The standard
Published in full, including the parts that make us look slow.
Toothachey is new and has no customer logos to show you. What it has instead is a method you can read before you pay, and hold us to afterwards. No competitor in this category publishes theirs.
Step one
Harvest, from the primary record wherever one exists.
Each event type has a public record behind it. We go to that record rather than to an aggregator that copied it, because a copy is where the errors live.
| Event | What we look for | Where it comes from | Base score |
|---|---|---|---|
| For sale | A listing appears on a broker or business-for-sale site. | Broker listings, business-for-sale marketplaces | 85 |
| Ownership change | New owner, DSO acquisition, or a name change on the practice record. | State filings, dental board records, the practice's own site | 80 |
| New opening | A location with no prior record appears, cross-checked against license issuance. | Dental board license issuance, new business filings, map listings | 75 |
| Associate hire | A new name on the team page, or an associate role posted. | Practice team pages, job postings | 60 |
| Hygienist post | An open hygienist role posted in the last seven days. | Job boards, practice career pages | 55 |
| Tech change | A change in the practice site's own markup: new booking widget, new PMS, full rebuild. | The practice website's public markup | 50 |
| Review collapse | Rating or review velocity drops sharply against the previous observation. | Public review counts and ratings, tracked over time | 45 |
Step two
Compare against last week, because a first sighting is not news.
Two of the seven types only exist as a difference.
We hold a roster per state
Last seen rating, review count, and a fingerprint of the practice website, per practice. Every run compares against it, emits a signal only on a material move, then writes the new snapshot.
Noticing you is not an event
If a practice enters our roster this week with a 3.9 rating, that is not a review collapse. It is a starting point. The earliest it can produce a signal is the following run.
Step three
Score it, using a model we will show you.
The score exists to order your Monday, not to impress you. It is arithmetic, not a model you have to trust.
Base, by event type
- For sale85
- Ownership change80
- New opening75
- Associate hire60
- Hygienist post55
- Tech change50
- Review collapse45
Then adjusted
- Inside the last seven days+ up to 15
- Practice larger than average by review count+ up to 10
- Source is a secondary aggregator, not the primary record- 20
Clamped to 0 through 100. Anything under 30 is noise and never reaches a file. We would rather send you eleven rows than thirty.
Step four
Verify, before anything ships.
This is the slow part and it is the reason the product is worth buying.
- 01
Weeks one through four of any territory: everything is checked
A person opens the source URL, confirms the fact is actually stated there, and confirms the practice identity matches. Nothing ships unchecked while a territory is new.
- 02
After that: full checks on the expensive ones, sampling on the rest
Every for-sale and every ownership change stays fully verified, because those are the highest priced and the most embarrassing to get wrong. Plus a random 20% of everything else.
- 03
A failed sample sends the whole batch back
If a sampled row does not hold up, that batch returns to full verification for two weeks. The penalty for sloppiness falls on us, not on your Monday.
- 04
A failed row is never deleted
It stays in our system marked unverified, so the same wrong fact is not harvested again next week and quietly shipped to somebody else. It just never reaches a subscriber.
Step five
Deliver, and admit it when a week is quiet.
Monday morning, your states only, ordered by score.
Boring on purpose
Practice, city, what happened, the date, the source link. It is built to be scanned in four minutes and worked from a phone in a parking lot.
A thin week is reported as thin
If your territory produced three rows, you get three rows and a sentence saying it was quiet. The first instinct of a rep who receives filler is to cancel, and they are right.
If we get one wrong
Tell us the row. If the source does not support it, we correct it in the next file, mark the original unverified, and credit the month if it was a for-sale or an ownership change. Those two carry the price and they should carry the consequence.