บันทึกจากสนามจริง/โฆษณา เชิงเทคนิค

CTR and Target ROAS under CPM vs CPC: what changes when marketplaces bill for impressions

Under cost-per-click billing, a better creative is worth nothing to your unit economics — it buys the same clicks from fewer impressions, and your cost per click does not move. Under cost-per-impression billing the same improvement cuts the price of every click by g/(1+g). Five lemmas on the transition, and why click-through rate is the input most marketplace sellers are least prepared to have priced.

18 สิงหาคม 202612 นาทีBhum Soonjun · DataGlass Labs Research
CTR gain worth, under CPC
0%

On cost per click, at fixed spend

Same gain, under CPM
g/(1+g)

Cut in effective cost per click

Profit elasticity to CTR
(π+S)/π

Against exactly 0 under CPC

CTR spread, our fleet
1.14×

Flattest of six measured inputs

โฆษณา

Effective cost per click, under impression billing
eCPC = CPM / (1000 x CTR)

That identity is the whole argument. Under cost-per-click billing, click-through rate does not appear in your cost per click at all — you pay what the auction cleared, and a better creative buys the same clicks from fewer impressions. Under cost-per-impression billing, click-through rate sits in the denominator of the price you pay. The same creative improvement that was worth nothing becomes worth a proportional cut in the cost of every click you buy [1].

To be explicit before anything else: no Southeast Asian marketplace has announced a move from click billing to impression billing, and this post does not claim one is coming. It is contingency analysis — the algebra of what would change, worked out in advance, because the input it promotes is the one sellers here are least prepared to have priced. In our fleet data across 84 Shopee shops and 9.11 million clicks between 1 February and 16 August 2026, click-through rate is the flattest of six inputs we measure: 1.14x between the best and worst daypart, against 2.24x for CPM and 2.01x for CPC. Sellers vary their prices, their baskets and their conversion rates far more than they vary their creative [4].

The thesis: under click billing, click-through rate is a delivery lever — it changes how much you can spend, not what a click costs you. Under impression billing it becomes a price lever, and profit responds to it more than proportionally. What would falsify this: a marketplace moving to impression billing without sellers’ realised cost per click moving inversely with their click-through rate.

What a Target ROAS actually instructs

A Target ROAS is not a promise about the return you will earn. It is a price ceiling on what the platform’s bidder may pay for traffic on your behalf [6]. Write CR for the probability a click converts, V for the attributed order value when it does, and R* for the target you entered. Expected attributed revenue per click is CR x V, and the target says that revenue must be at least R* times what you paid for the click. Rearranged, that is a bound on the bid.

The bid ceiling a Target ROAS implies
CPC_max = CR x V / R*

Example: CR = 4%, V = THB 900, R* = 6
CPC_max = 0.04 x 900 / 6 = THB 6.00

Two things follow that are routinely read backwards. First, the target enters the denominator, so it is an inverse control: raise it and the permitted bid falls. Raising a Target ROAS does not raise the return your campaign earns — it lowers the price it is allowed to pay, and the return is whatever the auction delivers at that price. Second, on Shopee GMV Max you never see this number. Per Shopee’s own seller documentation the campaign is auto-bid; you set a budget and a target, and the platform sets the per-click bid [2]. The ceiling describes what constrains the platform’s bidder, not a field you fill in.

A Target ROAS is a price ceiling, not a profit promise. Raise it and you lower the bid, not the return.

The same ceiling, priced per thousand impressions

An impression becomes a click with probability CTR, and under click billing you pay only when the click happens. So the expected cost an impression may carry is the click ceiling discounted by the click rate — which means the auction is already valuing your impression at CTR x CR x V, whether or not your invoice mentions CTR.

The impression ceiling
CPM_max = 1000 x CTR x CPC_max

Example: CTR = 2.6%, CPC_max = THB 6.00
CPM_max = 1000 x 0.026 x 6.00 = THB 156

This is why a settlement change is a smaller event than it first appears, and why it is not a redesign of the auction. The quantity being priced is identical under both bases. What changes is the moment money moves: click billing charges on the click, impression billing charges on the impression. Everything below follows from that one change, and nothing else.

Why a better creative is worth nothing under CPC

Take two sellers, X and Y, identical in every respect — same conversion rate, same basket, same Target ROAS, same daily budget, same clearing price. X improves its creative and lifts click-through by a fraction g. Y leaves its creative alone. Under click billing, at a fixed budget S and a clearing price CPC, the number of clicks you buy is simply S divided by CPC, and click-through rate is nowhere in that expression.

Under CPC, at fixed spend
Clicks       C = S / CPC          <- no CTR
Cost/click   eCPC = CPC           <- no CTR
Revenue      G = (S / CPC) x CR x V  <- no CTR
Impressions  I = S / (CPC x CTR)   <- CTR only here

So:  eCPC_X = eCPC_Y,  G_X = G_Y,  profit_X = profit_Y

X’s better creative bought the same clicks from fewer impressions. That is not worthless — it means X wins more auctions at a given bid and can therefore spend more, because both auctions rank on expected revenue per impression. But at a fixed budget, the gain is exactly zero. The price of a click did not move, and neither did the return per baht.

Under impression billing the same comparison inverts. Now the budget buys impressions directly and clicks are what the click rate makes of them, so every quantity that was immune to CTR becomes proportional to it.

Under CPM, same fixed spend
Change in cost per click  =  -g / (1 + g)
Clicks, revenue, ROAS     =  x (1 + g)

Example: g = 25% better click-through
  eCPC falls 25 / 125 = 20%
  clicks, attributed revenue and ROAS all rise 25%
What a creative improvement is worth, by billing basis, at a fixed budget.
Click-through gainCost per click (CPM)ROAS (CPM)Cost per click (CPC)ROAS (CPC)
+10%-9.1%x1.10unchangedx1.00
+25%-20.0%x1.25unchangedx1.00
+50%-33.3%x1.50unchangedx1.00
+100%-50.0%x2.00unchangedx1.00

Measured against a seller who holds its creative fixed and faces the same clearing price. The CPM columns are -g/(1+g) and 1+g; the CPC columns are the algebra above. Same auction, same seller, same budget — only the billing basis differs.

Cut in effective cost per click, under impression billing
+10% CTR
9.1%
+25% CTR
20%
+50% CTR
33.3%
+100% CTR
50%

The reduction is g/(1+g), so it is concave: the first improvements are worth the most per point of CTR. Under click billing every bar here is zero.

The profit effect is larger than the ROAS effect

The table above understates the case, because profit is a difference and not a ratio. A creative gain adds g times your gross contribution — not g times your profit. When ad spend is large relative to what is left over, those two are very different numbers.

Worked example — profit, not ROAS
Monthly ad spend           S  = THB 10,000
Contribution the ads earn  muG = THB 12,000
Profit after ads           pi = 12,000 - 10,000 = THB 2,000

A 25% click-through gain under CPM:
  contribution  12,000 -> 15,000
  profit         2,000 ->  5,000    (+150%)

Because the gain is g x (pi + S) = 0.25 x 12,000 = THB 3,000

A 25 per cent improvement in click-through produced a 150 per cent improvement in profit. That multiple is not a coincidence — it is the elasticity of profit to click-through, and it equals gross contribution divided by profit [1]. The closer a campaign runs to break-even, the larger it gets, without bound. A campaign at THB 11,500 of contribution on THB 10,000 of spend has an elasticity of 7.7; one at THB 10,500 has an elasticity of 21.

The nearer a campaign runs to break-even, the more a creative improvement is worth — and under click billing, it is worth nothing at all.

The competitive consequence is sharper still. Under click billing, X’s bid ceiling does not move at all when its creative improves — the ceiling formula has no CTR in it. Under impression billing the permitted bid itself rises by (1+g) at the same target, so X can outbid Y on every impression while still satisfying the identical Target ROAS, and pay less per click for the impressions it wins. The billing change does not just reprice the winner; it hands the winner a higher ceiling.

Where this argument breaks

Four limits, in the order they would bite.

  • The comparison holds one seller against another at a given clearing price. It is not a market equilibrium. If every seller improves click-through at once, the clearing CPM rises and the advantage competes away — what survives is the relative position, not the absolute cut in cost per click.
  • Impression billing usually arrives with view-through crediting: conversions credited to an impression that produced no click. That makes the unclicked impression both billable and creditable at the same time. Whether the credit compensates for the charge depends on the view-through conversion rate and value, which nobody outside the platform can observe in advance.
  • Every figure here is on attributed revenue, which is not the revenue advertising caused. The gap between the two is organic sales the attribution rule absorbed, and at a fixed target that gap converts directly into permitted spend. A wider attribution window is spend headroom, not performance.
  • The rates are per traffic opportunity, not campaign averages. Substituting a campaign-average click-through rate into these expressions is an approximation whose error grows with how much the underlying rates vary — and broad attribution makes the value term vary a lot.

And the standing caveat: this is derived from definitions, not measured against a live CPM marketplace, because there is not one to measure here. The lemmas are exact given their assumptions. The magnitudes are not claims.

Methodology

The one empirical claim in this post is the input-spread figure. It comes from DataGlass fleet data: 84 Shopee shops, 1 February to 16 August 2026, single-product campaigns only so that a campaign’s clicks can be attributed to one product’s economics, on broad attribution. 9.11 million of 9.21 million clicks matched (98.9 per cent). "Spread" is the ratio of the best to the worst of six dayparts on that input: click-through 1.14x, margin 1.13x, conversion rate 1.23x, basket value 1.42x, CPC 2.01x, CPM 2.24x. These are aggregates across our own connected accounts, not a population claim about Southeast Asian sellers, and no individual shop is identified. The five lemmas use no data at all — they follow from the definitions stated in the PDF.

ก้าวต่อไป

Know what a click is worth before the billing basis changes.

DataGlass reconstructs contribution per product from your Shopee, Lazada and TikTok Shop order-line data — cost of goods, commission, payment fees, vouchers and a returns reserve — then derives the Target ROAS each product can carry. Whichever way a platform bills, the number that decides is the same one.

แหล่งข้อมูลและอ่านต่อ

  1. 01
    DataGlass Labs — CTR & Target ROAS under CPM and CPC (technical note, 18 August 2026)

    The five lemmas this post summarises, with the proofs, the notation, and the view-through and incrementality results omitted here for length.

    /assets/papers/DataGlass_CTR_and_Target_ROAS_under_CPM_and_CPC_2026-08-18.pdf

  2. 02
    Shopee Ads — GMV Max and Target ROAS bidding (Seller Education Hub)

    Shopee seller documentation on GMV Max as an auto-bid campaign type where the seller sets a budget and a Target ROAS and the platform sets the per-click bid — the basis for treating the bid ceiling as a constraint on the platform’s bidder rather than a field the seller fills in.

    https://seller.shopee.co.th/edu/

  3. 03
    Shopee — Seller commission and fee schedule (Help Center)

    Commission by category, transaction and payment fees, voucher mechanics and Free Shipping Program cost-share — the cost stack that turns attributed revenue into the contribution term used in the profit worked example.

    https://help.shopee.co.th/portal/article/77790

  4. 04
    Bain & Company — e-Conomy SEA 2025: retail media

    Retail-media growth and cost inflation across Southeast Asian marketplaces — context for why the price variables in our fleet data move more than the behavioural ones.

    https://www.bain.com/insights/e-conomy-sea-2025/

  5. 05
    Sea Limited — Investor Relations

    Sea Limited disclosures on Shopee advertising revenue and take-rate, the commercial pressure behind any change to how marketplace ad inventory is sold.

    https://www.sea.com/investor/home

  6. 06
    INFORMS — Operations Research & Analytics

    The professional body on optimising a decision against an explicit objective under constraints — the framing used here, in which a Target ROAS is a constraint on a bidder rather than an objective in itself.

    https://www.informs.org/Explore/Operations-Research-Analytics

อ่านต่อจากคลังบทความ

  1. 30 พฤษภาคม 2026

    Target ROAS คืออะไร? ตั้งใน GMV Max ผิด ยอดขายโต แต่กำไรหาย

    Target ROAS คืออะไร? คืออัตราส่วนรายได้ต่อค่าโฆษณาที่ Shopee GMV Max ใช้เป็นเป้า — บนชุดอัตราค่าธรรมเนียมของ Shopee ที่มีผล 4 สิงหาคม 2569 สินค้าราคา 1,000 บาทที่มีต้นทุน 600 บาท เหลือ margin ก่อนค่าแอดแค่ 7.7% ตั้ง Target ROAS ที่ 5 จึงได้ contribution กลับมาแค่ 0.39 บาท ต่อค่าโฆษณา 1 บาท ทั้งที่แคมเปญรายงานว่าทำได้ตามเป้า บทความนี้อธิบายว่า Target ROAS คืออะไร วิธีตั้งใน GMV Max และสองสูตร — break-even ROAS กับ target-margin ROAS — ที่ช่วยรักษากำไร

  2. 1 มิถุนายน 2026

    ROAS คืออะไร? ทำไม ROAS สูงไม่ได้แปลว่ายอดขายเยอะหรือกำไรดีเสมอไป

    ROAS (Return on Ad Spend) คือ ยอดขาย ÷ ค่าโฆษณา ยิงแอด 1 บาท ได้ยอดขายกลับมา 20 บาท ROAS ก็เท่ากับ 20 แต่ ROAS สูงแปลว่าได้ผลตอบแทนต่อค่าโฆษณา 1 บาทสูงเท่านั้น ไม่ได้แปลว่ายอดขายรวมเยอะที่สุด และไม่ได้แปลว่ามีกำไรจริงหลังหักต้นทุนสินค้า ค่าธรรมเนียม Platform ค่าชำระเงิน voucher ค่าส่ง และของคืน บทความนี้อธิบายว่า ROAS คืออะไร ซ่อนอะไรไว้ และตัวเลขที่ร้านค้าควรดูจริง ๆ

  3. 29 พฤษภาคม 2026

    ยิงแอดได้ยอดขาย แต่ทำไมไม่เหลือกำไร? — ROAS ที่แพลตฟอร์มไม่บอก

    ROAS 5 รู้สึกเหมือนกำไร จนกระทั่งรู้ว่าแพลตฟอร์มคำนวณจากยอดขายรวมก่อนหัก fee, voucher, ค่าส่ง และ COGS แถมยังนับ order ที่ถูกส่งคืนหรือยกเลิกไปแล้วด้วย ตัวเลขบน dashboard จึงเป็นอัตราส่วนยอดขายที่ใส่ชุดกำไร บทความนี้อธิบายว่า platform ROAS ซ่อนอะไรไว้ และ True ROAS คำนวณยังไงบน Shopee, Lazada และ TikTok Shop

  4. 25 มีนาคม 2026

    วิธีคำนวณ True ROAS ของ Shopee เพื่อกำไร

    บันทึกเชิงระเบียบวิธี ROAS ที่ Shopee แสดงในแพลตฟอร์มอิงจาก gross revenue และมี bias เชิงโครงสร้างให้ใช้เงินเกินจุดคุ้มทุนเมื่อร้านโต True ROAS คือสูตรเดียวกันที่แทน input หนึ่งตัว และการแทนนั้นพลิกแคมเปญที่ดูชนะให้กลายเป็นขาดทุนในเกือบครึ่งของ catalog Shopee ทั่วไป พร้อมชาร์ต SKU สามโปรไฟล์ sensitivity analysis และ operating procedure ที่ใช้การแทนนี้ในจังหวะ production

  5. 29 เมษายน 2026

    การจัดสรรงบโฆษณาข้ามแพลตฟอร์มสำหรับผู้ขาย marketplace ใน SEA

    ผู้ขายหลายแพลตฟอร์มส่วนใหญ่แบ่งงบโฆษณาระหว่าง Shopee, Lazada และ TikTok Shop ตามสัดส่วนรายได้ในอดีต แต่คณิตศาสตร์บอกว่านั่นผิด การจัดสรรที่เหมาะสมคือทำให้ marginal ROAS เท่ากัน ไม่ใช่สัดส่วนในอดีต และช่องว่างระหว่างสองวิธีนี้บนบัญชีทั่วไปอยู่ที่ 4-7 percentage points ของ net contribution margin ต่อไตรมาส

  6. 8 เมษายน 2026

    วิธีลดค่าโฆษณาเสียเปล่าบน Shopee โดยไม่ทำให้ยอดขายตก

    บนบัญชี Shopee ทั่วไป 20–30% ของค่าโฆษณาขาดทุนเชิงโครงสร้าง ทั้งที่แดชบอร์ดจัดอันดับให้เป็นแคมเปญที่ชนะ การหยุด "แคมเปญที่ทำผลงานไม่ดี" จึงพลาดจุดรั่วไหลจริง บันทึกงานวิจัยว่าด้วยค่า default เชิงโครงสร้างสองอย่างที่ทำให้เกิดค่าโฆษณาเสียเปล่าแบบซ่อนเร้น และวิธี audit ที่ดึงมันออกมาได้โดยไม่เสียยอดขาย

หยุดเดา ให้ DataGlass ช่วยเพิ่มกำไร

ร่วมกับผู้ขายที่ใช้ DataGlass เปลี่ยนข้อมูลร้านให้เป็น Action ถัดไปที่เพิ่มกำไรสูงสุด สำหรับโฆษณา ราคา โปรโมชั่น และสต๊อกสินค้า