In partnership with
Invest in Popsmith, the Future of Popcorn
Popsmith is reimagining popcorn for a new generation.
With more than 100,000 customers, $20M+ in lifetime revenue, and retail partnerships with Williams Sonoma, Costco, and Sur La Table, Popsmith is building a premium brand in one of America’s most iconic snack categories.
Revenue is projected to reach $9M in 2026, with a nationwide Williams Sonoma rollout coming this holiday season.
Now, for a limited time, you can own a piece of the company and participate in what comes next.
See the Popsmith investment opportunity.
The reader you pay for.
Two issues ago: the payer’s reader, trained on the claims history. Last issue: the acquirer’s, pricing the perio your archive shows and your ledger doesn’t. This issue: the reader you pay for, on the five-line scorecard promised in Issue #2.
The Bench Test scorecard was introduced in The chair-hour is the unit of production: one product per issue, five lines, same discipline. Pearl’s Second Opinion is the widest-cleared dental imaging AI on the market, which makes it the right place to begin.
Independent evidence is thin across the category. Where a claim is vendor-only, this scorecard says so.
1) Evidence grade — B
Pearl’s foundational 510(k) K210365 (March 4, 2022) is the paper trail an owner should actually read. The FDA summary lists the standalone sensitivity band the model was cleared on — 76.39%–89.77%, with false positives per image between 0.46 and 4.85 (FDA K210365 summary) — and the MRMC reader study that got it through: 25 of 25 readers improved on periapical radiolucency, 22 of 25 on calculus, 17 of 25 on caries.
Since then Pearl has added at least seven more clearances, including the K242522, K242600, K243234 trio that swapped bounding boxes for polygonal segmentation for caries and periapical radiolucency (Pearl), plus panoramic, pediatric, and CBCT clearances. Third-party count as of early 2026: eight — the most in dental AI (review.dental).
What is missing is an independent effectiveness study. Pearl’s product page cites “94% accuracy, validated by an independent third party” and an operator lift from 82% to 98% in a published study, but names neither the third party nor the paper (hellopearl.com). Every dental-AI head-to-head I could find in 2026 acknowledged the same shortage — “published accuracy ranges for Pearl and Overjet overlap heavily and head-to-head studies under identical conditions are rare” (smbai.guide).
Grade: B. Regulatory footprint is real; independent clinical evidence is still mostly promissory.
2) Workflow fit — Minutes on the day
Pearl runs as an overlay on the imaging monitor the operatory already has, and the analysis returns in about 30 seconds per image (hellopearl.com). The integration list is genuinely broad: Dentrix, Eaglesoft, Open Dental, Curve, Denticon, CareStack, DEXIS, Carestream, Apteryx/XVWeb, tab32 — 20 to 40+ platforms depending on which source you trust (review.dental, Pearl).
Setup is not the 30-minute Monday-morning target the Dental AI Product Guide uses for phone and voice tools. The realistic sequence, per an independent comparison, is:
Weeks 1–2: pilot
Weeks 3–4: full operatory rollout
Through day 90: case-acceptance monitoring
Time added to the clinical day is small once configured — Pearl handles the remote configuration itself (Pearl) — but the workflow shift is in the treatment-coordinator conversation, not the operatory. Practices that leave the overlays running silently without showing patients the images “may see a fraction of the published ROI, sometimes essentially none” (smbai.guide).
3) Unit economics — Cost per chair-hour, and whose production it lifts
Public figures, all third-party because Pearl publishes none:
Line item | Reported figure | Source |
|---|---|---|
Second Opinion, per location | ~$299/mo (quotes range $250–$500) | |
Setup fee | ~$1,500 (includes historical radiograph analysis) | |
Second Opinion via partner channels | ~$349/mo | |
Practice Intelligence add-on | ~$549/mo ($494 prepaid) | |
Typical Pearl bundle | $700–$900/mo per location |
Applied to the chair-hour frame: an office running ~1,600 delivered clinical hours/year carries roughly $2.24 of Pearl subscription per chair-hour for Second Opinion alone, or $6.75 per chair-hour at the full bundle. Against the ~$154 per chair-hour fixed burn calculated in the chair-hour issue, Pearl is a 1.5%–4.4% overhead line — cheap on the P&L, expensive only if it is not being used.
Whose production it lifts is the harder question. The advertised 30% lift in case acceptance is vendor-reported (hellopearl.com). The independent frame says case-acceptance lift from AI overlays clusters at 15–25 percentage points off baseline only when the overlays are shown to the patient during consultation (smbai.guide). If the doctor is the beneficiary of the lift, the ROI accrues to whoever holds the ownership stake. If the associate is doing the consult, the associate’s production line moves — which is a Buyer Mirror problem for a seller-curious owner.
The audit trail issue from Your Radiograph Gets Read a Third Time also matters here. Pearl’s setup fee includes “analysis of your historical radiographs.” That is exactly the retrospective sweep an acquirer would run to price your perio gap. Which brings us to line 4.
4) Data terms — Who keeps the images
The FDA summary confirms Pearl’s architecture: an in-office client passes images to a cloud API where the computer-vision models run and append metadata to each radiograph (FDA K210365). Images leave the building.
Public HIPAA and BAA language is thin. What is verifiable:
Pearl operates under a HIPAA Business Associate Agreement with dental customers (smbai.guide).
The vendor does not publish image ownership terms, retention periods, storage locations, or training-use policy on its product page (hellopearl.com).
Third-party guidance is explicit: get the signed BAA before images leave the network, and get the retention window, storage location, vendor access list, and training-use answer in writing (smbai.guide).
An owner running last issue’s audit should treat the training-use answer as the important one. If a vendor’s BAA allows use of de-identified images for model improvement, that vendor’s retrospective sweep of your archive is a two-sided transaction: they price your perio gap, and they get better at pricing everyone else’s. Ask.
5) Verdict by buyer size
Solo practice. Pearl is a viable second reader on 2D images. At ~$299/mo it is a small line item, and Second Opinion alone is the lighter rollout (smbai.guide). Two conditions: the doctor commits to showing overlays during patient conversations, and the practice signs the BAA with retention and training-use language pinned down first. Capterra reviewers specifically flag over-flagging of interproximal decay and cervical burnout artifacts marked as pathology (review.dental) — the false positives are real, and they will need to be explained to patients in the room.
Group practice (2–5 locations). The economics still work — Pearl remains a low percent of chair-hour burn — but the Buyer Mirror consideration kicks in. Pearl’s historical-radiograph analysis on setup is precisely the archive sweep an acquirer runs during diligence. If a transaction is plausible in the next 36 months, run the retrospective yourself, own the output, and close the perio gap before someone else prices it into your multiple.
DSO. Pearl is one of two credible incumbents; the other is Overjet. The decision belongs on the data-terms line, not the accuracy line. Head-to-head accuracy is a wash in the current published record. Contract terms — image ownership, cross-tenant model training, portability on exit — are where negotiation leverage matters.
The one line
Pearl Second Opinion is the widest FDA-cleared dental radiograph AI on the market, priced as a small line on the chair-hour P&L, and viable for any independent owner willing to sign a BAA with the training-use question answered in writing. The productivity claims are vendor-reported; the workflow value is real when overlays are shown to patients, and largely absent when they are not. The overlooked risk for a seller-curious owner is that the same retrospective the vendor runs at setup is the retrospective an acquirer’s AI runs during diligence.
What Practice Ledger asks next time
Independent, non-vendor sensitivity and specificity from a named study, with dataset and reader panel.
Written training-use, retention, and portability terms in the standard BAA.
Case-acceptance measurement protocol that is auditable, not self-reported.
Chair-hour productivity impact separated from case-acceptance uplift.
Next week: Bench Test #2, Overjet AI Assist. Same five lines, the other incumbent, and the vendor that also sells the reader on the payer’s side of your claim. The Bench Test jumped the queue because the scorecard was ready and the invoices weren’t; the pricing issue and the annual-maximum story are still owed, in that order, and if either slips again I’ll say so here rather than go quiet.
Disclosure: I’m building a small group of practices myself (that’s the Diaries), so I sit on the buy side of this table, and I have a commercial interest in periodontal care. Nothing in this issue names it or is sold by it. The OVN Nexus brief linked below is free clinical education from my research platform; it discloses its own support on the page.
No industry sponsors, no affiliate links. If there is an ad at the top, it is a company that has never seen a bitewing; it buys adjacency, never a verdict. If this one landed, send it to the colleague who is about to sign a Pearl contract.
The clinical side of this desk: OVN Nexus, my research platform, now carries a free two-minute brief for dentists and their teams on what the oral-systemic evidence supports and where it stops, with a one-page discussion guide for the next team meeting. That page points here; this one points there.
— Thad
Sources: FDA 510(k) summary, K210365 (March 2022) · Pearl: Second Opinion product page, segmentation clearances announcement, integration and setup post, privacy policy · review.dental, Pearl review (pricing, integrations, Capterra reviewer complaints) · smbai.guide, Pearl vs Overjet comparison (rollout timeline, case-acceptance frame, BAA guidance) · The Practice Ledger: the chair-hour issue, the product guide, the third reader. Evidence grade for the pricing figures: third-party review sites, no invoices disclosed, and Pearl publishes no rate card. Falsifier: an invoice. Corrections append here, never silently rewrite.
Bench Test scorecard: evidence grade, workflow fit, unit economics, data terms, verdict by buyer size. One product per issue. Neutral data publication.