Research Audit · Independent Project
Research Audit
Sunnies Face — Lip Dip & Lip Treat Market Research Pack
Methodology: Independent review of source traceability, customer patterns, and research claims. Findings include documented corrections and remaining evidence gaps.
Starting Point
Research Hypotheses
Two hypotheses this audit evaluates against the evidence below — not established customer findings.
Potential purchase barrier
Whether texture and wear time justify the price, particularly in humid conditions.
Potential customer priority
Reliable color, comfortable wear, and easy application at a price that feels worthwhile.
Sources and limitations
This audit draws on public sources — brand sites, review platforms, marketplaces, Reddit, TikTok, and press coverage — together with researcher-supplied screenshots of public listings and threads, transcribed to text rather than provided as original image files. No client or private business data was used. Because the underlying marketplace and forum evidence was supplied as transcribed text rather than original screenshots or direct source links, this audit could not independently confirm that every entry is free of personal identifiers, or re-verify each source against its original page — a limitation of what this audit could check directly, not a completed verification.
Check 1 of 4
Sources
Opened Section 1. Checkmarks were not taken at face value.
Which sources did the pack say it checked but clearly did not, or could not?
YouTube is the one case worth naming directly. Section 1 marks it "Checked: Yes" with "10 relevant video titles located." That's true only in the narrowest sense — titles were located via search; no video was watched, and no comment or description text was ever read (every direct fetch returned HTTP 429). The row's own notes disclose this honestly, but a "Yes" next to "10 titles" can still read, at a skim, as "10 videos were reviewed." I've treated this as a labeling problem worth flagging here rather than a fabrication — the underlying disclosure was already accurate, just easy to misread out of context.
Who is overrepresented in this evidence? Complainers, fans, or one platform?
Complainers, by collection method more than by platform. The marketplace evidence (Shopee/Lazada) was deliberately gathered through star-filtered views (1★–3★ filters, "with media" filters skewed toward documented problems) specifically to surface negative slices — the pack says so itself, but it means the ~25-contributor formula-performance count is inflated relative to the product's own aggregate rating (cited elsewhere in the pack as roughly 94% five-star). Reddit and TikTok add a second, different skew: people motivated enough to post in a comment section or complaint thread aren't a random sample of buyers either way.
One question this evidence CANNOT answer, no matter how good it looks
What proportion of Lip Dip or Lip Treat buyers actually experience the dryness/durability complaint. No amount of additional star-filtered or search-driven contributor counting fixes this — it would take a random or representative sample, which nothing in this pack (or realistically available to an outside researcher) provides.
Verdict: good enough to work with.
Check 2 of 4
Patterns
Opened Section 4. The pack proposes eight groupings; three are audited below in detail.
| Pattern | People | Sites | Your verdict |
|---|
| Formula-performance complaints | 25+ | 7 | AcceptHolds, but the bare word "STRONG" travels badly on its own — it needs the collection-bias caveat attached every time it's quoted, not just once in this table. |
| Reformulation claims | 4 originally cited | 2 | EditTwo of the four cited contributors (P74, P75) have no exact words anywhere in the pack. Corrected — see below. |
| SG-vs-PH price/channel gap | 1 direct account | 2 | Reject (as a pattern)One account isn't a pattern by this pack's own definition. Demoted and moved — see below. |
Note: if the pack had given quote counts instead of people counts, that alone would have been an automatic edit. It didn't — every row already counts distinct contributors, not excerpts. The two problems found here are about traceability and definition, not inflated counting.
One change I made: merged, split, renamed or rejected. Which, and why
Two changes, both applied directly to the pack:
1 — Rejected and demoted a pattern. "SG-vs-PH price/channel gap" was listed in the Patterns table on the strength of one customer account (P2). A pattern, by this pack's own rule, needs convergence across people — one account with market-context evidence stapled onto it doesn't clear that bar, however well-sourced the context is. I removed the row from Section 4 entirely and left the single-account read where it's properly scoped: the pricing table (§2.3) and Segment 2 (§5, "The Import-Price Skeptic"). It no longer sits next to 25+, 6+, and 5+-contributor rows implying comparable strength.
2 — Corrected an untraceable citation. The reformulation-claims row cited P74 and P75 as supporting and disputing evidence. Neither contributor's exact words appear anywhere in Section 3 — they exist only in the researcher's fuller working notes, never reproduced in this pack. I removed P75 from the supporting count and re-labeled P74 as a flagged, not-independently-traceable claim rather than in-pack counter-evidence. The pattern's overall read (MODERATE, contested) didn't change — P112, which is reproduced, still carries the disagreement — but the citation no longer implies something a reader can't actually go check.
One piece of evidence I found that DISAGREES with the main finding
P170 (Reddit, Thread 4) directly disagrees with the smell/mold pattern: after extended, multi-unit use, the commenter reports no mold or rancid smell at all. It's already in the pack (§3.4, §4.1) — not something I had to go find — but it's worth restating here because the pattern's strength rating (MODERATE-to-STRONG) is a repetition count across platforms, not a resolution of this disagreement. P170 stands unaddressed by every commenter asserting the concern as settled fact.
Check 3 of 4 · Source Verification
Can You Trace It?
Three claims picked at random, including one from the executive summary. Exact words traced, not the easy ones.
| The claim | The exact words behind it | Source / URL | Verdict |
|---|
| Exec. summary: an independent gopicky.com account of the product developing a rust-like smell corroborates, without confirming, the Reddit-sourced smell/mold concern | "accumulates a smell akin to rust as time goes" | gopicky.com/product/43470/sunnies-face-lip-treat, reviewer "Deniugc" (P228) | Accept |
| §6 Market Read: PH buyers compare specific shades and formulas to specific named competitors | "hindi ganung katagal yung tint unlike sa Happy Skin" (P24); "Nag iba ung shade nya dati... I compared it sa nabili ko before" (P64) | Shopee PH listing screenshots | Accept |
| §4 Patterns (pre-audit): reformulation claims cite 4 marketplace contributors — P36, P70, P75, P78 | No quote for P75 exists anywhere in Section 3, or anywhere else in the pack | Cited only as "marketplace" — no listing, no reproduced quote | Flagged, then corrected |
How many of the three could NOT be traced?
1 of 3 (P75; its paired counter-citation P74 has the same problem and was caught in the same pass).
What did you do about them?
Corrected it directly in the pack rather than leaving a footnote for later: removed P75 from the reformulation pattern's supporting count in Section 4, and re-labeled P74 as a flagged, working-notes-only claim instead of in-pack counter-evidence. Both corrections are visible in the pack's own text, not just in this audit, so a future reader sees the audit trail without needing this document alongside it.
Remember: untraceable does not mean untrue. It means it cannot be said yet. That's why P74 and P75 were flagged and demoted, not deleted from the researcher's record — they may well be real; they just aren't verifiable from this document alone.
Check 4 of 4
Judgment
Opened Sections 6, 10, and 11. This is where checking stops and deciding starts.
The market
The desire, in your words: what they want to KEEP
The thing that got them buying it in the first place: bold color payoff and a blurred, matte-soft finish in one swipe, at a price that undercuts the imported brands they'd otherwise reach for.
The desire, in your words: what they want to LOSE
The tradeoffs they're now willing to complain about instead of walking away from: lips drying out or clumping over a wear, product that doesn't stretch as far as the price implies (especially bought on sale), and — on the Singapore side specifically — the sense of paying an import tax even with an official local channel now confirmed to exist.
| AI sophistication stage | Do you agree? | Why, or why not |
|---|
| 3–4 (comparison / differentiation stage) | Agree, with a caveat | The PH evidence genuinely earns this: four separate reviewers name Happy Skin, Romand, or Colourette unprompted — real comparison-stage behavior. The Singapore evidence doesn't clear the same bar — it rests on one reviewer benchmarking against Sephora liquid lipstick generically, which reads closer to stage 2 (still evaluating the category, not comparing named regional rivals). I'd keep the PH read as stated and treat the SG stage as unconfirmed rather than folded into one market-wide number that can be misread as applying evenly to both. |
One white space
| The idea | Do customers want it? | Is anyone saying it? | Can the brand back it? | Tier |
|---|
| Publish a clear, side-by-side humidity/wear-test comparison for the Lip Dip/Lip Treat line specifically | Weak-to-moderate — several reviewers already run and report this comparison themselves (P1, P2, P32), a real behavioral signal, but nobody has explicitly asked the brand for a published test | No — none of the four researched competitors publish this either; a category-wide gap, not a brand-specific one | Unclear — no test methodology, lab claim, or wear-time data was found on the brand's own site for the lip line | NOT PROVEN YET |
I agree with the pack's own tier here rather than upgrading it. "Customers are behaving as if they want this" is not the same as "customers have asked for this," and the brand's capacity to deliver a real wear-test claim is a genuine unknown, not just an unstated yes.
Check 4, continued
What You Still Do Not Know
| What we do not know | What to collect next | Where to get it |
|---|
| Whether the rancid-smell/mold concern reflects an actual recurring defect or an unverified, cross-platform-repeated narrative | Batch/manufacture-dated units tested for smell/mold onset over a fixed shelf period | A direct, controlled comparison of old vs. current stock; or a documented brand statement |
| Whether Lip Dip has actually been reformulated | A primary brand statement on formulation history, not a secondhand customer-service paraphrase | Direct written outreach to Sunnies Face customer service, exchange kept for the record |
| Whether the official Lazada Singapore channel is priced competitively enough to end the import-arbitrage behavior the one confirmed SG account described | Current Lazada SG pricing/promotions tracked over time, plus more than one SG-based customer account | Direct Lazada SG listing checks; further Lemon8/social search restricted to Singapore |
Summary
How many things were edited or rejected in total?
3(1) corrected the reformulation-claims citation to remove an untraceable contributor; (2) rejected and demoted the SG-vs-PH row from the Patterns table to a single-account observation; (3) re-flagged the paired P74 counter-citation as not-in-pack.
Looking back at the start of this audit: what was expected, and what actually turned up?
I expected the main objection to be texture and lasting power — that held up, and turned out to be the best-evidenced finding in the whole pack (25+ contributors, 7 source types). What I didn't expect was how thin the Singapore side would stay even after a dedicated pass to firm it up: there's now real market context (a confirmed official Lazada SG channel, SGD pricing), but customer-level Singapore evidence is still exactly one account. That's not a research failure — the pack says so plainly — but it means the PH-vs-SG comparison the brief asked for is really "a well-evidenced Philippines picture next to one Singapore data point," and that should be said plainly rather than let the document's polish imply a more balanced two-market comparison than the evidence supports.
Every claim I kept can be traced to a source I can point to.
I counted people, not quotes.
There is counter-evidence in here.
Every inference is labelled as an inference, not stated as fact.
I did not call any ad winner.
Independent audit. Completed 11 September 2026.