One tenth on research, nine tenths on hooks: Sonic Song’s NTD 30m split

The first thing Sonic Song did with NTD 30 million was spend a tenth of it finding out what he had.

Three million New Taiwan dollars went into market research, consumer surveys and the marketing activity that came off them, over the first month or two. The remaining twenty-seven million went into merchandise and game-equipment packages, which Song calls consumer hooks. Just under a million US dollars, split one to nine, in that order.

The property was Arena of Valor.

Arena of Valor is TiMi Studio Group’s international adaptation of Honor of Kings, the Tencent title that has spent a decade near the top of the mobile market in China. It opened on the Taiwan server on 12 October 2016 and in Thailand that December, distributed in both markets by Garena. By the end of 2018 it had passed 100 million installs outside China, and Taiwan alone supplied 29 percent of its international revenue.¹

Song’s brief was narrower than the game’s scale suggests: growth and exposure, Taiwan and Thailand, six months.

What he reports from that window is a three-fold increase in downloads, measured against the company’s own reports, followed by a stay of five to six years on the account.

The tenure is the quiet return on the tenth. A six-month brief that becomes a five-year account never goes back out to tender, and the client is renewing the sequence as much as the man, having watched the small spend earn the big one once already.

The split is the part an operator can copy.

Market research, consumer surveys and marketing activities, about one or two months, one tenth of the budget.

Research ran first, ran fast, and cost little enough to be wrong. The nine tenths that followed bought physical objects a player could hold, which is a bet nobody unwinds once the containers ship.

Sequence is doing the work here, rather than the ratio. A tenth spent afterwards is a post-mortem.

The obvious objection is that nine tenths on keyrings and headsets looks reckless for a property Tencent had already proven in China. It fails because the twenty-seven million moved only after the three million came back with answers; every carton was ordered against a survey, at a scale where the survey cost less than one shipment.

Song had spent twenty years in media, news and PR before this, and 2011 was his first year out of salaried work. The thirty million arrived with the decision to leave, and the split was the habit of a man spending someone else’s money in his first month of owning the risk.

The habit was tested from the other side a few years in.

A second mandate came to him at NTD 20 million, and he turned it away, because the projects already on his desk left him nobody to run it. Asked what the rejection cost, his answer is one word long: money.

Two thirds of an Arena of Valor sat on the table, and the constraint was headcount rather than method. The method he could have photocopied.

So the next spend went into people, on the same sequencing he sold to clients. He asked the upper end of his market for recommendations first, hired against them, and took half a year to assemble what he calls the dream team, on payroll but never called staff.

When I was focused on my own, independently, I didn’t realise the right people were already around me.

“When I was focused on my own, independently, I didn’t realise the right people were already around me.”

The screen he built for them runs on the same cheap-question logic as the surveys: five minutes to judge fit, thirty to confirm it, one minute of introduction carrying three achievements and three failures, then straight into live work with safety nets under it.²

There is one number Song no longer holds. Asked what the three-fold growth was measured from, he gives the honest answer instead of a reconstructed one: it has been too long to remember the actual figures. The multiple sat in the client’s reporting, which is where it stayed.

Every operator running a mandate on a platform they do not own meets that gap eventually, and it arrives fifteen years after the work, when the work is being valued.

The cheap half of the fix costs a fraction of what the research did. Screenshot the download curve, the survey deck and the shipment counts the week they are true, and keep them on the agency’s own drive rather than the client’s.

Song still runs the order he ran then, across four companies now, in markets where the brands belong to other people. The industries change, gaming to lifestyle to luxury, and he treats them as interchangeable as long as the base muscle is branding, advertising, IP and marketing, and the mandate pays.

The split never changes. Ask the small money first. Commit the rest once the answer comes back, whether the rest is twenty-seven million in merchandise or six months of hiring.

Song dates the NTD 30 million mandate to 2011, the year he left salaried work; the Taiwan server opened in October 2016. Fifteen years is long enough for two true things to settle onto one date, which is the same argument for keeping the receipt. Song sat with CÈ’s AI interviewer on 16 August 2026, by text; quotes lightly normalised for print, wording untouched, transcripts on file. The budget split, the three-fold figure, the NTD 20 million rejection and the tenure are his account; Arena of Valor’s dates, publisher and install base are from the public record. ↩
The five-minute screen has its own piece; see the related reading. The one wrong hire it let through, by Song’s account, passed every flag at the table and underperformed later for personal reasons, and he trusts the instrument anyway. ↩

Footnotes