes out. Breaking it up smaller and leaving it in is not a fix.
• ✓Natural clay. Not a pan at all, and handled in its own section below.

Be honest about the limits. Forking 40 square metres of beds through by hand is two people for the better part of two days, and it is the single most worthwhile day of physical work in any garden, so do it yourself and put the money into materials. Levelling a new-build platform, importing twelve cubic metres and carting spoil away is a machine and a team, and that is where our landscaping crew earns its keep. Compaction on an established lawn is a different job again, covered in our lawn rescue plan.

Step 3: Fix how the soil handles water

This is the step almost nobody knows about, and it explains more dead plants in Paarl than anything else on this page.

Bare soil in a Paarl February bakes into a crust and turns water repellent. Water beads on it, runs to the low corner of the bed and takes the fertiliser with it. The bed can be genuinely dust dry 100mm down an hour after you watered it, and the plant looks thirsty because it is thirsty, and so the owner waters more, and almost none of that reaches the roots either. This is not a myth and it is the most commonly misread symptom in a Winelands garden.

Three fixes, in order:

1. 1Re-wet it slowly. Short repeated runs over an evening or two, rather than one long soak. Dry repellent soil has to be reintroduced to water gradually, and a single long run mostly runs off.
2. 2Get organic matter into the top layer, so the crust cannot form in the first place.
3. 3Cover it. That is step 5, and it is the permanent answer.

A wetting agent off a shelf will get water in this week and will do nothing for you next February. It is a tool for a specific rescue, not a strategy.

Then the watering principle that follows from all of this: deep and infrequent, so roots chase moisture downward. Short daily runs are the most reliable way there is to build a shallow-rooted, fragile bed. Our guide to keeping a Paarl garden green covers the local rhythm, and the garden watering calculator works out session length and litres for your own area, so we will not repeat those numbers here.

One practical note worth its space: put drip line under the mulch, never on top of it. A dripper lying on bark waters the bark.

Step 4: Feed it last, and know what you are actually buying

Food, third, and only now. Sometimes a bed genuinely is hungry, and when it is, compost is the right answer and it works well. The problem is never compost. The problem is compost bought for a depth or a water problem.

• ✓Compost, for a bed that grows things and needs feeding. Good compost is dark, crumbly, smells of earth and is cool to the hand.
• ✓Topsoil, only when you are physically short of soil and need to fill or build up. Never used on its own, because topsoil bought by the load is very often subsoil with a good name: pale, structureless, and full of stones.
• ✓Kraal manure, well composted and used sparingly on vegetables. Poor value on ornamental beds, and never anywhere near fynbos.

Then the three-year argument, which is this article's thesis paying out. On hot, fast-draining, winter-wet soil, the organic matter you fork in is broken down and gone faster than most gardeners imagine. A one-off dig-in is a deposit, not an endowment. That is not an argument against digging it in. It is an argument that the dig-in is the smaller half of the job and step 5 is the larger half.

One opinion that contradicts what most people are told: do not improve a tree planting hole. A rich hole in poor ground becomes a pot. The roots circle inside it, find the wall of native soil, and the tree never leaves home. Plant a tree into its own soil, backfill with exactly what came out, and put the compost and the mulch on the surface over a wide circle instead. The same logic makes fynbos, proteas and ericas a special case, covered at the end, and our water-wise planting guide has the detail.

How deep to work the ground is this article's job. How much material that comes to, in cubic metres, bags and bakkie loads, belongs to our compost, soil and mulch calculator.

Step 5: Cover it, and keep it covered

If you do one thing in this whole article, do this one. A 50 to 75mm layer of bark, wood chip, straw or even gravel over the soil does more for a Winelands bed than anything you can dig into it, and it keeps doing it after the dug-in compost has gone.

• ✓It stops the crust forming, which is what caused the water problem in step 3.
• ✓It keeps the soil cool and damp, so the biology that builds structure stays alive through February.
• ✓It feeds the soil slowly from the top down as it breaks down, which is how soil is built in nature and nowhere else.
• ✓It suppresses the weeds that would otherwise be drinking your water.

Keep it off the stems. A mulch ring piled against a trunk holds moisture against bark that wants to be dry, and that is how you rot a young tree at the collar. Leave a hand's width clear around every stem.

Mulch is not permanent, and this is the part people miss. It breaks down, which is the point of it, so it wants topping up roughly once a year going into spring. That annual top-up is the cheapest and highest-return hour in the gardening year, and it is one of the standing jobs on our garden maintenance rounds. If we have done tree work on your property we can chip the branches into mulch on site, which is the cheapest mulch there is.

What the materials cost

Worked through, making and mulching a new bed comes to roughly R90 to R160 per square metre in materials as a rough 2026 guide. That is worth holding next to the all-in planted figure of R150 to R500 per square metre in our landscaping ideas guide, because it tells you something uncomfortable: the groundwork is the cheapest part of a new bed, it is the first part people cut, and it is the only part that cannot be added later without lifting the plants back out.

On bags against bulk, the short version is that bagged material is convenience pricing and the crossover comes quickly, so anything over about a cubic metre wants a delivery. The compost calculator will tell you which side of that line your job falls on.

Soil for a lawn is a different job

Three differences that matter. A lawn is a surface, so you get exactly one chance at the soil under it and then it is closed for twenty years, which is the opposite of a bed you can improve from the top every autumn. Everything above about loosening applies before the turf goes down and almost none of it applies afterwards, because once there is grass on it the only routes into the soil are aeration and top-dressing. And the bed tests in step 1 do not transfer, because turf has its own diagnosis and its own thresholds.

So: our instant lawn service prepares the ground as part of the job, and that preparation is where most do-it-yourself lawns quietly fail. Compaction, aeration and thatch on an existing lawn belong to our lawn rescue plan. Top-dressing depth, quantities and cost belong to our lawn dressing guide.

Clay, and what actually works on it

First, narrow it honestly, because most Winelands gardens are not clay. If the wet-and-dig test left water sitting rather than running off, and the handful test gave you a smooth ribbon that took a fingerprint, this section is yours. Otherwise skip it.

1. 1Raise the bed. This is the first and best answer, because you are giving roots depth above the problem instead of trying to change the problem. Two hundred to 300mm of raise over a broken surface does more than any amendment will.
2. 2Add organic matter, slowly and repeatedly rather than in one go. It genuinely does improve clay structure, and it works from the surface as much as from digging.
3. 3Match the planting to the ground you have, which is what our water-wise planting guide recommends and it is right.

Three straight opinions, because this subject attracts more bad advice than any other in gardening. Never dig sand into clay: the ratio you would need to actually change the texture is far beyond what anyone will buy, and anything short of it makes the problem worse. Gypsum is sold as a clay-breaker and it only does anything on sodic soils, which is not what most Winelands gardens have, so treat it as a specific remedy rather than a general one. And never dig a deep hole in clay and fill it with compost, because you have built a sump that fills in June and drowns the plant. That is the single most common way we see clay kill a shrub.

Franschhoek sits higher and wetter, so winter waterlogging is a live consideration there in a way it is not on the Paarl valley floor.

New-build ground, and what the builders left behind

A new stand is the purest possible case of a depth problem. The topsoil was stripped and sold, or spread thin at the end. The whole site was compacted by machines, because that is exactly what a building platform is for. And the garden is often 100 to 150mm of imported material sitting directly on that platform.

The specific finds, worth naming because everyone who has one recognises it: a band of sand and cement washout where the mixer stood, a buried line of offcut brick and plaster, and skim-coat lime through the top layer along the walls. We have taken 200mm of builders rubble and topsoil out of ground that looked like a lawn, which is the same find our paving guide warns about from the other direction.

The diagnostic is the same spade test with a different reading. On new ground the spade often goes through the imported layer easily and then stops abruptly. That clean stop is the platform. The trap is that a new-build garden is also the one where the owner is most likely to be told the soil is fine, because somebody spread 100mm of dark material over it and it looked right on handover day.

What to do: break the platform before anything is planted or turfed, get the rubble out rather than burying it deeper, and be sceptical about any load of topsoil that arrives pale, structureless or full of stones. This is the point where hiring it out is genuinely the right call, and Klapmuts is where we do most of it. Our landscaping team handles the machine work and the spoil.

The five checks, and when to do each one

Cheap, quick, and each one has a window that closes. Written as much for someone who has hired the job out as for someone doing it.

1. 1Before anyone quotes. Dig one hole to full depth in the worst corner and photograph the side of it. That hole determines the whole specification, and it is the only thing that will settle a disagreement later.
2. 2Before anything is planted. Put a spade in across the bed and check it reaches 300mm without real effort. After planting you will never get to do this again without lifting everything out.
3. 3On the delivery. Look at what came off the truck. Compost that steams or smells sour is unfinished, and topsoil that is pale and structureless is subsoil. Both are far easier to refuse than to remove.
4. 4An hour after the first watering. Dig a small hole and check where the wet stops. A bed that is dry at 100mm an hour after watering has a water problem that no amount of further watering will solve.
5. 5Next autumn. Measure the mulch. If it is under about 35mm the beds are running on borrowed time, and a top-up is the cheapest intervention available to you.

A Simondium bed, rebuilt

Back to the job at the top. About 40 square metres of beds around a four-year-old house on a subdivided smallholding portion near Simondium, replanted twice and lost twice, with roughly R4,000 spent on plants and bagged compost along the way.

What the three tests found. The spade went in about 140mm across the whole north bed and stopped dead on a compacted layer left behind by the building platform. After a twenty minute run with a hose the surface was wet in the top 40mm and dust dry at 100mm, with most of the water having run to the low corner of the bed rather than into it. And soil brought up from below the pan came out dark, crumbled into aggregates, and had worms in it.

That last finding is the whole point. The soil under that garden is genuinely good, exactly as the Drakenstein valley's reputation suggests. No root had reached it in four years.

What we did. Forked the whole 40 square metres through by hand to about 300mm, breaking the pan rather than turning the profile over, and lifted about 0.4 cubic metres of builders rubble and one length of buried conduit out of it. No rotavator, because on a pan at 140mm a rotavator just makes a smoother one at 180mm. Then we wet the ground slowly over two evenings before anything was planted, in short repeated runs, because dry repellent soil has to be reintroduced to water. Compost went through the worked layer, bark mulch went on at 60mm, and the drip line went under the mulch rather than over it. Two of us, three days, in March.

Materials came to roughly R90 to R160 per square metre as a rough 2026 planning figure, so somewhere around R3,600 to R6,400 for the 40 square metres. That is less than the plants the owner had already buried. Twelve months on, the mulch had broken down to about 35mm and needed its top-up, a spade went in to about 280mm without effort, and nine of the eleven surviving original plants were still there.

Two lessons. The soil was never the problem, and the Simondium reputation for rich valley soil is perfectly true, which is exactly why compost was the wrong purchase three times running. And the owner spent more on replacement plants over two seasons than the entire rebuild cost in materials, for want of one hole dug in the worst corner of the bed before anything was bought. If you would rather someone else did the forking, that is what our landscaping team is for.

Doing it through the year

Autumn, March into May, is the soil month. The ground is still warm, the killing heat has passed and the rains are coming, so anything worked and planted then roots before summer. It is also the only sensible time to fork, because dry summer ground is punishing to work and saturated winter ground smears and compacts under your own boots, which undoes a whole day of effort.

Winter is for planning, and for watching where the water actually goes on your property, which is free information almost nobody collects. Spring is the mulch top-up, before the dry season lands. Summer is for nothing at all except keeping the cover on, and specifically not for digging, which opens moisture-holding soil to the air on the worst possible day for it.

The one-line rule: work the soil in autumn, cover it in spring, leave it alone in summer. Our Western Cape garden calendar covers everything else the year asks for, and our Paarl water-wise guide has the local rainfall behind the timing.

When not to improve your soil

Five situations where we will tell you to stop, and the first one matters most.

• ✓Fynbos, proteas, ericas and restios. Rich soil is the failure here rather than the fix. Keep it poor and sharp, never add manure or phosphorus, and understand that in this article's own terms food is not third for fynbos, it is zero. Depth and cover still apply, because a fynbos bed still wants mulch. Our water-wise planting guide has the full rule.
• ✓A mature, established tree or shrub that is doing fine. Digging under a canopy destroys the surface roots keeping it alive. If you want to do something for it, mulch wide and thin and leave the spade in the shed.
• ✓Ground about to be paved or turfed with artificial grass. The requirement there is the exact opposite and compaction is the goal, as our paving guide sets out.
• ✓Pots and containers. Not a soil problem at all. They want a bought potting medium, because garden soil in a pot compacts and drowns, which our succulents guide covers.
• ✓Chasing a laboratory number on a domestic bed, for the reasons in step 1.

And the honest downside, which almost nobody publishes. Soft, moist, well-planted, well-mulched ground is exactly what mole-rats are looking for. Improving your soil properly makes your garden more attractive to them, and that Simondium bed saw its first heaps the following winter. It is a trade we would make every time, but you should hear it from us rather than discover it in July, and our mole guide covers what to do about it.

The five steps are really one idea: soil is fixed in the order depth, water, food, and the layer you keep on the surface outlasts anything you bury. If you take nothing else from this, take the three tests in step 1 and the mulch top-up in step 5.

We prepare and rebuild garden soil as part of new gardens and makeovers across Paarl, Stellenbosch, Simondium and the Winelands, and we deliver compost, topsoil and mulch. We are equally happy to tell you that your beds need a fork and a bakkie load of bark rather than a landscaping quote. Send us your measurements for a quote, or see our landscaping service.