FARMTest TikTok hooks and decide which formats to scale
Run a six-post creative test, diagnose where viewers leave, and decide whether to improve a hook, expand a format or stop producing it.
Choose one creative decision before making the next batch: a stronger opening, a clearer demonstration or a different format. Keep enough of the video consistent to learn which change deserves more production time.
This guide uses an interview-practice app as an example. The scripts and numbers are teaching examples; use your own product and account results.
Start with a six-post hook test
Suppose the app lets people record and replay an interview answer. Compare two openings:
- Feature first: “Record your interview answers and play them back.”
- Problem first: “Know your answer, then freeze when you say it aloud?”
Both lead into the same task: answer one question, replay it, improve one sentence. Make three original demonstrations for each opening, using different interview questions. Keep the audience, offer, approximate duration and product action similar.
| Pair | Demonstration topic | Version A | Version B |
|---|---|---|---|
| 1 | Introduce yourself | Lead with recording feature | Lead with freezing mid-answer |
| 2 | Explain why you want the role | Lead with replay feature | Lead with an answer sounding vague |
| 3 | Describe a setback | Lead with practice workflow | Lead with losing the thread |
This is a practical starting batch, not a required sample size. Schedule around the account's usual operating windows and compare posts at the same age—for example, seven days. Record the first frame, opening line, post link, average watch, relevant questions and production time.
If using two accounts serving similar audiences, rotate the assignment: account A gets feature-first then problem-first; account B gets problem-first then feature-first. Compare within each account before combining results. Keep unrelated language or audience groups separate.
Read the result without letting one hit decide
Here are example results for six 24-second videos:
| Pair | A views | A average watch | B views | B average watch |
|---|---|---|---|---|
| 1 | 800 | 7.2 seconds | 900 | 9.0 seconds |
| 2 | 1,000 | 8.0 seconds | 1,100 | 9.5 seconds |
| 3 | 6,000 | 7.5 seconds | 1,300 | 9.2 seconds |
A wins total views, 7,800 versus 3,300. B holds attention longer in every pair. Its median post-level average watch is 9.2 seconds versus A's 7.5. Average watch divided by video length is different from the percentage of viewers who finish.
Next move: make another problem-first batch and inspect A's unusually successful third post for a transferable idea. These six posts suggest a direction; they do not establish a causal winner. If B also attracts more specific product questions at similar production cost, that strengthens the practical reason to continue.
Turn an observation into the next test
Watch the video beside its retention curve when available. Use the point of confusion to choose an edit.
| What you observe | Inspect | Next test |
|---|---|---|
| Viewers leave near the opening | First image and first sentence | Show the problem immediately; remove the logo introduction |
| Viewers leave during explanation | Repeated lines or an unclear screen | Cut one repeated point; enlarge the important app action |
| Viewers leave before the answer | How long the payoff takes | Show the result earlier, then explain it |
| Strong viewing, mostly unrelated jokes | Audience and product connection | Keep the format; replace the topic with a buyer's real problem |
| “What app is this?” repeats | Whether name and next step are visible | Show the name/icon during the demonstration; answer genuine questions |
| Several posts repeat familiar advice | Topic freshness | Use the same structure to answer an adjacent question |
| Posts across unrelated formats drop together | Publication, reporting freshness and platform notices | Check those first, then compare against each account's normal results |
For a production-style test, compare a continuous camera demonstration with an edited version using overlays. Keep the question and publishing route consistent; compare comprehension and production effort as well as viewing.
Low views can hide a content problem
In our team's farming playbook, An Nayal (@consumerxai) reports returning to seven accounts abandoned after roughly 300-view posts. With improved content, including new hooks and pacing, the same accounts reportedly performed normally again. This is an operator report, without controlled before-and-after data.
The lesson is useful: review recent formats and try a stronger explanation before writing off an accessible account. A fixed view count cannot distinguish weak creative from an account restriction. Investigate an actual platform notice through the account troubleshooting guide.
Manage a format's lifecycle
| Stage | Your decision |
|---|---|
| Discovery | Collect repeated examples and audience questions; select a candidate with VSC |
| Validation | Run a small original batch and repeat the promising direction |
| Expansion | Produce more distinct questions within the winning structure; track whether cost and results hold |
| Fatigue | Slow new production when comparable results weaken and the substance repeats |
| Refresh | Test a new angle, opening or demonstration that fixes a specific weakness |
| Retirement | Stop commissioning that execution and keep the useful learning for the next format |
For the interview app, a “mistake → replay → better answer” series could expand across different interview questions. If every episode repeats “speak more confidently,” refresh with a visible technique: shorten the opening sentence or replace a vague claim with an example. Retire the execution if the refresh produces little useful response for its cost; that decision does not require abandoning the account.
Publish the batch and make the decision
Name exports by topic and hook, such as setback_problem-first_v1.mp4. Upload them to PhoneFarm's Videos library, assign the account and phone, and confirm each published result. Keep unpublished attempts outside the creative comparison.
Finish the review with one sentence: “Repeat problem-first hooks for three new questions; replace the long introduction; review at seven days.” Put that decision into the next production calendar.