CaseScreen — intake screen for dental labs

1. One-liner

Inspects every dental case at the door — flags missing shade, bad photos, void margins before a technician touches it.

2. Trend signal — why now?

Dental labs lose real money on remakes, and the data says most of those remakes are avoidable garbage-in problems that show up at intake — not craftsmanship failures.

• 42% of early remakes correlate with missing or unclear Rx data; 28% with uncalibrated shade photos taken under non-dental lighting. “Missing shade photos hide in WhatsApp until the remake” — labs discover the missing detail after the technician already invested time. (evismart.com, 2026)
• The remake economics are brutal. National remake rate runs 4–7%; nearly a fifth of dentists report >4%. Per Dental Lab Network owners, a lab has to fabricate 5–7 units at full fee to recover the loss on one no-charge remake. Spear Education puts the compounding annual cost “in the tens of thousands.” (Spear, dentallabnetwork.com, 2026)
• AI vision crossed the price/quality line in the last 12 months. Off-the-shelf vision models now grade clinical dental photographs and flag scan issues — incomplete coverage, undercuts, margin voids — reliably enough to ship. (Frontiers in Dental Medicine, 2026; Yucera AI integration review, 2026)

Provenance:

• Signal 1 (demand): 42% of remakes tied to missing/unclear Rx, 28% to bad shade photos hidden in WhatsApp — evismart.com — 2026
• Signal 2 (economic): US dental lab market $7.6B across 4,375 fragmented labs (no player >5% share); LMS incumbents charge $79–250/mo but none do intake image QC — IBISWorld / dentallabnetwork.com — 2026
• Signal 3 (feasibility): AI vision now reliably grades dental photos and flags scan coverage/margin issues at off-the-shelf API cost — Frontiers in Dental Medicine / Yucera — 2026 Category: Tech-unlock

3. The opportunity

Every restoration a lab makes starts with three inputs from the dentist: a scan (STL) or impression, shade photos, and a prescription (Rx). When any of those is wrong or missing, the lab has two bad options: stop and chase the dentist (delays the case, annoys the customer) or guess and proceed (risks a remake the lab eats). In practice small labs guess, because chasing is friction and the case is already in the queue.

The incumbents — LabStar, Evident, Magic Touch, EasyRx ($79–250/mo) — solve case tracking, billing, and dentist portals. Their portals enforce text fields (did you type a shade? did you pick a material?) but nobody looks at the actual image. A dentist can type “A2,” attach a yellow-tinted bathroom-light photo of the wrong tooth, and the portal waves it through. The remake is born at that moment and discovered three days later.

CaseScreen is the missing layer: an AI inspection gate that reads the actual files — scan completeness and margin integrity, shade-photo lighting/angle/in-frame-tooth, and Rx field completeness — at the second the case arrives, across whatever channel it arrived through (portal, email, WhatsApp forward). It returns a pass/flag in seconds and auto-drafts the one “we need X before we start” message back to the dentist. The lab fixes garbage before a tech touches it.

4. Target market

• Primary customer: Owner/operator of an independent dental lab in the US, 2–25 technicians, $300K–$3M annual revenue. The owner is usually a master technician who also runs the floor and personally eats remake losses.
• Why they buy: In their words — “Sometimes feel as if we just bend over and take 1 for the team!” Labs “frequently receive cases without specifying details like material or shade.” Every avoided remake is 5–7 units of recovered margin and a saved customer relationship. They feel this weekly.
• Rough TAM reasoning: ~4,375 US dental labs (IBISWorld). Conservatively 2,000 are the right size — too big to ignore remakes, too small to build custom QC. At $249/mo that’s a ~$6M ARR ceiling on US independents alone; DSO-owned lab groups and the UK/AU/India lab markets extend it well past $5M.
• Why now for them: Remake rates haven’t improved and the labor to chase cases got more expensive; meanwhile their dentists send more digital (scanner-native) cases than ever, so the inputs are now machine-readable files instead of physical impressions — which is exactly what makes automated inspection possible.

5. Product sketch (MVP)

• Drop-zone + channel capture: lab forwards/uploads a case (STL + photos + Rx PDF) or connects an email/WhatsApp inbox; CaseScreen ingests the bundle.
• Scan check: flags incomplete arch coverage, obvious margin voids, and missing antagonist/bite scan.
• Shade-photo check: flags photos that are too dark/yellow (non-dental lighting), out of focus, no shade tab in frame, or wrong/ambiguous tooth.
• Rx completeness check: flags missing material, missing shade, missing tooth number, contradictory instructions.
• Traffic-light verdict per case: green (start it), yellow (start with note), red (do not start — info missing).
• Auto-drafted dentist message: one-tap “Hi Dr. , before we start case # we need: a shade photo with the tab in frame under daylight, and the antagonist scan.” Sent by email/SMS.
• Remake-cause log: every flagged-vs-shipped case feeds a simple monthly report: “you avoided 11 likely remakes; your top dentist for bad shade photos is Dr. ___.“

6. AI angle — what’s load-bearing

The product is only a product because a vision model can now look at a compressed WhatsApp shade photo and an STL and say “this will cause a remake.” Remove the AI and you’re left with the existing LMS portals — text-field checklists that already exist and don’t solve the problem. The load-bearing work is multimodal: grading image quality and clinical adequacy of photos, assessing scan coverage/margin from 3D files, and parsing free-text/handwritten Rx into structured completeness checks. That judgment is exactly what the lab owner does by eye today and can’t scale to every incoming case.

7. Localization angle (if any)

N/A — US-first global play. The wedge is the English-speaking, scanner-heavy US independent lab market. The same product extends cleanly to UK/AU/Canada with no localization, and India’s large outsourced-lab sector is a later expansion — but there’s no payment-rail or language quirk to exploit at launch. Geography is not the wedge here; the AI inspection is.

8. Business model — path to $1M–$5M ARR

• Pricing: $249/mo flat per lab for up to ~400 cases/mo; $399/mo for high-volume labs; usage overage above cap. Anchored just above LabStar/Evident base ($79–250) because this prevents losses rather than just tracking cases — ROI is one avoided remake/month.
• ACV: ~$3,500/lab/year blended.
• Rough math to $1M ARR: ~285 labs at $3,500 ACV. That’s ~6.5% of the US independent-lab universe.
• Rough math to $5M ARR: ~1,400 labs (US independents + DSO lab groups + UK/AU), plus an upsell tier that auto-routes verified-green cases straight into the CAD queue.
• Expansion path: seat/volume growth as labs grow; “verified intake” badge labs advertise to dentists; a dentist-side companion app (shoot-the-shade-photo-right guidance) sold back through the lab; remake-analytics tier for DSO lab groups managing many sites.

9. Go-to-market wedge — first 100 customers

• Dental Lab Network + LMT Magazine community: the forum threads quoted above are where lab owners already vent about remakes. Post a teardown (“we analyzed 200 remakes — here’s what the dentist got wrong”), DM the ~40 owners in active remake threads, offer a free “remake autopsy” on their last 20 cases. Target 5–10 design partners.
• Cold-email the lab directory: NADL (National Association of Dental Laboratories) member list + Google Maps scrape of ~2,000 independent labs. Personalized: “labs your size eat ~$X/yr on avoidable remakes — send us your last 10 cases, we’ll show you which would’ve flagged.” Expect 3–5% to a free audit, 25–30% of audits to paid.
• Scanner/LMS adjacency: intraoral scanner reps and LMS resellers already sell into these labs and don’t compete with intake QC — co-sell or referral fee. One good reseller relationship = a steady trickle of warm intros.
• “Remake autopsy” as the hook: the free analysis of a lab’s recent remakes is the entire wedge — it’s concrete, it’s their own money, and it demonstrates value before they pay a dollar.

10. Build complexity — justification

Medium. The plumbing (file ingest, email/WhatsApp capture, dashboard, dentist messaging) is standard web stack. The hard parts are (a) reliable vision grading of shade photos and (b) usable scan/margin assessment from STL files — both buildable on off-the-shelf vision APIs plus a thin domain-tuned layer, but they need a domain-expert technician in the loop to define “good enough” thresholds and to label a starter set of good/bad cases. A technical founder with a master-technician advisor ships a credible v1 in ~3–4 months.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	QC tooling for labs; not a medical device making clinical claims, no patient PHI required beyond images the lab already holds.
Ethical — no harm / dark patterns	✅	Reduces remakes; advisory flags, lab stays in control.
Market exists (evidence above)	✅	$7.6B, 4,375 labs, documented remake pain and spend.
1–5 person team can build this	✅	Tech founder + technician advisor.
Launchable with <$50K / ₹40L	✅	Off-the-shelf APIs; main cost is design-partner time.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	16/20	Real, recurring, costs measurable money weekly; labs already absorb the loss silently. Just short of hair-on-fire because many cope by guessing.
Demand evidence	15	12/15	Hard numbers on remake causes + verbatim owner complaints + paid incumbents in the category. Missing: nobody yet pays specifically for intake image QC.
Build feasibility	15	11/15	Vision grading is the gnarly part; needs labeled data + thresholds, but no research breakthrough. ~3–4 months.
Distribution clarity	15	12/15	Named communities, scrapeable directory, “remake autopsy” hook with concrete conversion math.
Revenue mechanics	15	11/15	Pricing benchmarked to LMS + clear ROI per avoided remake. Conversion from audit→paid is the open assumption.
Time to first revenue	10	8/10	Design partners can pay within 6–8 weeks; the audit hook shortens the cycle.
Defensibility	10	5/10	Moat is accumulated remake/flag data + workflow lock-in over 12 months; an LMS incumbent could bolt this on, so speed matters.
Total	100	75/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required

Key assumptions to validate (3–5)

1. Assumption: Off-the-shelf vision models flag bad shade photos and incomplete scans accurately enough that labs trust the verdict (low false-positive rate). How to test: Run 200 historical cases (100 that became remakes, 100 that didn’t) through a prototype; measure precision/recall on catching the remake-causers.
2. Assumption: Lab owners will pay ~$249/mo for prevention, not just tracking. How to test: Offer 20 labs a free remake autopsy; count how many ask “can I keep using this?” and accept a paid pilot.
3. Assumption: Labs can route a meaningful share of incoming cases through CaseScreen despite the WhatsApp/email mess. How to test: With 3 design-partner labs, instrument one week of intake; measure what % of cases can actually be captured automatically vs. need manual upload.
4. Assumption: The auto-drafted dentist message gets used, not ignored (the value depends on the loop closing). How to test: Track send-rate and dentist-response-rate across design partners.

Risk flags

1. Platform/incumbent risk: EasyRx/LabStar already own the intake portal and could add image QC as a feature. Mitigation: move fast, win the data and the “verified intake” brand, stay scanner/LMS-agnostic so you’re the neutral layer.
2. Accuracy/trust risk: A false “red” that delays a perfectly good case erodes trust faster than a missed flag. Mitigation: bias toward yellow/advisory, keep the human in control, tune thresholds per lab.
3. Capture risk: If most cases arrive as compressed WhatsApp images forwarded by hand, ingestion is messy and adoption stalls at the upload step. Mitigation: make manual drag-drop frictionless first; automate channels second.

14. Structured verdict

Score:                  75/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder + master-technician advisor (domain expertise required)
Time to revenue:        6–8 weeks to first paid pilot
Capital to launch:      $8–15K ($ for vision-API usage + design-partner time)
Top 3 assumptions to validate first:
  1. Vision accuracy on catching remake-causers — backtest 200 historical cases (100 remakes / 100 clean)
  2. Willingness to pay $249/mo for prevention — 20 free remake autopsies, count paid-pilot conversions
  3. Real-world case capture rate — instrument one week of intake at 3 design-partner labs
Kill criteria:
  - Abandon if prototype precision/recall on the 200-case backtest can't beat a lab tech's eyeball QC
  - Abandon if <20% of free-audit labs convert to a paid pilot
  - Abandon if a major LMS ships equivalent image-QC before your v1 and bundles it free

15. Next step — 1-week validation sprint

• Day 1–2: Recruit 3 friendly lab owners from Dental Lab Network. Get 200 anonymized historical cases — 100 that became remakes, 100 that shipped clean — with their files (scan, shade photo, Rx).
• Day 3–4: Wire up off-the-shelf vision models into a throwaway script that scores each case green/yellow/red on shade-photo quality, scan completeness, and Rx completeness. Run all 200.
• Day 5: Measure: of the 100 known remakes, how many did the script flag red/yellow? Of the 100 clean cases, how many did it wrongly flag red? Go if it catches ≥70% of remake-causers with <15% false-red on clean cases. No-go if it can’t beat eyeballing.

The result is falsifiable: a confusion matrix on real historical cases, not a vibe.