A new AliExpress product got 8 orders on day 5, and the operator judged it “validated,” raising the daily budget from 30 yuan to 120 yuan that same day. By day 9, ROAS dropped to 0.8; by day 14, Zhutongche had spent four figures and closed 12 orders, with 400 units of inventory piling up in the warehouse. The problem is not media-buying technique—a sample size of 8 orders is simply not enough to prove this product is worth scaling.
Scaling up too early is more common than not spending aggressively enough, and it burns more money. The cadence chart below divides stages by cumulative data volume, not by days: after running for the same 7 days, 2000 impressions and 2 ten-thousand impressions are two completely different positions.
The Four-Stage Cadence Chart: Only One Goal per Stage
The thresholds in the table are suggested guidelines and must be calibrated against the baseline of your own category and site. Average click-through rates across different AliExpress second-level categories can differ by three times; directly copying someone else's numbers is just another way of guessing.
| Stage | Entry Threshold (Cumulative) | Sole Goal | Promotion criteria | Abandon threshold |
|---|---|---|---|---|
| Product testing period | Impressions ≥ 5000, clicks ≥ 150 | Obtain a click-through rate and add-to-cart rate sufficient for judgment | Click-through rate not below the category average, add-to-cart rate ≥ 5% | Click-through rate still significantly below the category average after 8000 impressions |
| Validation period | Cumulative 30 add-to-carts + first order | Confirm add-to-carts can convert into paid orders | Add-to-cart to first-order conversion ≥ 20%, no signs of same-reason returns | No organic orders after 60 add-to-carts, or 2 or more same-reason returns |
| Small-step scaling | Cumulative ≥ 20 transactions, and orders on 5 consecutive days | Verify that marginal cost does not rise after a budget increase | After a single increase of ≤ 30%, ROAS does not decline for 3 consecutive days | After a budget increase, ROAS falls below the break-even line for 2 consecutive days |
| Large-scale scaling | Stable daily orders; inventory and logistics are in place | Scale up without letting costs spiral out of control | Marginal conversion cost no higher than 1.2 times the average | Weekly marginal cost higher than 1.5 times the average |
Look at the second column. The stage is determined by data volume, not by the number of days. Increasing budget at 2000 impressions and only increasing budget at 2 万 impressions are two fundamentally different decisions; if you draw conclusions before the data arrives, all subsequent scaling actions are just gambling.
How to set the advancement criteria: only when the three numbers are in place are you allowed to move to the next stage.
- During product testing, check whether the denominator for CTR is sufficient.When impressions are under 5000, day-to-day fluctuations in CTR are meaningless—3% one day and 6% the next just means the traffic mix is changing. When the sample is insufficient, the action is to increase impressions: change keywords, adjust bid tiers, and replace the main image with a version that makes the comparison clearer—not adjust price, because price changes also alter the conversion denominator and make judgment harder. If impressions are sufficient but CTR is still low, cut it directly; don't try to rescue it by lowering the price.
- In the validation phase, look at the conversion from add-to-cart to first order, as well as signs of returns.Many add-to-carts but few first orders means you're stuck on price or shipping. Check the shipping template and final price first, then consider coupons; if there are two or more returns for the same reason (size, material, color difference), immediately stop advertising and go fix the Listing—don't use ad spend to suppress the return rate. If add-to-carts reach a cumulative 60 and there are still no organic orders, there's no need to wait any longer on this product.
- In the scaling phase, look at marginal conversion cost, not total ROAS.The metric definition is: incremental spend after this round of budget increase ÷ incremental conversions. Total ROAS gets diluted by earlier low-cost traffic, so it looks very healthy, while the incremental portion may already be losing money. If the incremental metric deteriorates, stop first, rather than continuing to increase.
How to add budget during the scaling phase: tiers, observation windows, and rollback lines
There are two hard rules for increasing budget: a single increase must not exceed 30%, and after increasing, observe for at least 48 to 72 hours—that is, one complete conversion cycle. Do not touch the budget again within the window just because there are no orders for half a day; repeated adjustments will make the system relearn, and the basis for judgment will disappear.
When a rollback signal is triggered, the first step is to reduce the budget back to the previous tier, not to pause keywords first. Pausing keywords will lose the promotion weight already accumulated, and the cost to recover is higher than reducing budget. Only when a single keyword's independent loss exceeds 30% of the campaign's total spend should you pause it separately.
Conversely, if ROAS does not decline for three consecutive days after increasing by 30%, you can move up another tier. The rhythm of scaling is to move up in steps, not to double the budget all at once—after doubling, what you get is not growth, but a bill you can't attribute.
How to reuse the cadence table when running multiple stores and platforms in parallel
The skeleton is universal, but the criteria need to change by platform. AliExpress store operations mainly look at on-site advertising data and dispute rate; a rise in dispute rate directly affects the stability of promoted placements; Shopee store operations and Lazada store operations need to add shipping timeliness and negative reviews, and one late shipment can interrupt an ongoing scale-up; Mercado Libre store operations must be viewed by site, because conversion baselines differ across country sites, and criteria cannot be directly applied across sites; Wayfair store operations, meanwhile, need to include return costs in the marginal cost calculation, as its return processing cost accounts for a notably higher share of total costs, and looking only at ad spend will lead to misjudging profit and loss.
The real breaking point for parallel multi-store operations is not in advertising. If AliExpress multi-store management, Shopee multi-store management, Lazada multi-store management, and Mercado Libre multi-store management share the same login environment or proxy exit, increased CAPTCHAs, secondary verification, and risk control prompts will directly interrupt the first 7 days of a new product—and these seven days are exactly the most critical window for accumulating samples during the product testing period. Isolate the environments first, then discuss scale-up; for specific methods, refer toHow to manage multiple e-commerce accounts: key points for login environment, sub-accounts, and operation log settingsthe four-layer criteria in it.
There is another discipline for reusing the cadence table across stores: only 1 to 2 products may be in the small-step scale-up stage at the same time, and all others must remain in the product testing period. If multiple products are scaled up at the same time, budget and operational attention will be diluted, and ultimately no single product's data can be read clearly. Wait until a single store is proven and the criteria are stable before replicating to multiple stores; do not reverse the order. For the specific stage breakdown, seeHow to operate multiple e-commerce stores? Stage tasks from single-store validation to multi-store replication. If you are worried that problems will first arise at the account level,Multi-Store Management SOP from Account Isolation to Fund Consolidationcan fill in the two parts on risk control and funds.
Execution: At the start of each workday, check these three lines first
- Current stage: Mark which box each active product falls into: testing, validation, small-step scaling, or full-scale scaling.
- What is still missing from the promotion criteria: If impressions are lacking, change keywords; if add-to-carts are lacking, change the main image and price band; if sales volume is lacking, check shipping costs and final price. Write specific items, not “observe further”.
- Whether to add or stop today: Only adjust the budget when a promotion or rollback signal has clearly been triggered; in all other cases, do not touch it.
The weekly review answers only one question: has there been a stage shift this week? Products with no shift have only two paths—add samples, or cut them via the abandonment line. Staying put is the most expensive option: it ties up inventory, budget, and your attention at the same time. As for whether the cadence table should run in a self-built spreadsheet or a unified workbench, that depends on the number of stores; for how to decide, seeHow to Choose Between a Self-Built Spreadsheet, Multiple Browser Instances, and a Unified Workbench; if you are an agency taking over someone else's store, first followAgency Operation vs. In-House Teamthe criteria in it to calculate costs clearly, then decide whether to take on active products with messy historical data.
FAQ
Impressions keep failing to rise—should I just give up?
First distinguish whether you're not getting impressions or not getting clicks. The former is mostly a problem with keywords and bid tiers: switch to a new batch of long-tail keywords and raise the bid to a level that can get you onto the first two pages, then run for three days; the latter is a problem with the product or the main image. If you can't even get impressions, then talking about elimination means eliminating keywords, not products.
The platform's major promotion has compressed the cycle. Can we skip the validation period and scale up directly?
No, but it can be shortened. During the major promotion, traffic is cheap, and the validation period's sample accumulation speed will be much faster. The criteria don't need to change; what changes is when you see the criteria. The risk of skipping the validation period is: after the major promotion ends, real conversions fall back, and you have no historical data on hand to judge whether this product itself works.
With 4 stores testing products at the same time, how should the cadence schedule be arranged?
Prioritize rather than spread evenly. Set a budget cap for each product, first run until the impression threshold is met, pick out the products with sufficient data to enter the validation period, and keep the rest running on a low budget. The only two stages that truly need human monitoring are the validation period and small-step scaling; the product testing period can be handled in batches.
After scaling up, the add-to-cart rate has dropped. Is it a product problem or a budget problem?
Look at the source structure of the newly added traffic. After increasing the budget, if the system pushes traffic toward a broader audience, the drop in add-to-cart rate is normal dilution; focus on whether the marginal conversion cost has exceeded the line. If the source structure hasn't changed but the add-to-cart rate has dropped significantly, it's more likely that the price or inventory status changed during the scaling-up period.
When an agency operations team takes over a store that has already run ads, from which stage should the cadence start?
Don't restart from scratch. First do a data health check: re-aggregate the historical ad data by impressions, add-to-cart, and conversions, and see which stage the current product is actually in. For the parts that cannot be aggregated, just rerun a round according to the product testing period; it's faster than making judgments on messy data.
