GMV Max, and What It Does With Your Budget

GMV Max is TikTok's automated campaign type: you give it a budget and an objective, and the platform decides the placements, audiences and bids. It works when there is enough conversion signal for it to learn from, which means products that already sell; on a product with a handful of sales it spends the budget anyway, because that is what it was asked to do, and shows little for it. What you still control is the budget, the only hard limit, the objective, the catalogue, meaning which products at what price with what stock, and above all the creative, because automation decides who sees a video and cannot make the video better. Judge it on sales it brought that would not have happened anyway, not on the totals it reports.

A fair test runs it beside the manual campaigns on products that already sell, for long enough to collect real orders, and keeps the manual ones where they still deliver sales the automation does not.

Most verdicts on automated campaigns come from tests that could not have given a fair answer. A test set up with the decision rule written in advance gives one either way.

Choose products that already sell. Pick a handful with steady sales over the last month or two, enough stock for the test period and a margin that can carry advertising. Leave launches and slow sellers on manual campaigns.

Write down the baseline. For the chosen products and for the account as a whole: total sales, advertising spend, organic sales and blended advertising cost of sales over the previous four weeks.

Set the target from the break-even. Work out what return each product needs to cover its costs after TikTok Shop's fees and any creator commission, and set the campaign's target at or a little above it. A target far above break-even starves the campaign; one below it buys sales at a loss.

Fix the budget and the dates. A budget the business can afford to spend learning, and a test long enough to pass the learning period and then run several more weeks. Do not change either during the test.

Decide the rule before starting. For example: keep the automated campaign if account level sales rise and blended advertising cost of sales stays within an agreed band, with organic sales holding. Otherwise return to manual.

Read it at account level. Compare the test weeks with the baseline, product by product and in total. The campaign's own report is a supporting detail, not the verdict.

An automated campaign chooses which videos to show and to whom. It can only choose from what it is given, so the creative supply is the input that decides most of the result.

Volume and variety. A campaign with two videos has little to choose between and wears both out quickly. One with a dozen, covering different objections and openings, can find the combinations that work for different audiences. Aim for a steady flow of new videos for each product rather than one perfect one.

Every video answers something. The same rule as organic content applies: each video answers one objection a buyer has, opening inside the moment. A video made only to give the campaign more to choose from adds noise rather than options.

Creator content counts. Videos by creators who already sell the product often outperform the brand's own, because they do not look like advertisements. Where creators agree to their content being used in the brand's advertising, it widens the campaign's options considerably.

Refresh before fatigue. A video shown repeatedly to the same audience loses effect. Watch for a winning video's results declining over several weeks and have the next set ready before it does, rather than after.

Learn from what it chose. The campaign's reporting shows which videos it spent on and which sold. Those are the openings, formats and objections to make more of, which is the most useful thing an automated campaign teaches a brand. The video formats that sell covers how to make them.

An automated campaign learns from a stable set of conditions. Every large change to its budget or target asks it to learn again, and a campaign changed every few days never finishes learning at all.

Change rarely. Decide the budget and target, then leave them for at least a week or two unless something is clearly wrong. Daily adjustments in response to a good or bad day mostly react to noise.

Change gradually. When a campaign is working and deserves more budget, raise it in steps rather than doubling it overnight. A sudden jump sends the campaign looking for buyers further from the ones it has learned to find, and results often dip while it adjusts.

Move the target with the margin. When a product's costs or price change, its break-even changes, and the target should follow. A target left unchanged after a price cut is a target that now loses money on every sale.

Plan the peak weeks. Competition for attention rises sharply in the busiest weeks of the year, and costs rise with it. Decide in advance whether the campaign's target relaxes for those weeks to keep volume, or holds and accepts lower volume, and make the change once rather than daily.

Keep a log. Each change, its date and the reason, beside the weekly results. When a campaign's performance shifts, the log shows whether the cause was a setting, the creative, the stock or the market.

A campaign treated this way is managed by its inputs, the creative, the catalogue, the budget and the target, rather than by watching its daily numbers.

GMV Max spends against creative, so the video formats that actually sell decide what the budget has to work with. Where advertising sits among the other things that move units on the platform is in how to sell on TikTok Shop.