Let the livestream host improvise, but never the claims
Live selling has become the fastest-growing channel for Southeast Asian beauty brands and the least supervised advertising they run. Every session should be governed as an advertisement, with an approved claims list, spoken disclosure and returns measured beside revenue.
At nine o'clock on a weekday evening in Jakarta, one of our hosts may be three hours into a broadcast, holding a jar of Embunsari moisturiser up to a ring light and answering questions as fast as they scroll past. How long does it last? Is it safe for my daughter? Will it fade the marks on my cheeks? Is it halal? Can I use it with my prescription cream? She has a few seconds for each, an audience in the thousands and a sales counter running in the corner of the screen. No advertisement we have made was ever produced under such pressure, and until 2024 none was so lightly checked.
Live selling now produces about two-fifths of our online revenue, and online is 22 per cent of the group. I am in favour of it. It restores something that department-store counters and market stalls always had: a person who can show the product, answer the question in front of her and admit when it is wrong for someone. Shoppers trust it for that reason, and that trust is exactly why the claims made in it matter.
A broadcast that writes its own advertisement
A conventional advertisement passes through a brief, a script, legal and regulatory review and, in some markets, pre-clearance before anyone sees it. A livestream skips all of that. The host improvises, the creator adds her own enthusiasm, and the result is broadcast, recorded and sometimes clipped and shared for weeks. In law and in the shopper's mind it is still the brand speaking, and I think the brand must answer for it as though it had written every word.
When we reviewed 60 recorded Embunsari sessions in early 2024, unapproved claims appeared in 27. Most were not lies. They were the natural inflation of a salesperson in a hurry: fades dark spots in seven days, safe in pregnancy, endorsed by dermatologists, chemical-free because it is halal. Each was a claim we could not support, and some were claims no cosmetic should make. In 19 of the 44 sessions run by contracted creators, the commercial relationship was not disclosed in a way an ordinary viewer would notice.
Advertising codes in our markets are clear that a paid relationship must be evident to the audience. Singapore's code has said so for years, and authorities and industry bodies elsewhere in the region have moved in the same direction. Disclosure alone does not solve the problem, though. A host can disclose perfectly and still promise a result the product cannot deliver.
The claims card and the first minute
Our answer was to keep what makes live selling work, the host's own voice, and to fix what that voice may promise. Every host now works from a claims card: 41 approved statements about our products, written in Indonesian and Malay, and 16 statements that are prohibited however they are phrased. The card also carries approved answers to the questions hosts hear most often, including questions about pregnancy, medication and halal status, and tells them when to say that customer care will answer instead.
Disclosure is part of the contract, not a courtesy. A contracted creator must say in the first minute that Embunsari pays her, and a label must stay on screen throughout. A session containing an unapproved claim is not paid, and a second breach ends the contract. A compliance reviewer watches a sample of sessions live and can message a host during the broadcast; the rest are sampled from recordings within a week.
None of this was free. We paused contracted-creator sessions for three weeks to introduce it, and Embunsari's online revenue fell 24 per cent that month. Thirteen creators chose not to sign the new contract. Our head of e-commerce, who argued against the pause, was right that marketplace algorithms would punish our absence; it took two months to recover our rankings. I still think the pause was right, because the alternative was to pay people, for another quarter, to make claims we knew to be unsupported.
What a board should ask about live selling
Most boards of consumer companies see live selling as a revenue line. It also belongs in the risk register as a conduct risk, because it is the one place where the company's advertising is made by people it may not employ, in real time, at a volume no review team can match. Four measures help.
The first is the share of sessions sampled, live or recorded, and the rate of unapproved claims found. A company that cannot state that rate is not supervising the channel. The second is the disclosure rate in creator sessions, measured by someone other than the team that hires the creators. The third is returns on live orders compared with other online orders. Ours fell from 12.1 to 6.9 per cent after the changes, which told us how much of the earlier revenue had been borrowed. The fourth is claim-related complaints per ten thousand orders, which fell from 31 to 9.
I would add one judgement a board should test directly. When a high-selling creator breaks the rules, who decides whether to keep her, and on what evidence? If the answer is the person paid on her sales, the control is weaker than it looks on paper.
There are honest objections to all this. Smaller brands cannot afford a reviewer watching sessions, and a claims card can make a host sound stiff if it is written by lawyers rather than by people who sell. Both are fair. A small brand can sample recordings weekly at modest cost, and our cards are written by former hosts and checked by regulatory affairs, which is the right way round.
Live selling will keep growing, and its formats will change faster than any policy can anticipate. That argues for principles rather than rules about particular apps: the host may say anything true and approved, must disclose who pays her, and must stop when she does not know. The brands that last in this channel will be those whose hosts are allowed to be human and are never allowed to improvise a promise.