The metric most APAC brands are about to measure 9.9 by — and why it's wrong.
ROAS and total revenue are about to lie to every APAC marketing team this 9.9. Why both are structurally broken in 2026 — and the four reads that actually measure performance.
In short: In 14 weeks, most APAC marketing leaders will walk into a post-9.9 review with ROAS and total revenue as their headline metrics. Both are 2024 numbers and both are structurally broken in 2026 — ROAS is blind to conversation commerce and cross-platform discovery, total revenue confuses pull-forward demand with incremental demand. The four metrics that actually measure 2026 performance — incremental revenue per channel, chat-surface contribution, cohort quality, and AI-share of discovery — run on a marketing intelligence layer that sits on top of integrated D2C with conversation commerce attributed in. Build the layer in the 14 weeks before 9.9, or have the same conversation in October you had last October.
Contents
The post-9.9 review is one of the most consequential meetings an APAC marketing team runs. The numbers on the slide shape the brief for 11.11. They determine which channels get more budget, which agencies get renewed, which strategies survive into December. Getting the measurement wrong isn't a minor analytical problem — it cascades forward into every decision made in the back half of the year.
In 2026, most APAC brands will walk into that review with ROAS and total revenue as their headline metrics. Both numbers will look like they're telling the truth. They won't be.
Why ROAS is lying in 2026
ROAS only measures revenue inside the platform's last-click attribution window. In 2026 a meaningful share of buying decisions happens outside that window — in conversation commerce surfaces, in AI-mediated discovery, across multiple platforms. ROAS is structurally blind to all of it.
ROAS was designed for a world where the buyer clicked an ad and bought something on the same platform. The platform tracked it, credited it, and returned a tidy ratio. That world still exists — it's just no longer the whole picture. In 2026, a meaningful portion of the 9.9 buyer journey looks different.
A buyer sees a Shopee ad. Doesn't click. Searches on ChatGPT or Perplexity for a comparison. Finds the brand mentioned. Opens WhatsApp and asks a pre-cart question. Buys. That sale is invisible to Shopee's ROAS report. It never completed a click-to-purchase loop inside a single platform attribution window. The ad that started the journey gets zero credit. The integrated D2C layer that stitched those touchpoints together — if it existed — would have seen it. Without it, ROAS sees nothing.
There's also the double-counting problem. When a buyer's journey touches Shopee, Meta, and the brand's own site, each platform's attribution model claims the sale. Total ROAS across platforms can exceed 100% of actual revenue because every platform fired a last-click attribution and counted the win. This isn't a bug in any single platform — it's what happens when individual platform ROAS is used as a cross-channel measurement tool it was never designed to be.
In 2024, ROAS was an imperfect but useful signal. In 2026, with conversation commerce contribution, AI-mediated discovery, and cross-platform buyer journeys all growing, it's a structurally misleading number. High ROAS on a platform can mean great performance — or it can mean the platform's attribution model captured credit for a journey it started on a different surface entirely.
Why total revenue is also lying
Total revenue captures what was bought during the campaign window, not what was bought because of the campaign. Pull-forward demand reads as incremental, discount depth hides margin reality, and conversation commerce revenue gets miscategorised as organic. The headline metric quietly lies.
Total revenue as a 9.9 metric has three structural problems that compound each other.
Pull-forward demand looks like incremental demand. A buyer who was planning to purchase in October buys on September 9th because of the discount. Total revenue counts that as a 9.9 sale. But it isn't incremental revenue — it's October revenue that got pulled into September. Without a counterfactual model built on last year's data, there's no way to distinguish the two. Marketing intelligence can do this. A standard revenue dashboard can't.
Discount depth disguises margin reality. A 9.9 with heavy promotions can produce a headline revenue number that looks strong while margin per transaction is significantly below the brand's healthy baseline. Total revenue doesn't tell you whether the revenue was worth generating at the discount level it required. Brands optimising for total revenue can run a 9.9 that grows the top line and destroys the economics of the buyer cohort they acquired.
Conversation commerce revenue gets miscategorised. When a buyer reaches the brand through WhatsApp or platform chat and converts, that revenue often doesn't flow back into the campaign revenue attribution correctly. Depending on how the integrated D2C layer is configured — or whether it exists at all — chat-surface conversions are frequently categorised as direct or organic rather than as campaign-attributed. The total revenue number understates conversation commerce's contribution and overstates channel-agnostic demand.
A brand can run a 9.9 that looks excellent on total revenue and ROAS — and be making three strategic errors simultaneously: over-counting pulled-forward demand as new demand, under-counting chat-surface contribution, and optimising for a cohort of buyers who never come back for 11.11. The measurement layer they used was the problem, not the execution.
Wrong metrics vs right reads
| What Most Will Measure | What It Actually Tells You | What to Measure Instead |
|---|---|---|
| ROAS | Spend-to-revenue inside platform's attribution window only | Incremental revenue per channel (counterfactual model) |
| Total revenue | Sum of campaign-window sales, regardless of cause or margin | Chat-surface contribution + cohort quality |
| Last-click attribution | Whichever platform fired last got the credit, often double-counted | Integrated D2C attribution across surfaces |
| Q4 sales lift | Pull-forward + discount depth disguised as new demand | Cohort returning for 11.11 + 12.12 |
| Reach + impressions | Top-of-funnel volume, mostly noise | AI-share of discovery — buyers via AI-mediated paths |
What to measure instead — the marketing intelligence layer
Four reads on a marketing intelligence layer measure 2026 9.9 performance accurately — incremental revenue per channel (not credited), chat-surface contribution to revenue, cohort quality of acquired buyers, and AI-share of discovery. None of them run on a 2024 dashboard.
The four reads require a marketing intelligence layer built on top of AI on first-party data with conversation commerce and integrated D2C attribution plumbed in. They can't be pulled from a platform dashboard. They run on the data layer the brand owns, not the data the platform chooses to share.
Incremental revenue per channel.
Not credited revenue — incremental revenue. The counterfactual question: what revenue happened because of this channel that wouldn't have happened without it? Calculated by AI on last year's 9.9 data calibrated against this year's spend and buyer patterns. The difference between credited and incremental is often significant — particularly for channels that fire late in the journey and claim credit for buyers who were already decided.
Chat-surface contribution to campaign revenue.
The share of 9.9 revenue that touched a conversation commerce surface — WhatsApp, Shopee chat, Lazada chat, Instagram DM — at any point in the buyer journey before purchase. Not just direct chat purchases: any buyer whose path included a chat interaction before converting. Without integrated D2C attribution this number is invisible. In 2026, across brands with active conversation commerce, it's often 20–40% of campaign revenue — and it's being dropped entirely from the measurement.
Cohort quality of 9.9 buyers.
Not how many buyers were acquired — how good they are. Cohort quality tracks whether 9.9 buyers come back for 11.11 at full price, whether their AOV increases on the second purchase, whether they engage with brand communications in the 30 days after 9.9, whether they opt into WhatsApp. A brand can acquire 50,000 buyers on 9.9 and have 80% never come back — because the discount attracted a cohort structurally unlikely to repurchase at normal margins. Cohort quality is the leading indicator that tells you whether 9.9 built the business or borrowed from it.
AI-share of discovery.
The percentage of 9.9 buyers who reached the brand through an AI-mediated path — a ChatGPT recommendation, a Perplexity comparison, a Shopee or Lazada AI-native search result, a TikTok Shop algorithm recommendation rather than a direct ad click. As AI-mediated discovery grows as a share of top-of-funnel, brands without this read are blind to a channel they can't optimise because they can't see it. AI-share of discovery is built from first-party post-purchase survey data combined with traffic source modelling — not from platform dashboards that have no incentive to surface it.
Build the marketing intelligence layer in 14 weeks.
RedDot Solutions sets up marketing intelligence on your first-party data — AI on last year's 9.9, 11.11, and 12.12 — fast enough to inform this year's brief, and running through the campaign window with daily reads during peak. Integrated with conversation commerce and D2C attribution so chat-surface revenue stops getting dropped on the floor.
Talk to RedDot about marketing intelligence for 9.9 2026 →What this means for the 14 weeks you have left
14 weeks is enough to build the marketing intelligence layer that will inform the June brief, run weekly through the campaign window, run daily during peak, and produce a post-9.9 review with the four reads on the slide instead of ROAS and total revenue.
The marketing intelligence layer has a build sequence that fits inside 14 weeks if it starts now. The first four weeks are data — pulling first-party data from Shopify, Shopee seller data, Lazada seller data, post-campaign surveys from last 9.9, and any chat conversation logs that exist. Cleaning, structuring, and making it queryable by AI. This is the foundation; without it, none of the four reads are possible.
Weeks five through eight are modelling — the counterfactual incremental revenue model calibrated against last year's data, the cohort quality scoring framework built on last year's buyer cohorts, the conversation commerce attribution logic wired into the integrated D2C layer, the AI-share of discovery methodology agreed and instrumented.
Weeks nine through 14 are live operation — weekly reads through the campaign build, daily reads during the 9.9 window itself, and the four reads already baked into the post-9.9 review template before the campaign launches. Not built retrospectively after the campaign. Already in the measurement framework as the campaign runs.
The brief that comes out of week eight — the one the agency uses to build the 9.9 campaign — is a structurally different document when it's informed by marketing intelligence on last year's data. Different segments get prioritised. Different channels get budget. Different creative angles get briefed. The measurement layer isn't just about the post-9.9 review. It shapes the campaign itself.
The measurement decision you make now is the conversation you have in October
The post-9.9 review happens in October. The numbers on that slide come from decisions made in May and June — specifically, what measurement infrastructure was built before the campaign went live.
Brands that go into 9.9 with ROAS and total revenue as their measurement framework will come out of it with ROAS and total revenue as their post-9.9 review. They'll use those numbers to brief 11.11. Some of those numbers will be wrong in the same structural ways they were wrong this year. The cycle repeats.
Brands that build the marketing intelligence layer now — the four reads, properly instrumented — will come out of 9.9 with a post-campaign review that shows incremental revenue by channel, conversation commerce contribution as a real number, cohort quality as a forward indicator, and AI-share of discovery as a signal they can actually act on for 11.11.
The marketing intelligence layer isn't just measurement. It's the feedback loop that makes each campaign smarter than the last. Build it before 9.9 and it runs for 11.11 and 12.12 automatically — the data accumulates, the models improve, and the brief-shaping insight compounds. Wait until October to think about it and you've used the same lying metrics for one more cycle.
Frequently Asked Questions
What's coming next
Over the next few weeks I'll be publishing more on this — specifically on what the marketing intelligence setup actually looks like in practice, and a piece on conversation commerce attribution that goes deeper on how chat-surface contribution gets measured properly.
If you're a brand running on Shopee, Lazada, or D2C in APAC and any of this is hitting somewhere — these pieces are written for you.
Follow me here for the next pieces. Or read the full version on reddot.solutions.
If you're planning your 9.9 measurement framework now and want to talk through the four reads — and what it takes to build the marketing intelligence layer underneath them — we'd be happy to map it out.
RedDot Solutions is the tech agency for APAC brands — marketing intelligence, conversation commerce, integrated D2C.
— Balan K
Founder, RedDot Solutions
Balan K is the founder of RedDot Solutions, an APAC technology partner working with D2C and consumer brands on AI-native marketing infrastructure. Based between Singapore and Chennai.
reddot.solutions