Landscape analysis, demand data, China sourcing economics and margin modelling for a filtration supply business serving residential, commercial and trade channels in Centurion and greater Tshwane.
There is a real business here, but it is not the business we assumed when we started. The money is in consumables and mid-market residential filtration, sold nationally through ecommerce and locally through Duncan's installation capability. Atmospheric Water Generation is a poor fit for the Highveld and should be treated as marketing, not revenue.
Three channels off one inventory pool:
Own the replacement cartridge relationship. Whoever installs the system controls a 10-year consumable annuity at 80%+ margin. Sell the system near cost if you must — win the annuity.
This matters because it determines what we sell and how we talk about it. The honest version of the Gauteng water story is more useful commercially than the scary version — and it is defensible if a customer pushes back.
Gauteng does not primarily have a water quality problem at the treatment works. It has a distribution and continuity problem between the plant and the tap. Rand Water's bulk supply is generally compliant; the failure is in ageing municipal reticulation, leaking reservoirs and intermittent supply. That distinction is the entire product strategy.
| Structural failure | Evidence | What it creates | Product we sell |
|---|---|---|---|
| Ageing reticulation & pipe bursts | Johannesburg Water recorded 16 pipe bursts in six days; 43 structurally leaking reservoirs, estimated R1.3bn to repair | Sediment, rust and discolouration at the tap after every repair and re-pressurisation | Point-of-entry sediment filtration (Big Blue 20") |
| Intermittent supply / outages | Rand Water major maintenance 29 May–2 Jun and 17 Jul 2026 across Joburg, Tshwane, Ekurhuleni | Pressure loss → back-siphonage risk; households install storage tanks | Tank inlet/outlet filtration + UV |
| Storage tank adoption | "jojo tank filter" 210/mo, "water tank filter" 390/mo | Stored water goes stagnant, grows biofilm, collects roof debris | Tank kits: pre-filter + UV steriliser |
| Municipal financial collapse | Joburg losses R2.4bn/8 months; Tshwane ~R1.9bn/yr; Rand Water demanding a R2bn+ deposit | No capital for repair → the problem compounds and is not solved within our planning horizon | Justifies the whole category long-term |
| Borehole substitution | "borehole water filter" 720/mo, "borehole water purification system" 170/mo | Raw groundwater: hardness, iron, nitrates, bacteriological risk | Multi-stage borehole treatment trains — highest ticket |
| Hard water & scale | "limescale" 590/mo, "hard water" 480/mo, "water softener" 590/mo | Geyser, kettle, espresso and ice machine damage | Softeners and scale inhibitors — the commercial wedge |
Sources: Johannesburg Water / DA infrastructure statements (Feb–May 2026); Rand Water maintenance notices; Daily Maverick, IOL, Engineering News (2026); Blue Drop 2025 via DWS/PMG; search volumes from DataForSEO, Google Ads, South Africa, July 2026.
The Blue Drop 2025 assessment found metro drinking water systems generally performing well, with low-risk systems improving from 60.2% to 61.9% nationally. If we run "your tap water is poisoning you" messaging in Centurion, we are factually vulnerable and it will eventually cost us credibility and possibly an ASA complaint.
The defensible pitch: "The water leaving the plant is fine. What happens in the 40-year-old pipe between there and your kitchen is not — and neither is what happens in your tank when the supply goes off. Filter at the point you actually drink."
Pulled live from DataForSEO (Google Ads data, South Africa, English, July 2026). These are national monthly search volumes.
Google groups close variants at identical volume — "water filter system for home", "home water filter system" and 18 other phrasings all report 3,600. Summing the raw list inflates the market roughly 4×. The figures below are de-duplicated to distinct clusters. Treat ~36,500/month as directional, not additive.
"puritech" = 2,400/mo at $1.96 CPC. That single branded term is bigger than almost every generic product category. It tells us the market is brandable — people search for a trusted name, not a spec. That is the opening for a well-built brand.
Shower filters: 1,300/mo at $0.54 CPC, and almost nobody in SA merchandises them properly. FOB ~$3.80, retails R279–420. Tiny CBM, high margin, perfect ecommerce basket-filler and gateway product.
Local terms are dead. "water filter centurion" is 20/mo. Do not build the plan around local SEO. Build national ecommerce, and use Centurion proximity as a service and trust advantage, not a traffic source.
Source: DataForSEO Google Ads Search Volume API, location South Africa, language en, retrieved 28 July 2026. Raw dataset: research/data/keywords-za-core.json (59 terms) and keywords-za-longtail.json (461 terms).
Six categories, mapped to the problem they solve and the channel that sells them.
| Category | What it is | Solves | Channel | SA retail |
|---|---|---|---|---|
| Consumable cartridges | PP sediment, CTO carbon block, GAC/UDF, T33 polisher, RO membranes, 20" jumbo | Everything. Every system needs them every 6–12 months | All three — subscription | R89–R650 |
| Under-sink RO | 5-stage 75GPD w/ tank; 6-stage remineralising; tankless 400/800GPD | Drinking water taste, TDS, chlorine, dissolved solids | Ecommerce + installed | R1,999–R11,290 |
| Point-of-entry | Big Blue 20" 2/3-stage whole-house, brackets, pressure regulation | Sediment and rust after bursts; protects geyser and appliances | Installed (Duncan) | R2,760–R4,200 |
| Tank & borehole | UV sterilisers 12GPM, auto-backwash multimedia, softeners (FRP + Runxin valve) | Stored/stagnant water, hardness, iron, bacteriological risk | Installed — highest ticket | R5,500–R14,442 |
| Commercial | 400GPD / 1600GPD RO, scale-inhibitor heads, ice & espresso filtration | Limescale destroying equipment; consistent product quality | Direct B2B contracts | R2,400–R14,000 |
| Retail / impulse | Shower filters, faucet-mount, countertop gravity units | Low-commitment entry; skin/hair; renters who can't install | Ecommerce only | R279–R1,800 |
Lead the range on under-sink RO and whole-house sediment — that is where search demand, price tolerance and Duncan's installation skill all intersect. Carry softeners and borehole trains as high-ticket installed work. Use shower/tap filters to build the email list and cartridge subscriber base cheaply.
The market is wide open in the middle. There is a credible budget end and a credible premium end, and a large, poorly served gap between them.
| Player | Position | RO price band | Read |
|---|---|---|---|
| Puritech | Category leader. Vast catalogue — components through to full bottling plants. Sells Runxin, Vontron, Canature | R138 – R414,000 | Strongest brand (2,400 branded searches/mo). Also a China importer — so no sourcing moat against us. Broad but industrial-feeling; weak consumer brand experience |
| BWT (was H2O International) | h2o.co.za now 301-redirects to bwtshop.co.za — absorbed into the Austrian BWT group. Ecosoft product line. Gauteng + Cape Town offices, franchise model | R3,995 – R11,290 | Watch this A multinational has just bought into SA retail filtration. They will professionalise and out-spend. Premium anchor — good for us, we price under it |
| Little Luxury | Budget/DIY ecommerce. 6-stage RO at R1,999 | R1,999 – R2,599 | Sets the price floor. Competes on price alone; no installation. We beat them on service and trust, not price |
| FilterShop | Custom systems, camping/outdoor range | Mid | Already serves the outdoor/portable niche — relevant to the overlanding question |
| Local Centurion R&M | Plumbers fitting whatever the wholesaler stocks | n/a | Our trade channel These are customers, not competitors. Duncan already knows them |
A 6-stage under-sink RO sells for R1,999 at Little Luxury and R7,695 at BWT. Same fundamental technology, ~3.8× price spread. Nobody owns the R2,800–R4,000 installed-and-supported position — a properly branded system, professionally fitted by Duncan, with a cartridge plan. That is the space to take.
Geographic gap confirmed: the nearest affordable competitors are in Midrand and Hartbeespoort. For installed work — where the customer wants someone to physically arrive, fit it, and come back in nine months to change cartridges — a Centurion base is a genuine advantage over both. It is worth nothing for ecommerce, and everything for service.
All FOB USD figures are indicative, taken from public Made-in-China and Alibaba listings. They are placeholders for Mario to quote against — not a basis for committing capital. Everything else (freight, duty, FX, CBM, fee structure) is built on verified ProLogistics actuals from the Prebur O1/O2 containers.
| Input | Value | Confidence | Source |
|---|---|---|---|
| Duty — HS 8421.21 (water filtering machinery) | 0% | Verified | SARS Schedule 1 / JLog. Free on general, SADC, EU, EFTA, UK |
| Duty — HS 8421.99.90 (cartridges = parts) | 0% | Verified | Confirmed against SARS Sched 1 during Prebur, Apr 2026 |
| ProLogistics fixed transport, 40HQ | R106,800 | Actual | Prebur O1/O2 invoice reconciliation |
| ProLogistics fixed transport, 20ft | R80,000 | Estimated | Scaled from 40HQ — get Dheo to quote |
| LC / confirmation fee | 2.00% of goods | Actual | ProLogistics cost structure |
| Marine insurance | 0.275% of goods | Actual | ProLogistics cost structure |
| Foreign bank charges | R1,700 × 2 legs | Actual | E2W 30% deposit / 70% balance terms |
| FX planning basis | R16.50 / USD | In use | Prebur basis; spot slightly below = cushion |
| 40HQ practical load | 65 CBM | Actual | Prebur O2 loaded 60.5 CBM / 571 cartons |
Landed includes FOB, freight allocation by CBM, LC, insurance and bank charges. RRP is set by deliberately undercutting the local market anchor, not by applying a fixed multiple — the margin is what falls out of that.
| SKU | Order qty | FOB $ | Landed | Our RRP incl VAT | Market anchor | Trade | GM retail | GM trade |
|---|---|---|---|---|---|---|---|---|
| Consumables — the annuity | ||||||||
| RO membrane 75GPD | 600 | 4.50 | R78 | R449 | R650 | R281 | 80% | 72% |
| 10" CTO carbon block | 1,500 | 1.10 | R22 | R139 | R185 | R87 | 82% | 75% |
| 10" PP sediment cartridge | 2,000 | 0.40 | R10 | R89 | R120 | R56 | 87% | 82% |
| T33 post-carbon polisher | 1,000 | 1.00 | R20 | R135 | R180 | R85 | 83% | 77% |
| 10" GAC / UDF cartridge | 1,000 | 0.95 | R19 | R129 | R175 | R81 | 83% | 76% |
| 20" Jumbo PP (whole-house) | 400 | 2.20 | R48 | R195 | R260 | R122 | 72% | 61% |
| Under-sink RO — the volume driver | ||||||||
| 5-stage RO 75GPD + tank | 100 | 42.00 | R837 | R2,799 | R3,995 | R1,752 | 66% | 52% |
| 6-stage RO + remineralisation | 60 | 58.00 | R1,120 | R3,999 | R6,395 | R2,504 | 68% | 55% |
| Tankless RO 400GPD | 25 | 95.00 | R1,719 | R5,499 | R8,595 | R3,443 | 64% | 50% |
| Tankless RO 800GPD | 8 | 125.00 | R2,238 | R6,999 | R11,290 | R4,382 | 63% | 49% |
| Point-of-entry & borehole — Duncan's installed work | ||||||||
| Whole-house 2-stage Big Blue | 80 | 28.00 | R563 | R1,899 | R2,760 | R1,189 | 66% | 53% |
| Whole-house 3-stage + bracket | 40 | 42.00 | R837 | R2,799 | R4,200 | R1,752 | 66% | 52% |
| Tank/borehole UV 12GPM + pre | 25 | 85.00 | R1,589 | R3,899 | R5,500 | R2,441 | 53% | 35% |
| Softener 1054 + Runxin valve | 3 | 210.00 | R4,057 | R9,499 | R14,442 | R5,947 | 51% | 32% |
| Auto-backwash multimedia 1054 | 3 | 175.00 | R3,441 | R8,499 | R12,500 | R5,321 | 53% | 35% |
| Commercial — contract channel | ||||||||
| Scale-inhibitor cartridge head | 60 | 34.00 | R613 | R1,499 | R2,400 | R939 | 53% | 35% |
| Commercial RO 400GPD | 8 | 265.00 | R4,780 | R9,999 | R14,000 | R6,260 | 45% | 24% |
| Retail / impulse — ecommerce basket | ||||||||
| Shower filter | 400 | 3.80 | R72 | R279 | R420 | R175 | 70% | 59% |
| Countertop gravity filter | 60 | 18.00 | R368 | R1,199 | R1,800 | R751 | 65% | 51% |
| Faucet-mount filter | 200 | 4.50 | R87 | R299 | R450 | R187 | 67% | 54% |
| AWG — demo units only | ||||||||
| AWG 30L/day domestic (demo) | 1 | 720.00 | R13,049 | R24,999 | — | R15,652 | 40% | 17% |
| AWG 20L/day compact (demo) | 1 | 560.00 | R10,092 | R18,999 | — | R11,895 | 39% | 15% |
Trade price = RRP ex-VAT less 28%, for plumbers and resellers. Market anchors from Puritech, BWT/Ecosoft, Little Luxury and PriceCheck listings, July 2026. The semi-commercial 1600GPD RO has been removed from Order 1 — at the indicative FOB it prices above Puritech's R10,350 and is not competitive. Re-quote it before reconsidering.
MOQ 50–100 pieces per type. Trivially met. FOB $0.36–$1.19. This is the easiest, highest-margin part of the order.
MOQ 2–100 units depending on supplier. Most sit at 10–100. Our Order 1 quantities (10–100) clear this comfortably.
MOQ often 1 unit, but some factories quote 1 × 40GP container minimum. Confirm before assuming we can buy two demo units.
MOQ ranges from Made-in-China.com and Alibaba supplier listings, July 2026. Mario to confirm against actual E2W factory terms.
Order 1 ships in a 20ft container. The mix below loads 31.2 CBM against a 33 CBM practical capacity (94% full) at R676,991 landed — comfortably inside the R500–750k envelope.
Why not the 40HQ: at this budget the natural volume is ~31 CBM against 65 CBM of capacity. Shipping it half-empty means paying the full R106,800 fixed cost across half the volume — effective freight jumps from R1,712/CBM to R3,425/CBM, doubling the freight burden on every unit. Water filtration is bulky relative to value; genuinely filling a 40HQ needs roughly R1.3m of stock. That is the Order 2 decision, once we have sell-through data.
The first-pass mix came to 35.5 CBM — over a 20ft. Trimming it improved the container rather than weakening it, because the cuts fell on the weakest lines:
Net effect: capital down R95k, blended margin up from 64.7% to 66.1%, and return on capital up from 183% to 195%.
| Scenario | CBM | Fill | Freight/CBM | FOB goods | Landed capital | GP (60:40) | Return on capital |
|---|---|---|---|---|---|---|---|
| A — 20ft, Order 1 mix agreed |
31.2 | 94% | R2,565 | $35,175 | R676,991 | R1,318,172 | 195% |
| B — 40HQ, same mix rejected — under-filled |
31.2 | 48% | R3,425 | $35,175 | R703,791 | R1,291,372 | 183% |
| C — 40HQ, mix ×2 the Order 2 question |
62.4 | 96% | R1,712 | $70,350 | R1,297,383 | R2,692,943 | 208% |
Scenario C has better freight economics and a higher return, but it doubles capital at risk on a product range where we have zero sell-through data. Scenario A buys a full range test across all three channels for R677k. Once we know which SKUs actually move, Order 2 goes 40HQ and captures the R1,712/CBM freight rate on proven winners only.
Headroom: at R677k against a R750k ceiling there is roughly R73k of slack and 1.8 CBM of space. Hold it. It absorbs the FOB variance when Mario's real quotes land, and it is cheaper to add units to an existing order than to discover mid-negotiation that the budget is already committed.
| Category | Capital | % capital | Revenue @RRP | % revenue | GM | CBM |
|---|---|---|---|---|---|---|
| Consumables | R157,926 | 23% | R867,739 | 39% | 82% | 7.9 |
| Under-sink RO | R211,848 | 31% | R620,267 | 28% | 66% | 9.8 |
| Point-of-entry | R140,734 | 21% | R361,173 | 16% | 61% | 7.5 |
| Retail / impulse | R68,339 | 10% | R211,600 | 9% | 68% | 3.5 |
| Commercial | R75,004 | 11% | R147,767 | 7% | 49% | 1.9 |
| AWG | R23,141 | 3% | R38,259 | 2% | 40% | 0.6 |
| Total | R676,991 | 100% | R2,246,805 | 100% | 69.9% | 31.2 |
Consumables absorb 23% of capital and return 39% of revenue at 82% margin — while AWG absorbs 3% of capital to return 2% at 40%. Every rand moved from AWG into cartridges improves the container. That is exactly what the trim did.
| 0% (expected) | R676,991 | — |
| 5% | R706,011 | +4.3% |
| 10% | R735,030 | +8.6% |
| 15% | R764,049 | +12.9% |
Prebur lesson: a supplier mis-declared steel frames and we carried 10% duty until confirmed. Get Dheo to confirm the tariff line in writing before the deposit goes out.
| R15.50/$ | R641,016 | −5.3% |
| R16.50/$ (basis) | R676,991 | — |
| R17.50/$ | R712,967 | +5.3% |
| R18.50/$ | R748,942 | +10.6% |
At 66.1% blended GM the business absorbs a R18.50 rand without breaking — and even then lands just inside the R750k ceiling. FX is a risk to capital required, not to viability.
R80,000 is scaled from the verified 40HQ figure, not quoted. This is the weakest input in the model, so it is worth seeing how much it actually matters:
| 20ft freight | Landed capital | Blended GP | GM | Return on capital |
|---|---|---|---|---|
| R65,000 | R661,992 | R1,333,171 | 66.8% | 201% |
| R80,000 (assumed) | R676,991 | R1,318,172 | 66.1% | 195% |
| R90,000 | R686,992 | R1,308,171 | 65.6% | 190% |
| R100,000 | R696,992 | R1,298,171 | 65.1% | 186% |
Even at R100,000 — a 25% overshoot on the estimate — landed capital stays under R700k and margin holds above 65%. Worth confirming with Dheo, but it does not threaten the plan.
"restaurant water filter" gets 10 searches a month nationally. "commercial water filter" gets 30. This channel cannot be acquired through search or ecommerce. It is entirely a field-sales motion — which is precisely why it suits Duncan and precisely why it is defensible once won.
Not for health. For equipment protection and product consistency. That reframes the entire pitch from "clean water" to "your R80,000 espresso machine dies in 18 months instead of 8 years".
| Segment | Pain | Solution | Ticket | Recurring |
|---|---|---|---|---|
| Coffee shops / roasteries | Limescale kills boilers, group heads and heating elements; hardness ruins extraction. SCA specifies 50–70 ppm | Scale-inhibitor head + carbon; softener where hardness is high | R1,500–4,000 | Cartridge every 3–6 months |
| Restaurants | Ice machine scaling and failure; dishwasher spotting; carafe water taste | Ice machine filtration + under-counter RO for table water | R4,000–10,000 | 2–4 cartridge changes/yr |
| Guesthouses / B&Bs | Guest complaints on taste and smell; geyser scale; supply interruptions | Point-of-entry + tank treatment + UV | R6,000–15,000 | Annual service |
| Gyms / offices | Bottled water cost and logistics | Plumbed-in dispensers with RO | R5,000–12,000 | Service contract |
| Dental / medical / labs | Autoclave and instrument requirements demand low-TDS feed | Dedicated RO with monitoring | R8,000–20,000 | Compliance-driven, very sticky |
Sell the install at modest margin (45–55%), then put the customer on a scheduled service contract. A coffee shop on quarterly cartridge changes at R1,499 a head is ~R6,000/year of 53%-margin recurring revenue from a single site visit. Fifty sites is R300k/year of high-margin, near-zero-acquisition-cost revenue — and it is exactly the round Duncan is already driving.
Higher margin per unit, far higher acquisition cost, much weaker retention. Residential is where the ecommerce brand and search demand live; commercial is where the durable cash flow lives. Run both — but do not expect them to behave the same way, and do not let ecommerce metrics govern the commercial pipeline.
Do not build a business line on AWG in Gauteng. Carry one or two demo units for showroom presence, PR and high-net-worth curiosity sales. Budget it as marketing spend that happens to have residual value, not as inventory.
Centurion averages 55% RH, but falls to 44% in August — and that monthly average masks the diurnal swing. Winter daytime humidity routinely sits in the 20–30s.
Most AWG units carry a cut-off at 30–35% RH. Output roughly halves between 65% and 35% RH. We would be selling a machine that underperforms worst exactly when Gauteng water supply is most disrupted — the dry winter.
AWG consumes roughly 0.35–0.45 kWh per litre (efficient units ~0.2; many are 0.5–1.0).
At South African electricity prices that is on the order of R1 per litre in energy alone, before capital recovery. Municipal water is a few cents a litre. Filtered tap water via RO costs a fraction of a cent. AWG only makes economic sense against bottled water — a narrow, and shrinking, comparison.
40% GM retail, 17% trade — against 82% on cartridges and 66% on under-sink RO.
It also consumes disproportionate CBM (0.25–0.35 per unit), carries warranty and service risk on a refrigeration circuit, and has no consumable annuity worth having. Every metric we care about ranks it last.
You asked whether small serviceable AWG units exist for the overlanding community. They do — but the answer is still no, and the reason is worth knowing.
The units exist. 10–20 L/day portable and 12V AWG units are made in China (HurRain NanoTech A10, Guangdong Hongshu tabletop 12L, others) at roughly $550–$1,200 FOB, MOQ often 1 unit. Watergen's Mobile Box and ON Board are the established Western reference products. So supply is not the obstacle.
Southern African overlanders drive to the Kalahari, Namibia and Botswana. Dry-season relative humidity there runs 20–40% — at or below the cut-off where AWG compressors stop producing water at all.
The product fails hardest in exactly the places its buyers take it. A R25,000 unit that produces nothing in the Central Kalahari in July generates warranty claims, refund demands and reputational damage in a small, tightly networked, forum-driven community that talks constantly. This is the fastest way to poison a new brand.
Overlanders have a real, unmet water problem — it is just not an atmospheric one. They need to make questionable water safe: borehole water at a farm gate, river water, dubious campsite taps, water that has sat in a jerry can for two weeks.
That is a 12V mobile filtration product: 4-stage RO or 0.01µm ultrafiltration with UV, running off the vehicle's electrical system. There is already a live SA market — Pure Water 4 Life's MP-12V-4, Allmech's H2GO, Aqua Angel Pure Terrain, Little Luxury's cross-country unit, FilterShop's camping range — which is fragmented and, notably, has no brand leader. Duncan's electrical specialisation is a direct fit for 12V vehicle integration.
Sources: weather-and-climate.com (Centurion/Pretoria monthly RH); Nature Communications and PLOS Water on AWG performance vs RH; safaribookings/Kalahari climate data; Made-in-China supplier listings; SA4x4, 4x4community, Pure Water 4 Life, FilterShop.
| Risk | Severity | Mitigation |
|---|---|---|
| FOB prices are indicative, not quoted | High | Whole model is a placeholder until Mario quotes. Every margin figure moves with it. This is the single biggest unknown. |
| NRCS Letter of Authority | High | Any electrical component (UV lamps, pumps, AWG) may need an NRCS LOA before it can legally be imported and sold. SARS and NRCS have an MOA blocking entry without one. Check before ordering anything with a plug. |
| SANS 241 / performance claims | Medium | SANS 241 governs water quality, not devices — but if we claim compliance we must be able to prove it. Budget for independent lab testing of the lead SKUs |
| BWT's arrival | Medium | A multinational now owns the #2 brand. Expect professionalised marketing and pricing pressure. Compete on service and locality, not on spend |
| Duty reclassification | Medium | 0% is well supported but supplier mis-declaration burned us on Prebur steel. Written confirmation from Dheo before deposit |
| 20ft freight cost unverified | Low | R80,000 is scaled from the 40HQ actual, not quoted. Quantified above: even a 25% overshoot keeps landed under R700k and GM above 65%. Confirm with Dheo, but it is not a threat |
| Cartridge fitment fragmentation | Medium | The annuity only works if our cartridges fit our systems and ideally competitors'. Standardise on 10"/20" industry-standard formats — do not accept proprietary housings |
| Installation capacity | Low | Duncan is one person. Model assumes full sell-through; installed lines will throttle on his available hours. Plan a second installer before Order 2 |
| Sell-through timing | Low | 195% return on capital is on full sell-through, not annualised. Prebur ran 46% sell-through in ~5 months — a realistic figure here is 12–18 months to clear the container |
If Mario's real FOB prices come back more than ~35% above the indicative figures used here, the mid-market position gets squeezed between Little Luxury below and BWT above, and the whole thing needs rethinking. If they come back at or below these levels, this is a straightforward, well-understood import-and-distribute business with an unusually good consumables annuity attached.