Building a Retaining Wall

Two failure modes, and why crews confuse them

A retaining wall does not fail generically. It either rotates forward about its toe, with the base line staying more or less where you put it, or the whole mass translates downhill and the base course walks off the pad. Those are different mechanisms with different arithmetic, and — this is the part that gets walls rebuilt — different remedies.

Overturning is beaten with mass, geometry and tie-back: a wider effective mass, more setback, deeper embedment, geogrid that reaches past the failure wedge. Sliding is beaten with friction and restraint at the base: a properly graded and compacted pad, embedment that gives you soil in front of the toe, and enough vertical load bearing down on the base plane. Adding two more grid layers to a wall that is creeping on a slick pad buys nothing. Rebuilding a pad under a wall that is rotating because the grid stopped short buys nothing either.

Diagnose the mode first from what the wall is actually doing, then spend money. The signatures are distinct enough to read off a leaning face.

Overturning: what drives the rotation

The driver is the horizontal component of active earth pressure acting on the back of the wall mass. It builds as a triangle, zero at the top and greatest at the base, so the total thrust grows with the square of retained height and its resultant sits roughly a third of the height up from the base. Multiply thrust by that lever arm and the overturning moment scales with the cube of height. A wall that goes from three feet to six is not twice the problem, it is about eight times the problem.

Soil type sets how hard the wedge pushes. Clean, angular, free-draining granular fill has a high friction angle and pushes least. Silts and lean clays push harder, hold water, and keep pushing after the rain stops because they drain slowly and shrink and swell seasonally. Backfilling with the spoil that came out of the cut is the most common way a wall gets a driving load nobody calculated.

Water is a separate load stacked on top of that, not a modifier of it. Saturated backfill adds hydrostatic pressure that does not care about friction angle, does not care how well you compacted, and acts on the full retained height. It is also the only load on the job that can arrive overnight.

Overturning: what resists it

Resistance is weight acting through a lever arm measured back from the toe. For a gravity wall that weight is the units themselves plus whatever soil sits on the setback shelf created by the batter. That is why unit depth matters more than unit face area, and why a deep, heavy unit at a modest exposed height outperforms a tall stack of shallow ones.

Batter moves the resultant back over the base and shortens the driving arm at the same time. Every course set back adds a small amount of resisting moment for free, which is why a wall built with the manufacturer's setback and a wall built plumb are not the same wall even though they use identical pallets.

Embedment buries the bottom of the wall so the effective rotation point is below finished grade rather than at the surface, and puts soil in front of the toe. Geogrid changes the problem entirely: the reinforced soil mass behaves as a coherent block whose width is the grid length, not the block depth. A four-foot grid turns a twelve-inch-deep unit into a four-foot-wide gravity structure, and the resisting moment goes up accordingly.

Overturning: what makes it worse

Losing batter is the quiet one. A homeowner asks for a plumb face, or the crew shims the base course level with chips instead of correcting the pad, and the setback that was in the design is gone by course four. Nobody records it and the wall looks fine on handover day.

Grid cut to roll width rather than to design length is the expensive one. Geogrid that terminates inside the active wedge is anchored to soil that is moving with the wall — it holds nothing. The wall behaves as an unreinforced stack and starts rotating at whatever height the block alone cannot carry.

Then the accumulators: backfill that is site clay, a tension crack opening behind the top course and filling with surface water, a cap course laid dry so the top row walks, and freeze cycles jacking a saturated wedge forward a few millimetres per winter. None of those fails a wall by itself. Together they take a marginal design past its factor of safety in the third or fourth season, long after the crew has left.

Overturning: what fixes it

Length the reinforcement past the wedge. The failure surface rises from the heel of the wall at an angle governed by the soil's friction angle, and every layer of grid has to cross it with enough embedded tail beyond to develop pullout resistance. Common practice puts the minimum grid length at a substantial fraction of total wall height — often around sixty percent, more where there is a slope or surcharge above — with the manufacturer's design method and the project engineer governing the actual number. Roll width is a shipping dimension, not a design dimension.

Tighten vertical spacing before you lengthen units. Grid every second course carries a rotating wall better than grid every fourth course of twice the length, because the unreinforced band between layers is what bulges first. Keep the connection at the face intact — grid laid short of the block face, or laid over a course it was never tested with, loses the connection strength the design assumed.

The geometry levers are still there when reinforcement is not an option: deeper units, more setback, more embedment, and removing the surcharge that put the wall over the line. A terraced pair of short walls, properly separated, solves problems that one tall wall cannot.

Sliding: what drives the translation

The driving force is the same horizontal thrust, but resolved differently. For sliding, the lever arm is irrelevant — what matters is the total push against the total friction available on the base plane. That means a squat, wide wall which is nowhere near overturning can still be perfectly capable of walking forward, and it will do it without ever leaning.

Reinforced walls get a second version of the same check. The mass can slide at the bottom of the pad, or it can slide internally along a grid layer, shearing the block above from the block below at the weakest interface. That is why the check has to be run at each reinforcement level, not just at the base.

Sloping ground under the wall makes both worse. A base plane that tilts downhill converts part of the wall's own weight from a stabilising force into a driving one, and a stepped pad built with thin shims instead of full course steps gives the mass a ramp.

Sliding: what resists it

Friction between the bottom course and the compacted pad, and the friction available on every internal interface above it. That friction is the vertical load times a coefficient set by the two materials in contact — which is why anything slick between the block and the pad is a direct subtraction from the resistance, and why a heavier reinforced mass slides less than a light one.

Passive resistance from the soil in front of the toe, which is what embedment actually buys you. It is worth real force, and it is worth nothing the day someone digs a trench along the face for a fence line or drops a new patio grade below the buried course.

Block lips and pins transfer shear between courses and keep the face aligned. They do not resist sliding at the pad interface, and a wall relying on the lip to hold the base course in place is relying on the wrong component.

Sliding: what makes it worse

A pad built on soft or wet subgrade. Excavate to competent material and prove it, because a pad compacted on a spongy bottom compacts the spongy bottom, not the stone. If the subgrade pumps under the plate, it will pump under the wall.

Rounded material in the pad. Pea gravel and rounded river stone roll; angular crushed stone with a graded fine fraction locks. The pad is a friction surface as much as a bearing surface.

Anything sheet-like under the base course. Geotextile or plastic run beneath the bottom units, laid there to keep the pad clean, gives the wall a bearing on a low-friction plane. Wrap the drainage stone and the retained soil interface; do not wrap the wall's own footprint.

And water again — saturated soil in front of and beneath the toe reduces effective stress, cuts friction, and softens the passive wedge exactly when the driving thrust is at its peak.

Sliding: what fixes it

Key the wall in. Embedment is normally specified as a full buried course as a minimum, with more required on a slope, in front of a descending grade, or where the toe soil is weak — a common figure is a fraction of exposed height, and the design method or the engineer sets it. Step the pad on grade changes in full course increments so every step is level, then re-establish the string and check it before the second course goes on.

Build the pad properly: excavate wider than the units by a margin each side so the wall never bears on the shoulder, place a compacted crushed-stone pad of the specified thickness, and compact it to refusal in lifts. A pad that is thick, angular and dense is doing three jobs — bearing, friction, and drainage under the base course.

Where friction is short, add mass. Extending geogrid increases the vertical load on the base plane as well as the width of the mass, which is the one remedy that helps both failure modes at once. A toe berm helps too, and is far cheaper than the trench somebody was about to cut in front of the wall.

Checking both modes before the pallets land

Both checks come out as ratios: resisting moment over overturning moment, and resisting friction over driving thrust. Common practice looks for a comfortable margin on each rather than a bare pass, with bearing pressure checked alongside them; the governing values come from the design method in use and from the jurisdiction, so confirm rather than assume.

Run them on the wall you are actually going to build — the real retained height including the buried course, the real backfill you can source, the real surcharge, and the drained case and the saturated case both. The saturated case is the one that tells you whether the drainage is a convenience or a structural component.

If the numbers come back marginal, the cheapest fixes are geometric and get made before anything is ordered: more embedment, more batter, a deeper unit, longer grid, or splitting one tall wall into two terraces with the separation the design calls for.

When the check does not pass with any of that, the answer is a designed wall, not more blocks.

The levelling pad and the base course

The first course decides the wall. Every deviation in level is carried upward and amplified by the batter, every gap under a unit becomes a point load, and a base course laid out of level along its run cannot be corrected later without pulling the wall apart. Set it level block to block and along the string, check each unit front to back and side to side, and seat units by tapping them down into the pad rather than shimming them up off it.

Pad geometry is fixed before the first block: excavate to competent subgrade, place the specified thickness of compacted angular crushed stone, and carry it wider than the unit on both sides so the wall is never bearing on a disturbed edge. Segmental units themselves are covered by ASTM C1372, Standard Specification for Segmental Retaining Wall Units, which is what you are buying against — it says nothing about how well you set them.

Embedment is part of this course, not an afterthought. Burying a full course means the block count includes courses no one will ever see, and it means the exposed height the client describes is not the height the wall is designed for. Settle embedment and batter first, then count.

Water: the load nobody budgets for

Every wall is designed drained. The pressure numbers assume the wedge behind the face is not holding water, and the entire margin of safety depends on that being true. Block the drain and the wall picks up a full hydrostatic load it was never checked against, at the moment it is least able to take it. That single change can push a wall from comfortably stable to failing without anything else on the job being wrong.

So the drainage has to be built as a system, not a habit. A continuous chimney of clean, open-graded stone immediately behind the units for the full height. A perforated collector at the base, laid to fall, daylighted at intervals and at every low point — and outlets marked, because an outlet nobody can find is an outlet nobody will clear. Filter fabric between the drainage stone and the retained soil so fines cannot migrate in and blind the aggregate; the stone stays clean or it stops being drainage.

Cut the water off above as well. Grade the surface behind the wall to shed away, cap the top of the backfill with a low-permeability layer, and never route a downspout or a swale into the reinforced zone. Freeze-thaw does the rest of the damage where drainage is marginal: saturated fines behind the face expand, walk the courses forward a fraction each winter, and open the joints that let more water in the following year.

Surcharge: the load that arrives after you leave

Anything sitting on the retained soil adds pressure over the full height of the wall rather than just at the top, which means surcharge feeds overturning and sliding equally. A driveway, a parking bay, a slope rising behind the wall, a pool deck, an upper terrace too close to the lower one — all of them change the design, and most of them are visible before the job starts if anyone asks.

The ones that catch crews out arrive later. A vehicle parked behind a wall built for a lawn. A stockpile of screenings dropped on fresh backfill during construction, which is a construction surcharge applied to a wall whose grid has not yet been covered. A delivery truck tracking within the reinforced zone. The wall may survive it, but the check that said it would survive did not include it.

Terracing has its own trap: an upper wall set close behind a lower one is a surcharge on the lower one, and the two must either be separated by a distance the design method specifies or be designed as a single system. Where the separation is not there, the lower wall carries the upper — usually without being told.

Backfill, in lifts, and away from the face

Backfill in lifts you can genuinely compact — thin enough that the equipment reaches the bottom of the lift, which for hand-guided plate work means shallow lifts, not the depth that empties the barrow fastest. Compaction is normally specified as a percentage of maximum dry density determined by ASTM D698, Standard Test Methods for Laboratory Compaction Characteristics of Soil Using Standard Effort, and on walls of any consequence it is worth testing rather than judging by feel.

Keep heavy compaction off the face. Within the strip immediately behind the units, use hand equipment and work parallel to the wall, moving from the block outward. Running a big plate hard against the back of the course drives the units forward, and the bulge shows up two or three courses below the one you were standing on, which is why crews often blame the wrong lift.

Grid gets pulled taut and staked before it is covered, never left slack and never buried with a wave in it — slack grid does nothing until the wall has already moved enough to take it up. Place fill onto the grid and spread outward toward the face; do not turn tracked equipment on exposed reinforcement, and get the specified cover down before anything drives on it.

Backfill material matters as much as the effort. Free-draining granular fill in the reinforced zone reduces the driving load, drains, compacts predictably, and does not shrink and swell. Site clay does the opposite of all four.

Gravity or reinforced, and where the engineer line sits

A gravity wall resists everything with its own mass and geometry — unit weight, unit depth, batter, embedment. It works well at modest heights with good soil, no surcharge, and honest drainage, and it stops working at a height that depends on all four of those, not on a single number printed anywhere.

A reinforced wall extends the mass backward with geogrid, so the structure being weighed against the earth pressure is the reinforced soil block rather than the blocks alone. The tradeoff is that far more of the wall's performance now sits in work that gets buried: grid length, layer spacing, connection at the face, compaction between layers. It is a design, and it fails as a design when any of that is shortened on site.

The height at which a jurisdiction requires an engineered design and a permit varies, and should be treated as varying. Many places set a threshold on exposed height — a figure in the range of a few feet is common — but the number differs between jurisdictions, and it is routinely reduced or waived entirely where there is a surcharge above, a slope above, tiered walls, or a structure within the zone of influence. Some authorities require engineering for any wall carrying a surcharge regardless of height. Confirm the threshold with the authority having jurisdiction before quoting, because the difference between a stacked wall and an engineered one is priced in at bid time or absorbed later.

Where a design method is called for, the segmental industry design manuals published for these systems are the reference the engineering will be built on; the manufacturer's tested values for their specific unit and grid combination govern the connection details.