Why this matters
The advanced social media level judged creator work on what it caused. At expert level it arrives about whole channels, from people who already hold an answer; the job is to decide, before any result, which evidence settles the budget and what each report may claim. The guide to measuring what TikTok adds beyond last click sets out that evidence for a paid channel; the guide to measuring YouTube's contribution to the business does it for an organic one.
The tackle retailer and its two channels
Heithosk Tackle is an invented online retailer of fishing rods, reels and bait that also takes phone orders, keyed in with a delivery postcode.
Heithosk Tackle: the reports on the table
TikTok, monthly average last quarter, at £8,000 spend and a £65 average order:
| View | Orders credited to TikTok | Revenue ÷ spend |
|---|---|---|
| Last click in web analytics | 70 | £4,550 ÷ £8,000 = 0.57 |
| TikTok click-through, account's chosen window | 150 | |
| TikTok view-through, account's chosen window | 330 | |
| TikTok's reported total | 480 | £31,200 ÷ £8,000 = 3.9 |
YouTube: one unpromoted long video a week. Description links have carried campaign tags since the quarter began; pinned comments and the channel page link have not.
| YouTube, last quarter | Count |
|---|---|
| Tagged YouTube sessions | 1,800 |
| Orders, last click from those sessions | 22 |
| Further orders assisted by a tagged YouTube session | 31 |
| Untagged sessions referred from YouTube | 900 |
- Orders are up 34% on the same quarter last year. TikTok ads started 9 months ago; the channel is 2 years old.
- One order in 5 arrives by phone. Customers are never asked how they found the shop, and branded search goes unread.
- 12 delivery regions form 6 pairs whose orders moved together over 2 years; TikTok can target each one.
- Contribution after goods, delivery and fees: 40% of revenue. The budget is decided in 14 weeks, before spring.
The finance lead: "TikTok returns 0.57. Halve it." The paid social manager: "It returns 3.9. Double it." The owner: "Sales are up a third and the channel is the only new thing we did."
Credit, signals and tests
Keep three kinds of evidence apart. Credit is what a report assigns; neither last click nor TikTok's column measures IncrementalityConversions caused by marketing.. Signals such as Branded SearchSearch queries that contain a company's or product's brand name. and customers' answers corroborate without proving. Tests withhold the channel and compare, through a geo holdout or a platform conversion lift study. An organic library stays findable, so YouTube is reported as a range with the missing test named.
Task
- Say what each view can and cannot show, and answer the three claims.
- Decide how view-through credit appears in the monthly report.
- Choose the outside signals to start this week.
- Choose a geo holdoutA test that runs a campaign in some regions and holds it back in comparable ones, then compares total outcomes, where people cannot be randomised one by one. or a TikTok lift study for the budget, with the decision rule written first.
- Say what stays unchanged while the test runs, YouTube included.
- Write this quarter's YouTube report, and say what next quarter adds.
Worked answer
What each view shows. Last click is blind to an ad seen but not tapped, so it cannot tell what TikTok adds. TikTok's total credits buyers already on their way and cannot see phone orders, so it cannot tell either. The owner forgets that TikTok ads started the same year. None of the three settles the budget.
View-through on its own line. Most of TikTok's total is View-Through ConversionA conversion credited to an ad that was shown but not clicked, within a set view window. credit, the part a test must confirm. The report keeps last click, click-through and view-through as separate lines, never summed, with the attribution setting and its date in the measurement plan and the test that checks it named beside it.
Signals to start now. An open how-did-you-hear question at checkout and in the phone script, coded with separate TikTok and YouTube codes and stored beside the tracked source, as the self-reported attribution guide sets out. A branded search baseline taken this week, nationally and by region. Tags on pinned comments and the channel page link. Keep each fixed: the trend is what gets read.
The test. A geo holdout. No pixel sees phone orders, a contested budget line needs evidence from Heithosk's own order system rather than the party being judged, and the regions pair well. Pause TikTok in one region of each pair, read web and phone orders by postcode, and work through the holdout test readiness checklist first. A lift study follows later at similar spend, showing how much of the effect TikTok sees, so cheaper lift studies can carry routine checks. Write the rule from the margin now, using the ROAS calculator.
Heithosk Tackle: the arithmetic fixed before the holdout
- View-through share of TikTok's total: 330 ÷ 480 = 69%.
- Break-even revenue return at a 40% contribution: 1 ÷ 0.40 = 2.5.
- Incremental revenue needed at £8,000 a month: £20,000, about 308 orders at £65.
- Low end of the interval at or above 308 incremental orders a month at full spend: keep the budget and repeat the holdout next year.
- Interval straddling 308: hold spend and extend the test, or repeat it at the lower spend.
- Whole interval below 308: make the finance lead's cut and test the lower spend next season.
Held still while it runs. The attribution setting, other budgets, prices, promotions and YouTube's rhythm stay unchanged. Branded search in holdout regions against TikTok regions then shows TikTok's share of any rise; only what every region shares is left for YouTube, the season and everything else.
The YouTube report. This quarter it claims a lower bound, with channel health and the gaps named. Next quarter the coded answers add the upper bound: customers who named YouTube but arrived another way. End each report with a decision on formats or investment.
Heithosk Tackle: the YouTube report this quarter
| Layer | This quarter |
|---|---|
| Tagged, last click | 22 orders |
| Including assisted journeys | 22 + 31 = 53 orders |
| Named YouTube but arrived untagged | Not yet measured: the question starts this week |
| Costs | Owner's time, recorded from now |
Gaps stated: 900 untagged YouTube referrals, app clicks arriving as direct, phone orders the site cannot see, no customer answers yet and no holdout, which an organic library rarely allows.
Common mistakes
- Choosing the method after the result. The test and its rule come first.
- Summing the TikTok lines. View-through stays apart until tested.
- Running the lift study and the holdout together. Each contaminates the other.
- Reading the holdout on web orders alone. Phone orders were the reason for it.
- Handing YouTube the branded search rise. TikTok moves it too.
- Reporting YouTube as one number. Give bounds and the gaps.
Test these judgements in the checkpoint on evidence for what social channels add.
