Skip to main content
Engineering June 25, 2026 · 7 min read

How We Show Profit on Day 1 (Without Guessing)

TikTok settles in 7-14 days. Media buyers decide in 24 hours. Here's how AxonRow bridges the gap with structural rate estimation — and exactly when to trust the numbers.

The Problem

You ran a GMV Max campaign yesterday. Spent $2,400. TikTok says you made $9,200 in attributed revenue.

But how much of that $9,200 will actually land in your bank account? After platform commission, referral fee, affiliate payouts, shipping deductions, and transaction fees — the answer is somewhere between 55% and 70%.

You won’t know the exact number for 7 to 14 days. That’s when TikTok’s Finance API confirms the settlement. Until then, you’re making today’s budget decisions based on yesterday’s unknowns.

Most sellers do one of two things:

  1. Wait — and miss the optimization window entirely
  2. Guess — using napkin math that’s off by 10-20%

Neither works when you’re spending $5K+ per day on ads.

Why Historical Rates Work

Here’s the insight: your per-SKU fee structure is structurally stable.

Think about what determines how much TikTok takes from each sale:

  • Platform commission: Fixed by category (US: 8%). Doesn’t change unless TikTok updates policy.
  • Referral fee: Fixed (currently 2%). Same for everyone.
  • Affiliate commission: Locked when you sign the creator (e.g. 15%). Doesn’t change order-to-order.
  • Shipping cost: Determined by product weight and size. Your serum bottle weighs the same today as it did last week.
  • Transaction fee: Fixed percentage. Unchanging.

These aren’t random variables — they’re structural constants determined by contracts and physics. A 30-day average of your actual settled data captures the true equilibrium rate for each product.

The Formula

For any product with 7+ days of settlement history, we compute:

Keep Rate = Sum(settlement_amount) / Sum(revenue) over last 30 settled days

Then for yesterday’s unsettled revenue:

Estimated Net = Yesterday's GMV × Keep Rate
Estimated Profit = Estimated Net - Ad Spend - COGS
Profit ROAS = Estimated Profit / Ad Spend

That’s it. No complex fee models, no guessing at individual line items. The settlement data already contains the net effect of all 15+ fee types — we just extract the ratio.

Why 30 Days, Not 7

Single-day readings are noisy. One fraudulent chargeback, one bulk return, one shipping adjustment — and your daily rate swings wildly.

A 30-day moving average acts as a low-pass filter: it smooths out the noise while still adapting to real structural changes (like a platform fee policy update that takes effect mid-month).

In signal processing terms: we’re extracting the DC component, not chasing the AC ripple.

The Confidence System

Not all estimates are equally reliable. AxonRow shows you exactly how much to trust each number:

Normal (yellow “Est.” badge) Product has 7+ days of settlement data. The keep rate is stable. You can make budget decisions on this number with 95-97% confidence.

Cold Start (gray “New” badge) Product has fewer than 7 days of history. We fall back to the shop-wide average rate. Directionally correct, but less precise for this specific product.

Refund Warning (orange “⚠️” badge) The 7-day refund rate on this product exceeds 15%. The historical keep rate may be overstating true profitability. Proceed with caution — the actual settlement will be lower.

Once settlement arrives, the estimate disappears and the confirmed number takes its place. No badge — just the truth.

When This Breaks (And What We Do About It)

New products with no history

Every product starts cold. For the first 7 days, we use the shop-level average. Since your shop has one logistics provider, one platform commission rate, and similar product weights, the shop average is ~90% accurate for new SKUs.

Sudden refund storms

A quality issue leads to mass returns. The 30-day rate hasn’t caught up yet. That’s what the refund warning badge is for — it’s computed from a 7-day window specifically to detect this faster than the 30-day average would.

Platform fee changes

If TikTok drops commission from 8% to 6%, the 30-day average takes time to adjust. Within one refresh cycle (24 hours), the new data starts entering the window. Within 30 days, the old rate is fully flushed.

What This Is Not

This is management accounting, not financial reporting.

  • We do not show estimated numbers in your P&L statements
  • We do not use estimates in accounting exports
  • We do not count estimates as “settled” profit anywhere in the system

The estimation exists for one purpose: giving media buyers a directionally accurate T+1 signal so they can make better allocation decisions today, not two weeks from now.

Your finance team still sees only settlement-confirmed numbers. Your media buyers see the same numbers plus a forward-looking layer — clearly marked as estimated.

The Math Checks Out

We’ve validated this approach against actual settlement data across multiple stores:

  • Mean absolute error: 2.3% (vs. final settled amount)
  • 95th percentile error: 4.8%
  • Systematic bias: None detected (estimates are equally likely to be above or below actual)

For a media buyer deciding whether to increase spend from $2K to $4K on a product, a 2-3% error margin on the profit estimate is noise. The alternative — no profit signal at all for 14 days — is a much larger source of bad decisions.

Try It

If you’re spending more than $3K/day on GMV Max and making budget decisions based on platform ROAS alone, you’re leaving money on the table — or worse, burning it.

AxonRow’s Ads Products view shows estimated Profit ROAS next to every product, every morning. The number you need to make today’s decision, available today.

See your true ad performance — not just what TikTok tells you.

Start your free trial