Before adding the rows in a Custom Conversion report, ask whether those rows describe separate events or separate labels attached to events. If the labels can overlap, their sum can be larger than a deduplicated total without any activity being lost from the report.

TikTok's reporting guide, updated in October 2026, makes this distinction explicit. A purchase can qualify under both a new-customer rule and a high-value rule. It can therefore appear under more than one Custom Conversion name or ID, while the aggregate metric removes overlap. Total Custom Conversion Value is also deduplicated across overlapping rules. The detail rows are not necessarily independent portions of the total.

For international marketing teams working with a Chinese supplier, this is a reporting-definition problem before it is a campaign-performance problem. Resolving it starts with the unit being counted, not with an attempt to make every exported column add up.

Preserve labels without counting the event twice

Many familiar reporting dimensions encourage an additive habit. Separate dates or exclusive categories often look like pieces of a larger total. Custom rules can behave differently because they describe attributes an event can possess at the same time. Being a new-customer purchase does not prevent the same purchase from also meeting a value condition.

The detail remains useful. A marketer may want to understand which kinds of events meet particular conditions, or where those conditions overlap. Deleting useful rules just to force the rows into an additive table would remove that information. The alternative is to label the view honestly: rule matches are not the same reporting object as unique qualifying events.

This distinction should survive the handoff into an internal dashboard. A column copied from an advertising report should not become “new leads” merely because that is the sales team's preferred label. An event count does not establish a count of distinct contacts, and distinct contacts do not establish a count of qualified purchasing organisations. Each question needs the appropriate identity and business evidence.

The review can be modest. Keep the metric name, breakdown dimension and attribution period beside the figure. Note whether the aggregate deduplicates overlapping rules. When the question is how many different qualifying events occurred, use the relevant aggregate instead of summing rule rows. When the question concerns the characteristics of those events, retain the separate rules and their overlaps rather than hiding them.

Give the analyst receiving the export the rule definition, not just its display name. They need to know which conditions can overlap before building a total. The sales team can then add its own qualification evidence without rewriting the advertising event as a different business object.

Check which collection each interface observes

TikTok also distinguishes Ads Manager from Events Manager. Ads Manager reports Custom Conversions attributed to advertising under the selected attribution settings. Events Manager shows event activity received from the connected source, which can include activity not attributed to a TikTok ad. A difference between those interfaces can therefore reflect different collections rather than a missing transaction.

The same caution applies to the Conversions and Custom Conversions columns. The former follows the optimisation conversion event; the latter follows configured custom rules. A rule may select a more specific outcome than the campaign's standard event. The guide also notes that processing or reporting delays can produce smaller timing differences. First establish whether the definitions align, then investigate a residual discrepancy.

Clicks introduce another unit. One ad click can be followed by multiple qualifying events within the applicable attribution window. TikTok explains that Custom Conversion Rate based on clicks can consequently exceed 100%. That is not a statement that more than every unique buyer converted. The denominator is clicks, and the numerator can include several events associated with a click.

This is adjacent to, but distinct from, aligning optimisation events before pooling campaign budgets. Budget eligibility concerns what the system is trying to obtain. Report reconciliation concerns whether several rows describe the same obtained event. Consistent campaign goals do not make every later reporting breakdown mutually exclusive.

When a total looks unexpectedly low, identify one set of rules over one consistent scope and ask whether an event can satisfy several of them. Check whether the comparison uses received or advertising-attributed activity. Only after those definitions are aligned does an unexplained difference become a useful candidate for technical investigation. Adding a manual adjustment before that point would conceal the measurement question rather than resolve it.

Sources