For most established South African brands the answer is both, and the argument is really about sequencing and margin rather than loyalty to a channel. A marketplace sells you demand at a known price. Your own store sells you margin, customer data and a business that can be valued, at the cost of having to generate demand yourself. The mistake almost every retailer makes is comparing the marketplace commission against zero, when the correct comparison is the commission against what it costs you to acquire the same customer independently.
The reason to run this arithmetic now rather than next year is that the South African market has just crossed a threshold, and the competitive picture inside it has changed materially in twelve months.
Where the market actually stands
The most recent industry research into South African online retail, published on 02 September 2026, sets the context for any channel decision made this year. The headline figures are worth having in front of you before the arithmetic starts.
South Africans are expected to spend around R159 billion online in 2026, a growth rate of roughly 22.5 percent, against nominal growth of about 4 percent across retail as a whole. Online retail is recording its first full calendar year at approximately 10 percent of total retail turnover. The R29 billion in additional online turnover expected this year is almost the size of the entire South African online retail market in 2020, which sat at around R30.2 billion.
On platform usage, the research puts Takealot at 35.3 percent of online shoppers, followed by Shein at 21.5 percent and Checkers Sixty60 at 15 percent. Amazon was used by 12.7 percent of online shoppers before Prime launched locally, which it did on 03 June 2026 at R59 per month or R399 per year, making South Africa the twenty-seventh Prime market. Category leadership is now fragmented: Shein leads clothing, Checkers Sixty60 leads groceries, Takealot leads gifts, and Dis-Chem leads toiletries and cosmetics.
Two further details matter for anyone weighing channel strategy. Takealot Group recorded its first full-year trading profit fifteen years after launch, with adjusted earnings before interest and tax of R171 million. And subscription propositions have become a battleground, with Amazon Prime at R59 a month, Shoprite’s Xtra Savings Plus at R99 a month for unlimited free deliveries, and TakealotMore accounting for more than a quarter of Takealot Group gross merchandise value within two years.
Read together, these say something specific. The market is growing fast enough to support new entrants, the platforms are now competing on retention rather than acquisition, and the middle of the market is more contestable than it was. That is a better environment for an independent retailer than the one that existed three years ago, and a worse one for a retailer whose entire relationship with the customer is mediated by somebody else’s app.
What a marketplace actually costs
Marketplace fee structures are usually discussed as a single commission percentage. They are not a single percentage, and the difference between the headline number and the effective rate is where margin quietly disappears.
On Takealot, sellers face a monthly account fee in the region of R400 excluding VAT, a success fee set by product category that published seller guidance commonly places somewhere in the region of 4 to 18 percent, calculated on the VAT-inclusive selling price rather than the ex-VAT price, plus a per-unit fulfilment fee based on size and weight if you use their distribution network, plus storage fees on stock that sits too long, plus returns handling, plus whatever you choose to spend on promotions and sponsored placement.
Amazon’s South African structure is similar in shape: a professional selling plan at a comparable monthly figure, referral fees set by category that published guidance places broadly in the region of 6 to 15 percent with a minimum charge per item, and fulfilment and storage charges if you use their network.
Both platforms publish their own fee estimators, and any seller making a real decision should run their actual products through the official calculator rather than a blog table, because the category assignment is what determines the rate and the size class is what determines the fulfilment fee.
The useful exercise is to stack the fees on a realistic product rather than quoting the commission alone. Take a R299 homeware item sold through a marketplace and fulfilled by it. VAT comes off first. Then a category success fee calculated on the inclusive price. Then a fulfilment fee that, for a small light item, might be in the region of R30 and for a bulky one is several times that. Then a share of returns, which run far higher in apparel and similar categories than in others, and which cost you the fulfilment in both directions.
Stacked together on a low-priced physical product, it is entirely normal for the platform’s total take to land in the region of a third of the selling price before you have paid for the goods. That is not a criticism of the platforms. They are providing warehousing, delivery, payment, customer service, returns handling and, most valuable of all, a customer who was already there. It is simply a different number from the headline commission, and it is the number your pricing has to survive.
What your own store actually costs
The comparison people run is marketplace fees against nothing, because their own website feels free once it is built. It is not.
A WooCommerce store carries a build cost, hosting, maintenance, a payment gateway taking roughly 2.9 to 3.5 percent on cards and around 1.5 to 2 percent on instant EFT, several of them adding a flat rand fee per transaction on top, courier costs you now negotiate and manage yourself, packaging, your own customer service time, and the returns handling that the marketplace was doing for you. All of that is real and most of it is manageable.
The item that decides the outcome is not on that list. It is customer acquisition cost. A marketplace bundles demand generation into its commission. Your own store does not. Every order on your website is either the result of organic visibility you have built, advertising you have paid for, an email you were permitted to send, or a customer who came back on their own.
For most South African retailers, first-sale acquisition on their own website costs more than the marketplace fee. Which is why the decision does not turn on the first sale.
The crossover point is repeat purchase
Run the comparison over a customer lifetime rather than a transaction and the picture inverts.
On a marketplace, you pay the full fee on every order, forever. The tenth order from the same customer costs you exactly what the first one did, because you never acquired that customer, the platform did, and you are renting them each time. You usually cannot email them. You often cannot see who they are.
On your own store, the acquisition cost is front-loaded. If the customer returns, the second order carries a gateway fee and fulfilment but no acquisition cost. Contribution margin on repeat orders is dramatically better, and it compounds.
So the real question for any product is: does this customer come back, and how often? A product with a genuine repeat pattern, consumables, supplements, pet food, coffee, skincare, replacement parts, printer supplies, is a product where owning the customer relationship is worth paying to acquire. A one-off, price-shopped purchase in a commoditised category with no reason to return is a product that may legitimately belong on a marketplace, where the demand already exists and the fee buys you the traffic you would otherwise have to purchase.
Most retailers have both kinds of product in the same catalogue and treat them identically. That is the expensive mistake.
The four things a marketplace will never give you
The customer. You get an order, not a relationship. No email address you may market to, limited demographic insight, no ability to build a segment and speak to it, no way to bring that customer back except by paying the platform again.
Brand. Marketplace listings are designed to make products comparable, which is the opposite of what a brand is for. Your product sits in a grid next to eleven near-identical alternatives, differentiated by price, rating and delivery time. Everything you have invested in positioning is flattened into a thumbnail.
Control. Category fee changes, algorithm changes, promotional requirements, account suspensions and the platform’s own private-label products competing in your category are all decisions taken without you in the room. A business whose entire revenue runs through one marketplace account has a single point of failure it does not control.
A saleable asset. This is the one founders tend to discover late. A business built entirely on marketplace sales has revenue but very little that transfers: no customer list, no brand equity in a channel you own, no organic traffic, no direct relationships. A business with its own store, its own audience and its own data is worth materially more when someone eventually values it.
The model that actually works
The retailers doing this well in South Africa do not choose. They run a portfolio and give each channel a job.
Marketplace as discovery and liquidation. Use it to reach buyers who would never find you, to test new products against real demand, and to move overstock and end-of-range items without discounting on your own brand. Accept the margin. Do not build the business on it.
Own store as margin and relationship. Full range, bundles, subscriptions, loyalty, anything customised, anything where you want to control the experience. The place where the customer becomes yours.
Deliberate differentiation between them. This is where most retailers get lazy. If your own store is simply the same catalogue at the same price with worse delivery than the marketplace, nobody has any reason to buy from you directly. Give the direct channel something the marketplace cannot match: bundles and sets, a subscription option, personalisation, extended warranty, loyalty pricing, first access to new products, or simply a better buying experience for people who care about your brand rather than the cheapest matching listing.
A migration path. Every marketplace order is a chance to introduce the customer to the brand within the platform’s rules. A well-designed insert in the parcel, a registration or warranty process, a brand experience that makes them look you up next time. Read the platform’s policies carefully and stay inside them, because losing the account to win a mailing list is a bad trade.
What your own store has to do to earn the difference
If you are going to give up the marketplace’s demand, the website has to hold up its end. In practice that means five things, and most independent South African stores are weak on at least three.
Findability. Product pages that can rank, clean structured product data, and a category architecture that matches how people actually search. If your entire catalogue is invisible outside of paid traffic, you are paying acquisition costs forever with no compounding.
Checkout that does not leak. Mobile-first, minimal fields, guest checkout, visible delivery cost early, and no surprises at the last step. South African traffic is overwhelmingly mobile and the checkout is where independent stores lose to platforms that have optimised this for a decade.
Payment coverage that matches the market. Card, instant EFT, and the buy-now-pay-later options that a meaningful share of South African shoppers now expect at the R300 to R3,000 order values where they change conversion. Platforms offer all of it as standard. If your checkout offers card only, you are losing orders you never see.
A delivery promise you can keep. The marketplaces have trained the market on next-day and same-day in the metros, and Prime now offers South African members unlimited free next-day delivery with no minimum spend, plus same-day in Johannesburg, Pretoria and Cape Town on orders placed before midday. You do not have to match it, but you do have to state your actual lead time clearly and hit it. Vague delivery language on a product page converts worse than an honest five-day promise.
Retention that runs. Email and messaging to customers who gave you permission, post-purchase sequences, replenishment reminders on consumables. This is the machinery that makes the whole model work, because it is what turns an expensive first order into a profitable fifth one.
One emerging factor worth watching
Discovery is beginning to move again. A growing share of product research now starts inside an AI assistant or an AI-generated answer rather than a search results page, and those systems lean heavily on structured, well-described product data and on third-party sources that discuss products credibly.
The practical implication for a South African retailer is not to chase a new channel. It is that the quality of your product data, the descriptions, the attributes, the structured markup, the specifications, is becoming a discovery asset in its own right, wherever the product is listed. Retailers who publish thin product pages with a manufacturer’s paragraph and three attributes are invisible to those systems. Retailers with genuinely complete product data are not. That work is worth doing for its own sake, and it happens to be the same work that improves conversion on the pages you already have.
How to decide, concretely
Take your top ten products by volume. For each one, calculate the contribution after all marketplace fees, and the contribution after all costs on your own website including a realistic acquisition cost. Then add the variable that decides it: how many times does a typical customer buy this product in three years?
Products with a repeat pattern and an acquisition cost you can control belong on your own website, with the marketplace used for discovery. Products with no repeat pattern, thin margin and heavy price competition belong wherever the demand already is, and there is no shame in that.
What does not work is the default position most retailers drift into, which is to run both channels at the same prices with the same catalogue, invest in neither properly, and conclude that ecommerce is difficult. The market is growing at more than 22 percent a year. It is not difficult. It is simply unforgiving of channel strategy that was never actually decided.
If you are weighing up a marketplace-first or direct-first approach, or your own store is not currently good enough to justify moving volume onto it, our team builds and optimises WooCommerce stores for South African retailers, including the product data, checkout and payment work that decides whether the direct channel earns its margin.
Frequently Asked Questions
How big is South African online retail in 2026?
Industry research published on 02 September 2026 puts expected online spend at around R159 billion for the year, growing about 22.5 percent, and records the first full calendar year at roughly 10 percent of total retail turnover.
What does Takealot charge sellers?
Takealot charges a monthly account fee in the region of R400 excluding VAT, a category-based success fee commonly published in the region of 4 to 18 percent, calculated on the VAT-inclusive selling price, plus per-unit fulfilment fees based on size and weight, storage fees on slow-moving stock, and returns handling. Run your own products through the official fee estimator before committing.
Is it cheaper to sell on my own website than on a marketplace?
Only if you can acquire the customer for less than the marketplace fee. The commission bundles demand generation, which your own store has to fund itself. The economics usually favour your own store over a customer lifetime rather than on the first sale, because repeat orders carry no acquisition cost.
Should I sell on both Takealot and my own store?
For most established brands, yes, with a clear job for each channel. Use the marketplace for discovery, new product testing and moving overstock, and your own store for full range, bundles, subscriptions and repeat customers. Differentiate the two deliberately, or customers have no reason to buy direct.
What is the biggest risk of selling only on a marketplace?
Concentration. Category fee changes, algorithm shifts, promotional requirements and account suspensions are all outside your control, and you cannot contact your own customers to route around them. A business built entirely on marketplace revenue also has far less transferable value if you ever sell it.
Does Amazon’s arrival change things for South African retailers?
It broadens competition and raises customer expectations on delivery and subscription benefits, with Prime having launched locally on 03 June 2026 at R59 per month or R399 per year. It also adds a second large demand pool for sellers. The strategic implication is unchanged: use the platforms for demand, and own the customer relationship wherever the product justifies it.
Make the direct channel worth it
Product data, checkout, payment coverage and retention are what earn the margin you give up to a marketplace. We build South African stores that hold up on all four.
Take a look at our ecommerce web design, or contact the team.
