The SEA marketplace in 2026: a fragmented arena with one survival rule

The Google–Temasek–Bain e-Conomy SEA 2025 report, published in October 2025, projects Southeast Asia's digital economy to surpass US$300 billion in gross merchandise value by year-end. Inside that aggregate, e-commerce alone is on track for roughly US$185 billion GMV and US$41 billion in revenue. Across the same period, video commerce — a category Bain defines to include live-stream selling, creator-driven content commerce, and short-form video shopping — has grown to account for approximately 25% of regional e-commerce GMV, a share that did not exist in 2020 and was below 10% as recently as 2022.

The macro trajectory is unambiguous. The seller-side picture is the opposite. The same period that produced regional GMV growth produced systematic margin compression at the operator level, driven not by individual seller mistakes but by structural shifts in how the platforms allocate visibility, price impressions, and split discount funding. This note documents those shifts, the per-platform-per-market matrix they produce, and the operating rule that — across every cell of the matrix — separates sellers who hold contribution margin from sellers who do not.

The argument is empirical where it can be (citing the Bain report, Sea Limited's 4Q25 and 1Q26 investor disclosures, Alibaba's SEA segment filings, and the platforms' own seller-facing documentation) and acknowledged where it cannot be (the per-account margin distribution we describe is drawn from accounts we model directly; the methodology section at the end of this note describes the sample and its bounds). What follows is not a marketing post on competition. It is a research note on what changed in the SEA marketplace from 2022 to 2026, why the change is structural rather than cyclical, and what an operating layer that survives it actually does at the SKU level.

There is no SEA marketplace playbook. There are eighteen — three platforms × six markets — and the operating layer that wins models all of them at once.

Thesis: in an arena now fragmented across eighteen distinct platform-market cells (three dominant platforms — Shopee, Lazada, TikTok Shop — across the six SEA-6 markets of Indonesia, Thailand, Vietnam, the Philippines, Malaysia, and Singapore), sellers who run a single account-level operating view lose; sellers who run per-cell economics keep margin. The fragmentation is not a bug to be normalized away. It is the moat. The seller who can model all 18 cells at once compounds an advantage that is hard to copy precisely because it is administratively expensive — the same characteristic that prevents naive consolidation is the characteristic that makes the moat durable.

The 2026 baseline

Five years ago, the e-Conomy SEA reports treated the region as a growth story whose primary explanatory variable was internet penetration. By 2025, that variable is largely solved. Per the Bain report, internet penetration across the SEA-6 averages above 80%, and the marginal e-commerce buyer is no longer the new internet user. The variable has shifted to share-of-wallet and frequency, both of which are now determined by a different set of mechanics: platform recommendation systems that govern visibility on a saturated product surface, live-commerce momentum, creator-affiliate match-up, and the ad-auction prices that sellers themselves bid up by competing for the same impressions.

Read carefully, the table above is not a story of uniform regional growth. Indonesia and Vietnam carry the regional volume; Singapore and Malaysia have higher per-capita GMV but a smaller absolute footprint. Video-commerce share is not uniform — Indonesia and Vietnam sit at 25–30%; Singapore and Malaysia are below 20%. Shopee is the dominant platform across five of the six markets, but the gap to the second-largest platform varies meaningfully (Lazada is materially closer to Shopee in Singapore and Malaysia than in Indonesia or Vietnam). A seller building a SEA strategy from the regional aggregate is averaging across markets that, by 2026, are no longer averagable — different penetration curves, different live-commerce maturity, different consumer-spending profiles, different ad-auction densities.

Three structural shifts

The growth-to-margin-compression transition is the cumulative output of three concurrent shifts in how the platforms operate. Each shift is documented in the platforms' own filings or seller-facing documentation; none is a matter of opinion. The shifts compound, and the compounding is the operating-side story.

1. Ad auctions matured.

Bain's 2025 commentary identifies retail media as one of the two channels (alongside video commerce) reshaping marketplace discovery. On the platform side, Sea Limited's 4Q25 and 1Q26 investor disclosures — and the Reuters reporting that followed on 3 March 2026 — document that Sea's operating expenses rose materially during 2025 as the company invested in AI infrastructure for search relevance, recommendations, and advertising auction optimization. The seller-side translation is direct. In 2022, a Shopee account in Thailand could maintain meaningful organic flywheel with ad spend below 6% of revenue, because the auction was thin and the recommendation system had not yet been tuned for monetisation density. By 2026, the same impression share typically requires 12–18% of revenue on ads. The shift is not cyclical — it reflects an active platform investment in monetising every available impression, and the operating direction is one-way.

A second-order effect compounds the first. Shopee's default ad targeting allows broad-match keyword expansion, in which the platform automatically surfaces a campaign for related queries the seller did not bid on. Broad-match expansion captures a meaningful share of low-intent traffic — informational queries, comparison-shopping queries, queries for substitute products — and the attributed orders convert at materially lower true-ROAS than exact-match keywords the seller chose deliberately. The platform's relevance model gets better at expanding; the seller's cost-per-acquired-sale rises with it. Lazada's Sponsored Discovery product and TikTok Shop's feed-based ad placements exhibit the same dynamic in different forms.

2. Platform-funded promotions became seller-funded.

Shopee's Help Center, in the seller voucher mechanics article published at help.shopee.co.th, makes the funding source explicit: "When a buyer redeems a Shop Voucher, the discount amount is deducted from your sales as a marketing cost." Lazada's Open Platform documentation distinguishes between platform-funded and seller-funded promotions and notes that the Free Shipping Logo program requires the seller to opt in to bear part of the shipping cost in exchange for the badge's ranking signal. TikTok Shop's Seller University documents Affiliate Plan commissions in the 1–80% range (seller-set per SKU), with the Open Affiliate Plan defaulting to 10–25% on most categories — a commission that stacks on top of the 1–5% platform commission, before any seller-funded voucher applies.

Read these documents end to end and the conclusion is uncomfortable. The discounts buyers see, the free shipping that drives conversion, and the affiliate spend that drives reach are, by documented design, paid by the seller. The platform's role is increasingly an auction house — it allocates impressions; the seller funds the inventory of attention. This is not a complaint or a policy critique. It is the operating fact that the contribution-margin reconstruction has to account for, and it is the fact most platform-side dashboards do not display.

3. Cross-platform arbitrage closed.

In 2018, a savvy multi-shop operator could exploit a price gap between Shopee and Lazada for several days, sometimes longer. The arbitrage was meaningful, durable, and required only attention. By 2026 it has effectively closed. Three changes drove the closure. First, buyers comparison-shop across all three platforms in seconds, often through TikTok Shop's in-feed surface that displays competitor prices alongside the in-feed product. Second, third-party automation that mirrors competitor prices is widely deployed by competing sellers; price gaps surface within minutes and are mirrored back within hours. Third, the platforms themselves have rolled out competitive-parity rules on their highest-tier programs (LazMall, Shopee Mall) that explicitly police downward against off-platform price points. What remains is not cross-platform price arbitrage. It is operational arbitrage — the gap between the seller who knows per-SKU per-channel contribution margin and the seller who watches GMV.

The platform-market matrix

Each cell of the resulting 18-cell matrix has its own commission schedule, its own program-specific fees, its own holiday-calendar voucher mechanics, and its own dominant ad-auction product. The table below captures three platforms across three representative SEA-6 markets — Thailand, Indonesia, and Vietnam — to illustrate the structural difference between cells. The Singapore, Malaysia, and Philippines cells are similar in shape but differ in absolute fee level and in voucher-tier escalation cadence.

The structural difference between cells is not a rounding error. A Thai seller comparing Shopee Mall against TikTok Shop on the same SKU is comparing one cost stack (Shopee commission ~6% + Mall premium ~5% + transaction fee ~2% + voucher subsidy ~5–8% during campaigns) against a fundamentally different stack (TikTok platform commission ~3% + Open Plan affiliate commission ~20% on affiliate-tagged orders + transaction fee ~2% + a returns reserve typically 4–8 percentage points above Shopee's, reflecting the impulse-purchase profile of the buyer base). The two stacks produce break-even ROAS targets that are 2–3 ROAS points apart on the same SKU. A seller running a single account-wide ROAS target across the cells is, by construction, mistargeting at least one of them.

Three worked examples

The structural shifts and the matrix are abstract until the operating arithmetic is specified. The three formula blocks below trace the same hypothetical scaling pattern — a representative Thai seller scaling monthly GMV from THB 200K to THB 800K across each of the three platforms — using the typical fee and voucher structure of each cell. The numbers are illustrative; they reflect a typical SKU-mix profile in our data, not a specific seller account. The point of the exercise is not the precise margin produced; it is the per-cell shape of the margin compression.

Pre-scale  THB 200,000 GMV · 35% gross · 8% ads · 3% normal voucher · ~2% fees
                     net contribution margin: ~22% of GMV  →  ~THB 44,000

At-scale   THB 800,000 GMV · 35% gross · 16% ads (auction-inflated)
                  · 9% campaign voucher (escalated tier) · 4% free-ship subsidy · ~2% fees
                     net contribution margin: ~4% of GMV  →  ~THB 32,000

Outcome:  GMV grew 4×; net contribution profit fell 27%.

Pre-scale  THB 200,000 GMV · 35% gross · 7% ads · 4% Mall fee premium
                  · 3% mandatory free-shipping subsidy · 4% normal voucher
                     net contribution margin: ~17% of GMV  →  ~THB 34,000

At-scale   THB 800,000 GMV · 33% gross (LazMall pricing parity drag) · 14% ads
                  · 4% Mall premium · 4% mandatory free-shipping · 10% campaign voucher
                     net contribution margin: ~1% of GMV  →  ~THB 8,000

Outcome:  GMV grew 4×; net contribution profit fell 76%.

Pre-scale  THB 200,000 GMV · 35% gross · 6% ads · 3% platform commission
                  · 18% affiliate (50% of orders, 20% commission) · 6% returns reserve
                     net contribution margin: ~2% of GMV  →  ~THB 4,000

At-scale   THB 800,000 GMV · 35% gross · 14% ads
                  · 3% platform commission · 22% affiliate (60% of orders, 22% commission)
                  · 8% returns reserve (live-commerce mix)
                     net contribution margin:  −12% of GMV  →  −THB 96,000

Outcome:  GMV grew 4×; net contribution profit collapsed into a loss.

The three examples diverge sharply. The Shopee non-Mall path scales with margin compression but stays profitable on the increment. The LazMall path scales close to break-even because the Mall fee premium and mandatory shipping subsidy compound with the campaign voucher escalation. The TikTok Shop path scales into a loss because the affiliate commission, layered on top of platform commission and a higher returns reserve, exceeds gross margin once campaign voucher escalation is added. The same SKU mix, scaled the same way, produces three distinct margin trajectories. A seller running an aggregate dashboard across the three cells sees average margin compression and concludes that "the market is competitive." A seller running per-cell economics sees that two of the three cells are losing money on the increment and one is not, and reallocates accordingly.

Sensitivity — where the conclusion changes

Stress-test the worked examples. The table below shows how the at-scale net contribution margin from worked example 1 (Shopee TH non-Mall) changes when one input moves while the others hold. The sensitivity is asymmetric — small movements in voucher tier and ad spend change the outcome more than equivalent movements in COGS or platform commission, because voucher and ad cost are themselves auction-driven and tend to escalate rather than fluctuate.

The sensitivity table makes a structural point. Small shifts in two of the largest cost inputs — ad spend and voucher tier — produce 50%+ swings in at-scale net contribution margin. The seller who lacks a real-time per-SKU per-cell view of these inputs is operating with a margin signal whose noise floor is larger than the operating signal. The seller who has the view can act on the table directly: cut ads two percentage points and refuse the next voucher-tier escalation, and the at-scale margin recovers.

The survival rule, applied to the matrix

Run the business on contribution margin per SKU per channel per market, not GMV. The rule is simple to state and administratively expensive to apply, which is why it is also the moat. Six operational implications follow from it directly.

• Which campaigns to scale: those whose post-attribution contribution margin per unit is positive on the increment, judged in the cell's own break-even ROAS, not against a global account target.
• Which discounts to publish: those whose voucher tier does not exceed the category's gross margin in the specific cell — campaign vouchers that work in Shopee TH may compress margin to break-even in LazMall TH on the same SKU.
• Which products to retire: those whose contribution margin is negative across every cell after a fair audit — not those that are negative in one cell while positive in another.
• Which channels to push a SKU into: the cells where the SKU's margin holds; not the cells with the highest aggregate GMV.
• Which campaigns to refuse: those whose participation cost (campaign voucher tier, mandatory ad spend, deeper free-shipping commitment) exceeds the SKU's contribution margin even before the campaign's GMV lift is realised.
• Which competitors to ignore: the ones competing on GMV. Their compression is on a different curve and following them is a margin-compression decision wearing a growth-strategy mask.

The cells are not averagable. The operating layer that wins models all 18 at once.

Limitations and where this argument breaks

A research note that does not name its scope is advocacy in disguise. The argument above has explicit limits.

• Account-size lower bound. The argument applies cleanly to accounts above approximately THB 200K monthly revenue. Below that, fixed operating costs (warehouse, fulfillment overhead, the seller's own time) dominate and the per-SKU optimisation framework underperforms simpler heuristics. A solo seller with 5 SKUs and THB 50K monthly revenue should not be running per-cell contribution-margin reconstructions; they should be picking their two best SKUs and watching cash flow.
• Account-size upper bound. The argument also has a top end. Top-tier sellers (LazMall enterprise tier, Shopee top-decile sellers) have negotiated commission schedules, custom voucher arrangements, and direct platform relationships that change the matrix in ways our public-fee-schedule modelling does not capture. The framework still applies; the input values are different and require recalibration.
• Cross-border sellers. China-to-SEA, Hong Kong-to-SEA, and Korea-to-SEA cross-border sellers face a different fee and customs cost stack — Shopee International Platform, Lazada Cross-Border, and TikTok Shop's cross-border seller programs each carry distinct economics not modelled here. The operating principle holds; the numerical bounds shift.
• Platform coverage. We treat Shopee, Lazada, and TikTok Shop as the three dominant platforms across SEA-6. This excludes Tokopedia (Indonesia-specific, large in absolute terms but not regional), the Bukalapak / Blibli long tail in Indonesia, Sendo and Tiki in Vietnam, and emerging surfaces (Instagram Shopping in Singapore, Meta Shop integrations) that may matter at the margin. Sellers with meaningful Tokopedia exposure should treat the analysis as a useful-but-incomplete starting point.
• Time horizon. The structural shifts described here are observable over the 2022–2026 window. The directional trajectory is stable but the rate is not — Sea Limited's 2026 AI investment may either accelerate or moderate the ad-auction cost trajectory depending on whether the platform passes efficiency gains to sellers (compressing the auction) or extracts them (further inflating the auction). Forecast confidence beyond 24 months is low.
• Margin distribution claim. The "in our data" claims are aggregated and anonymised; they reflect the accounts we model. They are not population claims about all SEA marketplace sellers, and they explicitly exclude the bottom and top of the size distribution as discussed above.

Methodology — the numbers behind the "in our data" claims

Public-data citations in this note are taken from the Bain e-Conomy SEA 2025 report, Sea Limited's 4Q25 and 1Q26 investor disclosures (filed via SEC Form 6-K and the Sea Investor Relations portal), Alibaba's Lazada SEA segment disclosures from the same period, the Reuters reporting on Sea's March 2026 quarterly results and the February 2026 Google–Sea AI partnership, and the platforms' own seller documentation: the Shopee Help Center (help.shopee.co.th and equivalent country domains), Lazada Open Platform documentation (open.lazada.com), and TikTok Shop Seller University (seller-th.tiktok.com).

Internal-data claims — the cost-input ranges in the worked examples, the typical scaling patterns, the "in our data" margin distributions, and the "20–30% of ad spend runs at a true loss" figures — are aggregated across approximately 400 active SEA-6 marketplace seller accounts across the DataGlass research methodology sample frame (Jan 2024 – Apr 2026, 28-month observation window).

The methodology section exists for one purpose: to make every numerical claim in the note inspectable in principle. A reader who disagrees with the conclusions should be able to point to the input that is wrong (the public-data citation, the sample, the cost-input range, the attribution model) rather than to the conclusion itself. This is the difference between analysis that can be argued with and analysis that cannot.