1. Write down the question and baseline
Choose one outcome, the observation dates and timezone, and the change you intend to test. Compare complete windows of equal length and similar posting cadence.
Record promotions, deliveries and unusual posts separately. This protocol is our proposed analysis method, not an official TikTok benchmark.
2. Keep platform counters separate
Record available views, profile visits and new follows for the same period. Ratios between aggregate counters are descriptive, not proof that the same people took each step.
Missing metrics remain unavailable. We do not apply universal healthy-account scores or fixed completion-rate thresholds.
3. Count website visitors, not repeated page loads
Our internal TikTok report deduplicates first-party visitor identifiers after administrator and known-bot filtering. A person changing browsers or clearing storage may still appear more than once.
Sessions and page views are auxiliary units, not additional customers. Consent, blockers and incomplete records can reduce observable traffic.
4. Separate assisted checkouts from verified purchases
A generic checkout after a TikTok page in the same session is an assisted milestone, not proof of a TikTok purchase. Paid attribution requires an actual paid order containing a TikTok deliverable and the exact stored session or visitor identifier, with a preceding visit.
We do not use a matching IP or email to claim acquisition causality. Unmatched paid orders remain visible separately.
5. Label mixed orders and time windows
An order containing TikTok and another platform is one mixed order. Its whole-order total is not TikTok-only revenue.
Our report separates currencies and uses payment time in the selected Berlin date window; its PAID-only table excludes currently refunded orders and is not a net-revenue statement. Visits outside the chosen window are not backfilled into the displayed cohort.
6. Do not confuse AI clicks with citation frequency
A ChatGPT referral is a recorded visit, not a census of recommendations. A crawler fetch is not a user referral.
Repeated, dated question tests can record whether a page was cited under the tested conditions, but cannot establish how often all users see it. No measured uplift or customer outcome is claimed on this page.
7. Decide whether spending addresses the goal
More purchased views or followers cannot establish organic reach, sales or program eligibility. Use the selector to distinguish a profile quantity from a video metric, or choose no purchase.
Current product pages, not this article, define prices and order terms.
Sources & methodology
Official sources, research and market data used for this guide.
- OfficialTikTok SupportHow TikTok recommends content
- OfficialTikTok SupportTikTok Studio
Separate measurement, advice and commercial services.
Meta describes likes, comments, saves, and other actions as signals whose likelihood Instagram predicts for Feed ranking; Meta does not publish one fixed universal weighting.
Compare one clearly labelled formula over a defined period, and separate content formats and organic from paid distribution.
BoostMaus offers visible likes, views, comments, saves, and reposts for supported public content.
Additional interactions do not guarantee Explore placement, virality, recommendation eligibility, sales, or organic growth, and they do not replace organic Insights.