Strategy
Excellence is not a claim a platform makes. It is a system it runs.
The question behind this page is not whether Specify has good ideas. It is whether Specify has a method for finding better ones after the people who founded it stop having them.
It is a strategy, not a record of results.
Practice that compounds
That method is currently run by people, by hand, on a small number of examples. There is no dataset large enough to measure anything, no ecosystem to compare a merchant against, and no engine deciding autonomously what to change. What exists is the sequence itself, and the discipline of writing down what it produces.
This page describes that sequence: how a standard gets written, where the gap between the standard and the current experience is found, and what has to be true before a change is allowed to be called an improvement.
Observation
Platforms rarely fall behind because they stopped shipping. They fall behind because nobody kept asking what good was supposed to mean, so the answer set at launch quietly became permanent.
The decay is usually structural rather than negligent:
None of this looks like failure while it is happening. Releases still ship. The product still gains features. What is missing is any mechanism that could tell the team that the definition of good has moved and theirs has not.
A product that never revises its definition of good has already chosen one.
Strategy
A gap cannot be measured against an instinct. Before Specify can say that a commercial experience falls short, it has to be able to say what it falls short of, in a form somebody else could disagree with.
A standard in this sense is not a style guide and not a checklist of features. It is a written position about what a commercial experience should do for a customer at a particular moment, held with enough precision that a real storefront can be inspected against it and found wanting.
Four things separate a written position from an opinion held confidently:
| Where it came from | The research, the observed behaviour or the recorded enquiry that produced it. A standard with no source is a preference wearing better clothes. |
|---|---|
| Where it applies | The kind of purchase, the moment in the journey, and the conditions under which it holds. A standard that applies everywhere usually says nothing. |
| What it looks like when it is met | Something observable in the experience itself rather than a feeling about it. Somebody other than the author has to be able to check. |
| What would make it wrong | The condition under which the standard gets rewritten. A position that cannot be overturned is a belief, and beliefs are what make platforms stale. |
Staying current does not mean following every trend. It means being able to tell, with evidence rather than instinct, which changes in practice are durable and which are fashion, and only then rewriting the standard.
You cannot close a gap you have not defined the far side of.
Strategy
Excellence in commerce is not one property. It is a set of separate questions, each with its own answer, and most of them are currently answered by accident.
Candidate areas, not a published library. Naming the areas is the first step; writing the standards is the work.
The written output
Standards are internal. The Idea Catalogue is what gets published from that work: a selection of practical ideas for specification journeys, product information, capability and constraints, each carrying an explicit label for how well supported it actually is.
None currently carry a strong evidence label
Strategy
The strategy is not a set of beliefs about commerce. It is a sequence that can be run again, by different people, on a different area, and produce a defensible change at the end of it.
Read the behaviour, not the trend commentary. What did the customer actually try to do, where did they stop, and what did the businesses that handled it well do differently.
Turn the finding into a position with a source, a scope, an observable signal and a condition that would make it wrong. Until this exists there is nothing to measure against.
Take a real journey and check it against the standard rather than against taste. The inspection has to be able to fail, otherwise it is a demonstration.
Run by people today
State precisely where the experience falls short and why that matters to the customer standing in front of it. A gap nobody can articulate is a gap nobody will close.
Design the change, with the reasoning attached to it. The reasoning is what makes the change reviewable later by someone who was not in the room.
Where a test is possible, run the improvement against the current experience and leave it alone long enough to say something. Where a test is not possible, and at Specify's volume it usually is not, the change is made on the strength of the standard and recorded as a judgement rather than a finding.
Next hypothesis
Compare the result against what the proposal predicted, including when the answer is that nothing detectable changed. A measurement that can only confirm is not a measurement.
Next hypothesis
Fold the validated change into the platform so that every merchant inherits it, and record what was tried and abandoned so the next pass does not repeat it.
Back to 02. Write the standard down
What survives measurement rewrites the standard, not only the product. This is the step that makes the next pass through the loop different from the last one.
None of this is fast. A single honest pass takes longer than shipping the same change on instinct would have. The argument for it is not speed. It is that the tenth pass is better informed than the first, and instinct does not have that property.
The loop is not how the product improves. It is how the definition of better improves.
Strategy
The same sequence is meant to operate on two different subjects, and only one of them is currently possible.
Specify runs the loop on its own product: how requirements are collected, how a limit is communicated, how a request is answered once it arrives. This is what happens today, by hand. The subject is Specify's own software, so there is nobody to ask for permission and no customer data involved.
The same inspection, pointed at a merchant's own commercial experience, showing where it falls short of a written standard and what could be prepared in response. That is the wider Automated Improvement Audit, and it does not exist yet. The narrow version running on this site reads enquiries, not storefronts.
The second level is what would make Specify valuable to a merchant rather than merely well built, and it is the one that does not exist yet. Naming it as a hypothesis is more useful here than describing it as though the work were finished.
Strategy
A worked example taken from this site rather than from a customer engagement. It is deliberately small, because the honest ones are.
Observation
Enquiries pasted into the audit on this site arrive as circumstances. They describe a space, an occasion, a constraint or a problem, and frequently name no product at all.
Research
By email and on the phone, a competent person asks two or three questions before answering anything. Those questions are never about what the customer would like to buy. They are about what would make an answer possible at all.
Standard
A commercial interface should establish what makes an answer possible before it offers one, and should show the customer what is still missing rather than guessing past it. Source: observed handling of the same requests outside any platform. Wrong if customers abandon the questions more often than they abandon a guess.
Gap
Measured against that standard, Specify's own application form failed. It collected a description and a contact address and asked nothing that would change the answer. It behaved exactly like the catalogue the rest of this site argues against.
Change
It now asks what is missing, in the order the missing information matters, and states plainly when it has enough to proceed. The form does to its own visitors what the platform proposes to do for a merchant's customers.
What can be said
That is the entire claim. Not that more people finish it, and not that the ones who do are better qualified. Those are questions the loop has not reached yet.
The sample is too small to measure. A change made against a written standard at this volume is a judgement, not a result. What this pass produced is not proof that the form is better; it is a recorded position that somebody can now test, and overturn, once there is enough traffic to do it properly.
Strategy
The fastest way to lose the value of a loop like this is to let its outputs be reported as its results. A standard is not evidence that the product is good. A prepared improvement is not evidence that it worked.
A position
A written standard, with a source and a revision condition. It states what Specify believes good looks like. It is the weakest thing on this list and the one most often presented as the strongest.
An inspection
A named gap between the standard and a real experience. It says something true about the current state. It says nothing about whether closing the gap helps.
A proposal
A prepared change with the reasoning attached. This is where most roadmaps stop and start describing the result as an improvement.
A measurement
What happened after the change, on enough volume to be capable of contradicting the proposal. This is the only rung that is evidence, and Specify does not currently stand on it.
A standard is a position, a prepared improvement is a proposal, and a measurement is the only one of the three that is evidence. Specify will never present a hypothesis as a result, on this page or in a merchant's account.
Strategy
Start with the disadvantage. A method compounds slowly, and slowly is a liability long before it is an advantage. A competitor who copies features ships faster than one who writes standards, and for the first year or two they look better for it. Everything below only holds if Specify lasts long enough for the difference to show.
A standard with a source and a revision condition can be read by somebody who was not there when it was written. Product taste cannot. Most commerce knowledge sits in the heads of a few experienced people and leaves when they do; writing it down is unglamorous, and it is the only reason any of it accumulates.
A record of what was tried and abandoned is worth more over time than a record of what shipped. Each abandoned improvement is a constraint on the next proposal. Copying a competitor's product hands you their conclusions without any of the attempts that ruled out the alternatives.
A standard written once is inspected against many different storefronts, and each inspection is an opportunity for it to fail. This is not an argument about scale: a standard that has survived ten honest inspections is worth more than one that has survived a hundred lazy ones.
What is being defended here is not a set of answers. It is the willingness to overturn them on evidence, which is the one property a competitor cannot acquire by reading the product and rebuilding it.
None of this is a moat today. It is an argument about what compounds, made by a company that has not been running long enough for anything to have compounded yet. The reason to state it this early is that a method has to be built in from the start or it never gets built at all.
Available now
Almost everything above is an intention. These are not.
There is one more, and it is the only evidence on this page that the method is real rather than intended: this website does not build if a hypothesis is written as a result. Every string of copy on every page is walked by an automated test that rejects unsupported outcome claims, monetary promises, and any description of an unbuilt capability in the present tense. It has a source, a scope, an observable signal and an enforcement point, which is the definition of a standard given three sections above. It has failed the build more than once, which is the only reason it is worth mentioning at all.
Everything else here describes how Specify intends to work, not how it has already worked. Keeping that line visible is the point of the page.
It is that Specify is built to keep finding out, and to rewrite its own standards when the evidence says the last ones were wrong. Every platform that fell behind was, at some point, confident it had the answer. The defence against that is not more confidence. It is a mechanism capable of contradicting the people who built it.
Excellence, on this account, is not the number of changes made. It is the ability to identify, validate and operationalise better commercial practice, repeatedly, long after the founders stop being the people in the room who know most about it.
Excellence is not a claim a platform makes.
It is a system that has to keep running after the launch.