A video can win attention and still lose money. It can also introduce a customer who returns through search days later, leaving a last-click report that understates the ad’s possible role. The answer is not to trust whichever dashboard tells the most flattering story. Read the entire customer path, reconcile it with the business’s own transactions, and state what the data can and cannot prove.
Judge the campaign by the intended customer action and the value retained after fulfilling it. Views, likes, and clicks help diagnose how people moved through the path. They do not by themselves show that the campaign created profitable demand. A good report keeps these layers separate.
This is part five of the TikTok Advertising Guide. The measurement article defines events and allowable acquisition cost; the testing article explains fair comparisons. For product sellers, the Shipping Profit Guide helps turn revenue into a more honest order-level result.
Read the funnel in the right order
Start with delivery: amount spent, campaign status, impressions, reach, frequency, and where the ad appeared. An ad with a low click count may have received little delivery. One with high frequency may be showing repeatedly to a small group. TikTok’s basic metrics guide distinguishes all clicks from destination clicks; use the measure that matches your question. Social interactions are not the same as visits to the offer page.
Next examine the transition from ad to page. Compare destination clicks with actual landing-page visits and the page’s own analytics. A large gap can indicate slow loading, accidental taps, blocked tracking, redirects, or a mismatch in definitions. Inspect the page on a real phone before changing the video. If people arrive but leave immediately, compare the video’s implied promise with the page’s product, price, and first screen.
Then follow the selected event: purchase, qualified lead, booking, or other action. Check whether it fired once, whether value and currency are correct, and whether the business record exists. A cost per conversion has little meaning if “conversion” is a button click that does not result in a submitted inquiry. TikTok’s custom-report guidance lets advertisers choose dimensions and metrics; build a report around the primary event and a few diagnostic steps instead of a wall of unrelated columns.
Finally, look past the recorded event. Was the order paid and delivered? Was the lead in a market you serve and did the sales team reach it? Did the customer return or request a refund? These outcomes are often outside the ad platform. Join them to the campaign through an order or lead record when you can, respecting privacy requirements and avoiding unsupported claims about individual attribution.
Understand what each rate actually divides
Cost per thousand impressions says what it cost to show the ad, not whether viewers were good prospects. Click-through rate compares clicks with impressions; ensure you know whether the clicks are destination clicks or all clicks. Cost per destination click divides spend by clicks to the specified destination. Landing-page conversion rate divides completed actions by actual visits. Cost per purchase divides spend by attributed purchases, which may not equal unique paid orders in the store.
These denominators matter. Suppose one video receives 100 destination clicks and 10 purchases, while another receives 40 clicks and six purchases. The second has a higher click-to-purchase rate; the first has more purchases. If both spent the same amount, the first may have lower cost per purchase. If the first produces more refunds, the apparent advantage can reverse. Do the arithmetic using the same dates and a clear definition for every count.
Use a diagnostic question for each metric. Low reach: is the audience or budget too restrictive? Views but few destination clicks: does the video invite the right next action? Clicks without visits: is the site or tracking failing? Visits without orders: are price, trust, product fit, or checkout the problem? Orders with weak margin: is the offer economically sound? This sequence prevents a team from endlessly editing the first three seconds when the real issue is shipping cost.
Reconcile platform reports with business records
Build a simple weekly table with spend, TikTok-attributed conversions, unique store orders or reviewed leads, canceled orders, refunds, gross sales, fulfillment costs, and retained contribution. Use consistent time zones and note whether the platform counts conversions by ad interaction date or event date. Document promotion codes and any offline sales process. If the platform reports 30 purchases while the store has 22 paid orders from the relevant period, investigate duplicates, test orders, rejected payments, attribution timing, and orders influenced by other channels.
Do not force the two systems to match exactly. An ad platform uses its attribution rules to credit interactions, while a store records transactions. TikTok’s Attribution Analytics guidance describes comparisons across attribution windows, time to conversion, and assisted conversions. Its overview guidance distinguishes TikTok-referred and TikTok-assisted paths. These reports can help explain a difference, but attributed credit is not the same as a causal estimate of additional sales.
View-through attribution is especially easy to overread. TikTok’s view-through guide describes conversions after someone views an ad without clicking within a chosen window. This can capture influence that click-only reporting misses, but the person may also have encountered email, search, or another ad. Keep click-through, view-through, and business-side results visible rather than blending them into one unexplained “sales from TikTok” number.
Turn revenue into a decision number
Return on ad spend is attributed revenue divided by ad spend. If an ad receives credit for $1,000 in revenue from $400 of spend, reported ROAS is 2.5. That is a useful efficiency ratio, but it omits product costs, shipping, fees, returns, and support. A 2.5 ROAS can be profitable for one offer and a loss for another. TikTok’s own ROAS guide distinguishes revenue efficiency from profitability.
Calculate retained contribution for the orders connected to the campaign: recognized revenue less refunds, product or service delivery costs, payment fees, shipping, variable support, and ad spend. For illustration, suppose a campaign brings $1,000 of paid orders, $100 is refunded, and the remaining orders cost $520 to provide. After $300 in ads, contribution is $80 before fixed overhead. These figures are fictional. If a store reports the original $1,000 as revenue and ignores refunds or fulfillment, it will make the wrong budget decision.
For services, use qualified leads, show-up rate, close rate, sales-cycle delay, and margin from actual clients. A lead that never answers the phone should not receive the same value as a signed customer. If only a few leads have had time to close, report the early signal and the incomplete sales window separately. Do not insert an optimistic assumed close rate into a final ROI claim without saying it is an assumption.
Ask whether the ads added demand
Attribution answers which interactions receive credit under a defined rule. Incrementality asks what would have happened without the ad. A surge in attributed orders during a holiday promotion might include existing customers who would have purchased anyway. A decrease in organic or email orders while ad orders rise might show some channel shifting. Neither observation alone proves cannibalization or lift; use them as reasons for a more careful test.
For advertisers with enough volume and suitable access, a holdout or conversion-lift study can estimate additional outcomes. TikTok’s Conversion Lift Study explanation describes randomized test and control groups. A small business may not have the scale for that method. It can still compare periods carefully, run a limited geographic test where appropriate, or look at total business demand alongside spend, while acknowledging seasonality and other campaigns. Avoid presenting a simple before-and-after chart as proof of causation.
Produce a report that leads to action
End the review with four sentences: what was spent, what customers did, what the business retained, and what remains uncertain. Include technical incidents, offer changes, and sample size. Then recommend one action: continue at the same level, run a specific follow-up test, repair the page or measurement, reduce spend, or stop. The report should make clear who owns the next action and when it will be reviewed.
Customer experience belongs in the result. Note delivery complaints, misleading expectations, refund reasons, and support workload alongside cost per order. A creative that brings cheap orders but repeatedly attracts the wrong buyers may damage the business. The Holiday Marketing for Dropshippers Guide provides an example of why campaign profit and fulfillment quality should be reviewed together.
The final article uses these results to decide whether to scale, revise, or stop the campaign without turning a promising week into uncontrolled spend.