Scale, Hold, or Cut: A Channel iROAS Head-to-Head for Sea Limited
Objective: Given a supplied 6-channel iROAS readout for Sea Limited's Shopee marketing mix, decide which channels to scale, hold, or cut, correctly separating branded search's low incrementality from its high platform-reported ROAS.
You're the paid media lead at Sea Limited reviewing a completed multi-channel geo-lift program across Shopee's Southeast Asia markets.
Head-to-head every channel's platform ROAS against its measured iROAS, then make a scale/hold/cut call per channel.
Before you start
What you'll need
Free path (everything below is enough to finish)
Free, handles a 6-channel comparison table and simple threshold rules without any specialist software
The process
3 steps
Step 01 of 03
Attribution asks which channel touched the conversion, incrementality asks which channel caused it. Only iROAS tells you where to put budget.
Shopee's dashboard reports CTV at 1.8x ROAS and branded search at 6.2x ROAS. The geo-lift test measures CTV's iROAS at 3.3x and branded search's at 0.7x. Which platform number was more misleading?
Procedure
- List each channel's platform-reported ROAS
- List each channel's measured iROAS next to it
- Sort by the size of the gap between the two, largest gap first
Channel PlatformROAS iROAS Gap Branded Search 6.2x 0.7x -5.5x CTV 1.8x 3.3x +1.5x
Healthy
Branded search's platform number is recognized as the misleading one, despite looking like the best channel on the dashboard.
Unhealthy
Reading CTV as the underperformer because 1.8x looks worse than branded search's 6.2x on the dashboard.
What this means
The channel with the biggest gap between platform ROAS and iROAS is the one your dashboard is lying about hardest.
So what do I do about it?
| Symptom | Action | Effort |
|---|---|---|
| A channel's platform ROAS is far higher than its measured iROAS | Treat the platform number as attribution noise, budget off the iROAS figure instead | 30 min |
Step 02 of 03
The lesson's benchmark table shows branded search scoring lowest (0.70x) industry-wide, people already searching for your brand by name were going to click the organic result anyway.
Branded search's iROAS came back at 0.7x for Sea Limited. Marketing wants to double the branded search budget because it 'converts so well'. What's the response?
Procedure
- Compare Sea Limited's branded search iROAS (0.7x) to the lesson's industry median for branded search (0.70x)
- Confirm the pattern matches, not an anomaly specific to this test
- Recommend a spend cut, not an increase, for branded search
Category SeaLimited iROAS Industry median Branded search 0.7x 0.70x <- matches, not an anomaly Non-branded search 1.6x 1.46x
Healthy
Branded search spend gets cut, not doubled, once the near-zero iROAS is confirmed against the industry pattern.
Unhealthy
Doubling branded search budget because its platform ROAS and conversion rate both look strong.
What this means
A high-converting channel and a high-incrementality channel are not the same thing; branded search is the textbook example of the gap.
So what do I do about it?
| Symptom | Action | Effort |
|---|---|---|
| Branded search shows a high platform ROAS but sub-1x iROAS | Cut branded search spend toward defensive-minimum levels, redirect the saved budget to non-branded or CTV | 30 min |
Step 03 of 03
The lesson's closing line: your ad platform's ROAS measures what happened near your ads, incrementality measures what your ads actually caused.
Given the full 6-channel iROAS table, which channels get a scale, hold, and cut recommendation for next quarter?
Procedure
- Apply a scale/hold/cut threshold to each channel's iROAS
- Write one line justifying each call using the iROAS number, not the platform ROAS
- Total the current spend on 'cut' channels to size the reallocation pool
Channel iROAS Call CTV 3.3x Scale Non-branded SEM 1.6x Hold Branded Search 0.7x Cut TikTok 0.9x Cut
Healthy
The scale/hold/cut calls are justified entirely by iROAS thresholds, with no channel's call flipped by its platform ROAS.
Unhealthy
Hedging on the branded search cut because the platform dashboard still shows a strong ROAS number.
What this means
A clean threshold rule applied consistently beats a case-by-case argument with whoever owns the highest-platform-ROAS channel.
So what do I do about it?
| Symptom | Action | Effort |
|---|---|---|
| A scale/hold/cut call keeps getting re-litigated by channel owners | Publish the threshold rule and the iROAS table together, so the call is mechanical, not political | 30 min |
Final deliverable
A 6-channel scale/hold/cut table with the platform-ROAS-vs-iROAS gap and a one-line justification per channel.
See a reference example
Nubank paid channel review (excerpt) Channel PlatformROAS iROAS Call Meta 4.1x 2.6x Scale Branded SEM 7.0x 0.6x Cut CTV 2.0x 3.1x Scale
Success criteria
You're done when you can:
- Correctly separates platform ROAS from iROAS for every channel
- Recommends cutting, not scaling, branded search despite its high platform ROAS
- Justifies every scale/hold/cut call using the iROAS threshold rule, not the platform number