Emerging
40A pattern that has cleared the publication floor and is worth telling you about, labelled as early rather than dressed up as settled.
// The part nobody else has
Every practice on Ascend makes every other practice better. None of them ever sees another's numbers.
This is the whole argument. Everything else, a determined competitor could copy in a year. This they cannot, because it is not a feature. It is what happens when a network of practices contributes evidence to a system built from the first day to keep every practice's numbers inside its own walls.
// How a pattern travels between practices
The network supplies the principle. Your practice supplies every number.
1
Evidence
Ascend observes what actually happened across the network. Not opinions, not a survey, not an industry benchmark report. Outcomes, recorded as they occurred.
2
Transfer
When a pattern holds, a separate question is asked: can the mechanism behind it apply elsewhere? That judgement is written in code and reviewed by a human. It is never guessed at run time, and anything unrecognised fails closed.
3
Adaptation
The principle arrives. Every number attached to it is then computed from your rows: your patient list, your measured rates, your history, your currency.
4
Validation
Ascend states in advance what it expects to move and over what window, then measures whether it did. A recommendation that cannot be checked afterwards does not ship as one.
// Confidence is a number, not an adjective
Three tiers, and Ascend tells you which one it is on.
Established
60Enough independent evidence that Ascend will act on it in a recommendation, with the reasoning attached.
Proven
90Held across the network over time and validated against what it predicted would move.
// The privacy contract
Structural, not promised.
Guaranteed: The network carries principles. It is incapable of carrying magnitudes: not revenue, not pricing, not patient counts, not conversion rates, not budgets.
Guaranteed: Patterns are assembled from a fixed vocabulary of tokens, so a number cannot physically appear inside one.
Guaranteed: Nothing is published until at least three independent practices and thirty separate observations support it.
Guaranteed: Nothing identifying your practice ever enters the pool, so there is nothing there to reclaim: you exist in it only as a one way hash, and only patterns that generalise are emitted at all. Ask us to stop contributing and we stop, and what you already contributed stops counting as current evidence.
Never: No "practices like yours". Ascend never compares you to a competitor, never ranks you against a peer group, and never builds a similarity profile of your business.
// Why this cannot simply be copied
Six decisions taken before the first practice joined. None of them can be added afterwards.
Every one of these had to be true on day one of the schema. A competitor who ships a shared learning feature next year inherits data that was never captured this way, and there is no migration back.
The same action, described three ways, counts once
Posted a whitening case, uploaded a smile transformation, shared a before and after. Three practices, three phrasings, one thing. Left as text those stay three weak signals forever, no matter how many practices join. Ascend normalises every observation to a fixed vocabulary at the moment it is captured, so the network gets stronger with scale instead of noisier. Anything it does not recognise is held aside rather than forced into the nearest bucket.
Anonymised three separate ways, not promised once
No practice identifier ever touches a contributed row. A contributing practice exists only as a one way hash. And only patterns that generalise are emitted at all, so the specific never leaves in the first place. The tables that hold contributions are not reachable by the application at all, only by the nightly job that writes them.
You are told your number, not the flattering one
A pattern might have eighteen contributing practices worldwide and three in your market. Ascend shows you the three. The larger number is the right one for the network and the wrong one for your practice, and printing it would be the easiest exaggeration in the product.
Knowledge is revised, not quietly forgotten
When two established patterns in the same slice start predicting opposite things, that is not noise to be aged out. The side with real outcome evidence behind it wins, and the other is superseded with a reason attached and kept on the record. Advice that is actively wrong now does more damage than advice that is merely old, so it is handled differently.
It learns which recommendations practices actually act on
Not only which patterns hold, but which kinds of advice get taken up and, as the evidence matures, whether taking them up worked. That runs through the same anonymisation and the same floors as everything else.
The model never goes looking
Retrieval is deterministic. A query hands the model a small, already validated set of patterns, ordered so a weak one can never outrank a strong one. The model writes and explains. It does not decide what counts as evidence, which is the job it would be worst at.
Any of the individual features on this site could be built by a competent team in a year. This could not, because it is not a feature. It is the compound result of every practice on Ascend contributing evidence to a system that was built from the first day to keep each practice's numbers inside its own walls, and none of that is retrofittable onto a product that did not start that way.
// What the network cannot do yet
The honest state of it.
A network needs practices before it has anything to say, and the founding cohort is how that starts. Here is where it genuinely stands.
A pattern needs 3 independent practices and 30 observations before it publishes. Below that it is held, and you are not shown a thin signal wearing confident language.
That is why the network is the third layer and not the first. Ascend ships with a Foundation layer of curated guidance and industry research, so on the morning you sign it already has something useful to say about what tends to work. It simply will not pretend that knowledge is something it measured about you.
Everything Ascend does from your own data works on day one and does not wait for anybody else. Reactivation, slot fill, the receptionist, the queue, the brief: all of it runs on your rows.
The network is the part that compounds. In month one it will have little to add. In month twelve, with a cohort behind it, it is the reason Ascend knows something your last vendor could not.
We would rather tell you that now than have you discover it in week three.
3
Independent practices, minimum
30
Separate observations, minimum
Both are policy constants in the platform rather than editorial judgement, which means nobody can decide to publish something early because it would look good.
// The ask
Send us your last twelve months. We will run your own numbers through the four leaks.
Not the worked example from earlier. Yours. You will see the actual figure for your practice, with every assumption named and arguable.
If the number is not big enough to justify the fee, that is a completely reasonable outcome and we will tell you on the call rather than chase you for a quarter.