More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+22 more

    TikTok Account Review

    1

    Audit your own TikTok from TikTok Studio analytics: KPIs vs your own baseline, what works, and 3–5 fixes as pass/kill tests.

    $9.99

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    TikTok Account Review

    TikTok Account Review

    Example session with this skill installed

    Review my TikTok account. I've pasted the analytics for my last 26 videos below. I have about 2,400 followers, post 3–4 times a week, and I sell handmade ceramics.

    • Read your context and instructions
    • Compiled the tiktok account review

    TikTok account review: ceramics brand

    Account: small home-goods brand (ceramics), 2,400 followers ·

    Period: 18 Aug – 29 Sep 2026 ·

    Posts: 26 (23 in the baseline, 3 provisional, younger than 7 days) ·

    Benchmark tier: 1–5K followers ·

    Data route: template (your CSV) ·

    Review date: 2 Oct 2026
    Sample label: full read (23 baseline posts, 21 videos and 2 photo carousels).
    Missing fields: non-follower share not supplied; person on screen, direct address, audio, captions, first frame and series not tagged, so those comparisons were skipped. The hook uses the true 3-second hold (your template has it), not the coarse "stopped at" second.

    1. Verdict

    What's working: real filmed products. Real footage (n=8) vs template mockup scenes (n=15): still watching at 3 s 35.0% vs 23.0%, save rate 0.97% vs 0.29%, follows per 1,000 views 3.1 vs 0.94 [Own data].
    2. The biggest leak: the hook. Your median 3-second hold is 24.6%: about 3 in 4 viewers are gone before second 3.
    3.

    The #1 fix: a test of 6 real-footage posts against 6 template posts on the same products. Pass: real footage wins on 3-second hold

    and save rate in at least 4 of 6 pairs.

    2. KPI scorecard

    KPI (plain words)Your medianYour p75 ("good for you")StatusIndustry sanity checkTrend
    Hook: % still watching at 3 s24.6%32.3%leak (below 30%)No organic benchmark. The only reference is ads data, a 2-second hook rate of about 30.7% [Data, Billo via Sepia Lab 2025, 11 ad accounts: tiny sample]— (first review)
    Hold: watch ratio (avg watch time ÷ length)22.3%26.7%own baseline onlyNo public benchmark: your own baseline—
    Completion: watched in full3.5%5.1%own baseline onlyNo public benchmark: your own baseline—
    Spread: share rate0.09%0.15%own baseline only (a quarter of posts earn 0)No public rate benchmark: your own baseline—
    Intent: save rate0.43%0.70%own baseline onlyNo public benchmark: your own baseline—
    Growth: follows per 1,000 views1.12.4own baseline onlyNo public benchmark: your own baseline—
    Discovery: share of views from Search1.2%2.1%own baseline onlyNo public benchmark by account size: your own baseline—
    ER by views ((likes+comments+shares+saves) ÷ views)7.8%11.6%strengthAbove the 1–5K-follower average of 4.40% [Data, Socialinsider 2026 TikTok Benchmarks, 2025 data] and 5.25% [Data, Socialinsider, H1 2026]—
    Views per post (context)7031079strengthAbove the 1–5K average of 350 [Data, Socialinsider, 2025 data] and 317 [Data, Socialinsider, H1 2026]—

    Hook leak, for this account: your engagement and views beat your tier, so people who stay do respond. The loss happens in the opening seconds, and it is concentrated in the template posts (23.0% vs 35.0% for real footage).

    3. What works

    Real footage (n=8) vs template (n=15): 3-s hold 35.0% vs 23.0%, watch ratio 31.9% vs 20.7%, save rate 0.97% vs 0.29%, follows/1K 3.1 vs 0.94, ER 12.7% vs 6.6%. This is a

    finding. Overlap check: 7 of 8 real posts also have no audience named, but that is the account's usual state (16 of 23). Among no-audience-named posts alone, real (n=7) vs template (n=9) still shows 3-s hold 33.5% vs 23.8% and save rate 1.01% vs 0.29%. The gap also holds at every length you use, and real posts are spread evenly across the window (4 early, 4 late) [Own data; medians computed from the intake CSV].

    Search comes from tutorials. Tutorial (n=5) search share is 7.3%, against 0.85% for all other posts (n=18, computed). This stays a

    lead: all 5 tutorials are 22 s long with no CTA, so the data cannot tell the format apart from the length.

    Top 3 posts: #4 (23 Aug, behind the scenes, real, 34 s): highest save rate (1.7%) and follows/1K 3.8 · #12 (5 Sep, teaser, real, 15 s): highest 3-s hold (42.1%) and follows/1K (5.0) · #13 (7 Sep, tutorial, real, 22 s): highest watch ratio (39.5%) and search share (11.8%). Common factor: all three are real footage with no CTA. All three bottom posts (#7, #15, #1) are template.

    Leads (too few posts to count): showcase (n=4) has the highest ER, 10.1%, and the 7 s length bucket (n=4) is the same set of posts. Behind the scenes (n=4) has the highest share rate, 0.17%. The 2 carousels average 755 views vs 703 for videos but get fewer follows/1K (0.47 vs 1.2).

    4. What doesn't

    Template mockup scenes (n=15): they hold the bottom 3 posts and post the lowest median on 3-s hold, watch ratio, save rate and follows.

    Teasers (n=5) earn no shares and no follows: median share rate 0.0% and follows/1K 0.0. All 5 also have no CTA. Among no-CTA posts only, teaser (n=5) vs the rest (n=9) gives follows/1K 0.0 vs 1.1 and share 0.0% vs 0.105% (computed). This is a finding, but a small one.

    Repost (you confirmed #3 and #8 are the same post): #3 (21 Aug, template, 259 views) vs #8 (29 Aug, real, 1281 views): 3-s hold 20.3% vs 41.3%, save rate 0.39% vs 0.86%, follows/1K 0.0 vs 3.1. The second posting changed footage and did far better. It is one pair, so it is a

    lead.

    Leads: posts that name the audience (n=7) do worse (3-s hold 22.6% vs 25.2%), but 6 of the 7 are template, so footage explains it at least as well. Hard CTA (n=9) vs none (n=14): save rate 0.29% vs 0.46%, but all 9 hard-CTA posts are launches or showcases.

    Cadence: a steady 3–4 posts a week. The busiest weeks show no clear drop (correlation −0.40 over 6 weeks, correlation only).

    5. Fix plan: 3 tests, ranked

    Dates assume 4 posts a week starting 4 Oct (computed). Test 1 runs first. Tests 2 and 3 alternate posts afterwards, because they change different things on different posts.

    #Problem KPIChangeN postsPassKillReview date
    1Hook (3-s hold 24.6%)Real filmed product vs template scene, paired by product line and length, with nothing else changed12 (6 + 6), posts 1–12, 4–23 OctReal wins on 3-s hold and save rate in ≥4 of 6 pairsReal wins ≤2 of 6 [Inference: section gives no kill; mirrors its first test]30 Oct 2026
    2Discovery (search share 1.2%)Keyword-first: product + who it is for in the first caption line, on screen at 0:00, and said in the voiceover6, odd posts 13–23, 25 Oct–12 Nov≥10% of views from Search at day 14 in ≥3 of 6None above 3%26 Nov 2026 (day 14)
    3Spread (share rate 0.09%)One line built to be sent ("send this to the friend who…")6, even posts 14–24, 27 Oct–13 NovMedian share rate ≥0.18% (2× your median) and ≥0.3%No post above your p75 (0.15%)20 Nov 2026

    Tests 2 and 3: footage barely moves share rate (real 0.08% vs template 0.09%) or search share (1.6% vs 0.80%), so using test 1's winning footage should not blur the result [Inference].
    After these: test the cold first frame (open on the product in frame 1) and a person speaking to camera in the first second. Both are ad-data ideas [Data, VidMob 2024, ads] and cannot be checked here because neither was tagged. Also test hard CTA vs no CTA on the same category.

    6. Evidence and caveats

    • 23 baseline posts. Three provisional posts (26–29 Sep) are excluded. Hook, hold and completion use the 21 videos; the 2 carousels count only for share, save, follows and ER, and their composite rank uses fewer KPIs.
    • Skipped comparisons (blank): person on screen, direct address, audio, captions, first frame, series. Disclosure is on for every post, so it cannot be compared. Voiceover is used on most videos but was not tagged per video.
    • Tags were applied as you gave them. #3 and #8 say "just landed… link in bio" but are tagged tutorial with no CTA, and that choice affects the tutorial and CTA groups.
    • Labels: [Own data] is this account · [Data] is a dated study · [Inference] is the reviewer's deduction. Nothing here is a cause; every difference is a likely cause to test.
    • Worth checking: any video without voiceover should use Commercial Sounds, because other music is not licensed for product promotion [Official, TikTok Music Terms]. If mockup scenes are AI-generated and look realistic, they need TikTok's AI label [Official, Community Guidelines Sep 2025].
    • Benchmarks fetched 26 Sep 2026. 2026 studies show views per post falling (−59% for 1–5K accounts in 2025 [Data, Socialinsider]), so your own baseline outranks them.

    Next review: 26 Nov 2026. The plan needs that date; read test 1 on 30 Oct. Bring the intake CSV with every post from 3 Oct onwards, tagged the same way, plus audio and first_frame per video and this review's CSV.

    Appendix: per-post table

    Values as script-formatted. Colour = rank against your own quartiles. Columns: 3-s hold · watch ratio · completion · share · save · follows/1K · search · ER.

    #DateCategoryKey attributesViews3-s holdWatch ratioCompletionShareSaveFollows/1KSearchERTest
    232026-09-24tutorialtemplate, cta none, aud. named no, video, 22s66824.6% amber24.5% amber1.8% red0.15% green0.30% amber1.5 amber7.3% green4.6% redT1 template arm · T2 (search lead)
    222026-09-22teaserreal, cta none, aud. named no, video, 15s68925.2% amber36.1% green6.7% green0.0% red1.3% green0.0 red1.1% amber8.0% amberT1 real arm
    212026-09-21launchtemplate, cta hard, aud. named yes, video, 15s70324.1% amber21.3% amber3.6% amber0.57% green0.43% amber1.4 amber0.30% red8.4% amberT1 template arm
    202026-09-19showcasereal, cta hard, aud. named no, video, 7s114336.4% green32.1% green6.5% green0.09% amber0.44% amber2.6 green1.3% amber12.4% greenT1 real arm
    192026-09-17behind_the_scenestemplate, cta none, aud. named no, video, 34s69224.8% amber20.7% amber1.4% red0.14% amber0.29% amber0.0 red0.20% red4.2% redT1 template arm
    182026-09-15tutorialtemplate, cta none, aud. named yes, video, 22s94922.7% red17.1% red3.2% amber0.11% amber0.21% red1.1 amber9.0% green6.2% amberT1 template arm · T2 (search lead)
    172026-09-14teasertemplate, cta none, aud. named no, carousel441———0.0% red0.45% amber0.0 red2.1% green4.3% red—
    162026-09-12launchreal, cta hard, aud. named no, video, 15s148841.5% green26.7% green5.5% green0.0% red1.0% green4.0 green1.8% amber13.7% greenT1 real arm
    152026-09-10showcasetemplate, cta hard, aud. named yes, video, 7s32616.5% red22.1% amber5.1% green0.0% red0.0% red0.0 red0.70% amber5.2% redT1 template arm
    142026-09-09behind_the_scenestemplate, cta none, aud. named no, video, 34s89525.3% amber24.7% amber2.1% amber0.0% red0.22% red1.1 amber0.40% red7.7% amberT1 template arm
    132026-09-07tutorialreal, cta none, aud. named no, video, 22s32233.5% green39.5% green0.80% red0.0% red0.93% green3.1 green11.8% green12.7% greenT1 real arm · T2 (search lead)
    122026-09-05teaserreal, cta none, aud. named yes, video, 15s119742.1% green32.5% green4.4% amber0.33% green0.50% amber5.0 green1.2% amber12.6% greenT1 real arm
    112026-09-04launchtemplate, cta hard, aud. named no, video, 15s103623.0% amber19.9% red3.0% amber0.10% amber0.29% amber0.97 amber0.90% amber5.7% redT1 template arm
    102026-09-02showcasetemplate, cta hard, aud. named no, video, 7s81823.0% amber22.3% amber5.2% green0.12% amber0.24% red1.2 amber1.3% amber7.8% amberT1 template arm
    92026-08-31behind_the_scenestemplate, cta none, aud. named yes, carousel1069———0.19% green0.28% amber0.94 amber0.70% amber7.7% amber—
    82026-08-29tutorialreal, cta none, aud. named no, video, 22s, repost pair 3/8128141.3% green25.4% amber6.9% green0.08% amber0.86% green3.1 green6.4% green10.7% amberT1 real arm · T2 (search lead)
    72026-08-28teasertemplate, cta none, aud. named no, video, 15s64218.9% red15.1% red2.9% amber0.0% red0.47% amber0.0 red0.10% red7.8% amberT1 template arm
    62026-08-26launchtemplate, cta hard, aud. named yes, video, 15s45822.6% red23.9% amber2.0% red0.0% red0.22% red2.2 amber0.80% amber6.6% amberT1 template arm
    52026-08-24showcasereal, cta hard, aud. named no, video, 7s127032.3% green22.0% amber4.1% amber1.3% green1.4% green1.6 amber0.0% red15.6% greenT1 real arm
    42026-08-23behind_the_scenesreal, cta none, aud. named no, video, 34s52632.2% amber31.6% green1.0% red0.76% green1.7% green3.8 green2.1% green12.7% greenT1 real arm
    32026-08-21tutorialtemplate, cta none, aud. named yes, video, 22s, repost pair 3/825920.3% red20.5% red3.5% amber0.0% red0.39% amber0.0 red4.3% green6.9% amberT1 template arm · T2 (search lead)
    22026-08-19teasertemplate, cta none, aud. named no, video, 15s37620.4% red16.5% red1.7% red0.0% red0.53% amber0.0 red1.2% amber8.2% amberT1 template arm
    12026-08-18launchtemplate, cta hard, aud. named no, video, 15s108824.6% amber19.4% red4.4% amber0.09% amber0.09% red0.92 amber0.20% red4.2% redT1 template arm
    242026-09-26behind_the_scenesreal, cta none, aud. named yes, 34s909provisional——0.33%0.55%3.3———
    252026-09-27showcasetemplate, cta hard, aud. named no, 7s807provisional——0.0%0.37%0.0———
    262026-09-29launchtemplate, cta hard, aud. named no, 15s643provisional——0.0%0.0%0.0———

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Most TikTok advice compares you with someone else's account. This skill compares you with you.

    Paste your TikTok Studio analytics (one copy per video), share screenshots, or fill a CSV template. Bundled Python scripts compute every rate, median and quartile, so no number comes from the model's head. You get a 1–2 page review:

    • A KPI scorecard against your own median and p75: hook (still watching at 3 s), watch ratio, completion, shares, saves, follows per 1,000 views, search share and engagement by views, with dated industry figures only as a sanity check.
    • What works and what doesn't: content types, real footage vs templates, lengths and reposts, always with the number of posts behind each pattern. Under 5 posts it is called a lead, not a finding.
    • A fix plan of 3–5 small tests: the problem KPI, the change, how many posts, pass and kill thresholds, and review dates fitted to how often you post.

    Built to be honest. It never promises growth, writes "likely cause" instead of "cause", runs a confound check before calling a difference real, and labels every claim with its evidence.

    Your data stays yours. The skill only uses what you paste or upload. It never logs in to TikTok, never reads other accounts and never scrapes.

    For creators and small brands with 20+ videos who want to know what to change next, not another generic checklist.

    How to use

    Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill passes our trust checks below.

    Security scanned

    Passed our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scanned

    Verified clean 7 days ago

    • Passed all security checks, Safe to install

    Requires

    Python

    Listed7 days ago

    What's inside

    Frequently Asked Questions