How to Test TikTok Selling Pricing and Discounts: Don't Only Look at GMV

First define the problem to be solved this time

GMV is gross merchandise value, not equal to profit. A price cut may increase orders, and may also change the product mix and after-sales pressure.

Before starting, write down the target audience, content or campaign scope, statistical time zone, and observation period. If Shop, showcase, ads, or LIVE are involved, also record the market, account role, and feature entry point; these capabilities may differ by region, eligibility, and backend version, and you cannot treat someone else's interface as your own available conditions.

What judgments official materials can support

TikTok distinguishes click attribution from view attribution, and provides attribution metrics such as Shop assisted browsing, add to cart, purchase, and revenue; Ads Manager distinguishes impressions, reach, all clicks, destination page clicks, conversions, and different conversion rate definitions. [Source 1] [Source 2]

These materials are used to confirm feature and metric definitions, and do not provide "follow this and growth is guaranteed" or a fixed traffic threshold. The process in this article is an execution framework organized from public definitions, and actual conclusions still need to be verified by your own content, products, account, and order records.

Execute in three steps

1. First define the profit floor: record the current state and constraints, and save pages, creatives, or data snapshots that can be reviewed later.
2. Record the original price and actual payment: adjust only one main variable at a time; if theme, budget, product, and publish time are changed at the same time, it will be difficult to explain the results later.
3. Observe refunds and repurchases after the campaign: review during the same observation period, and write the conclusion as "what to keep, what to stop, and what to verify in the next round."

Recommended data table to keep

Counts, people, ratios, and attribution results must be listed separately. Any ratio you calculate yourself must state the numerator, denominator, and window; fields that are unavailable should be marked directly as "unavailable," and do not fill them in with other numbers.

Hypothetical case: how to write a conclusion without exaggeration

Sales rising from 10,000 to 14,000 does not automatically mean the campaign was better; if discounts, ads, and refunds cost an extra 5,000, you must return to the complete revenue table to judge.

This case only demonstrates the judgment method, and is not a customer case or industry benchmark. When the sample is small, traffic sources change, or feature fields are missing, the conclusion should be written as "needs continued observation," and do not write correlation as causation.

Common misconceptions and checking order

• First check whether the account, content, product, or campaign has a clear status prompt.
• Then check whether the region, time zone, observation period, ad delivery, and external traffic are consistent.
• Separate content performance, product information, fulfillment, and business results to avoid using one total to explain all problems.
• When pricing, inventory, authorization, personal information, or platform policy are involved, the current backend and official terms shall prevail.
• Keep the original records, and in the next round verify only the single most likely and actionable cause.

Frequently asked questions

Is it enough to only look at sales when reviewing a discount campaign?

Not enough. You also need to combine order count, average order value, discount cost, ad spend, refunds, fulfillment cost, and new vs. returning customer structure. Different campaigns must be compared under the same attribution and statistical time zone.

When can a conclusion be drawn?

When the comparison objects, time range, field definitions, and main conditions are basically consistent, and multiple comparable samples have been accumulated, a temporary conclusion can be formed. The scope of application should still be noted; one performance result, one screenshot, or one popular case is not enough to prove a long-term pattern.

Continue reading

You can view the same series of TikTok data, content, and e-commerce methods in the 96SMM Operations Guide. When account features, product eligibility, and ad settings are involved, the current region and account display shall prevail.