Application Organization Worksheet
Three questions that take your notes or a provisional and say which of the organizations seen in 32 US software applications the explanation fits, and what that organization's detailed description walked through. Read it below, copy it, or download it.
Built from observed structure in 32 US software non-provisional applications, and tested on 16 held-out families that were not among the 32. Figures are of documents or of families. Method and limits: how we read these filings.
Most applications in the sample were mechanism-led (20 of 32), so that is the likely starting point, and these questions help you notice whether yours is one of the exceptions. On a held-out test, answered from a provisional, they matched the organization of the later application 7 of 8 times, the same score as always answering mechanism-led, because 7 of those 8 were mechanism-led. Treat the answer as a starting suggestion, not a prediction. In an earlier test round, the screen question picked out both applications built around a person at a screen (2 of 2).
What the Worksheet Does
Every one of the 32 applications read followed the same section order. What differed was how the detailed description organized the explanation. Four organizations appear in the sample, and the questions below ask about three features of an idea that they tended to go with: whether a person's actions at a screen are part of the mechanism, whether a reader needs a concept explained before the mechanism makes sense, and whether the core is a sequence of operations or a structure.
The worksheet asks those three things of material you already have, in order, and the first yes decides. It describes the shape of that material, and the test below is the reason to read its answer as a description rather than a forecast.
How the Questions Did on Held-Out Families
The questions were written from the 32 applications. They were then answered from the provisional alone for 16 other families, in two sets of 8, and each answer was compared with the organization a separate evaluator assigned to the later application. The baseline is the answer that needs no questions: always say mechanism-led, the most common organization in the 32 (20 of 32).
| Held-out set | The questions | Always mechanism-led |
|---|---|---|
| Second set, the reported test (nothing revised after it ran) | 7 of 8 | 7 of 8 |
| First set, used to revise the questions | 5 of 8 | 5 of 8 |
- The misses. The second set's one miss was its one problem-led application, answered mechanism-led. The first set's three misses were all on the problem-led question: two provisionals answered problem-led were followed by mechanism-led applications, and one problem-led application was answered mechanism-led.
- Where it did better than the default. The first set held 2 interface-centered applications among its 8 families, and the questions picked out both. The default misses both by construction. The second set held none, so it could not repeat that result.
- What was not measured. The scoring counted a mechanism-led answer as a match whichever emphasis it named, procedure-led or component-led. Nothing in this test shows whether the emphasis was the right one.
The test's own conclusion is the one to carry: the organization of the later application is mostly settled when that application is written, not by anything in the earlier material. Use the worksheet to see the shape of what you have. Do not use it to predict what will be filed.
The Worksheet
Start Here
How to use the answer: as a starting suggestion. Most applications in the sample were mechanism-led (20 of 32), and these questions help you notice whether yours is an exception. On a held-out test they matched the later application's organization 7 of 8 times, the same as always answering mechanism-led. Basis: observed in 32 US software applications.
The Material You Are Reading From
Name what the answers come from (notes, a design document, a provisional, something else) and its date. Answer every question from that material alone, not from what you plan to add later.
Question 1: A Person at a Screen
Take the questions in order; the first yes decides. For each answer, quote or point to the passages it relies on, so someone else can check it. Where the material cannot answer, write "cannot tell" and note what is absent. Do not guess.
Is any step of how the idea works something a person sees or does at a screen? Yes · No · Cannot tell
A screen counts only when it is a step of the mechanism, not a way in or a way out of it. A settings page or a results view is a no when the idea works the same without it. The test: would the account of how the idea works have holes if the screens were left out? If the material describes no screen at all, the answer is no.
Question 1: Passages Relied On
Quote or point to the passages. A yes means the explanation fits the interface-centered walkthrough, and the worksheet ends here.
Question 2: Explanation Before the Mechanism
Two conditions, answered separately. The explanation sits toward the problem-led end of the scale only when both hold. The one-fifth figure is a starting point, not a measured boundary; how the share is counted was revised once, on held-out material.
Roughly what share of the material explains the setting? Estimate
Count only passages whose subject is the setting, the existing approaches, or what a term means. Leave out any explanatory sentence inside the description of a step or a part, and leave out boilerplate. Is the share about a fifth of the material's length or more? Give the estimate and the passages counted.
Does the mechanism use a concept that has to be explained before the mechanism can be described? Yes · No · Cannot tell
Name the concept if there is one, and where the material first explains it. A long background that a specialist would skip does not make this a yes: length is not need.
Result of Question 2
Both conditions hold: toward the problem-led end. Either one does not: go on to question 3.
Question 3: Sequence or Structure
Material that reaches this question fits the mechanism-led organization. The answer names the emphasis inside it. On the held-out test the emphasis was not scored.
Is the core of the idea mostly a sequence of operations, or mostly a structure? Sequence · Structure · Cannot tell
A structure is parts and how they connect, or elements and the rules that relate them, such as a layout or a schema. For a learned model, answer for what the material says is new: a training or inference procedure is a sequence; an architecture or an input representation is a structure. A pipeline drawn as boxes is still a sequence when the boxes are stages run in order.
Question 3: Passages Relied On
Quote or point to the passages. A sequence points to the procedure-led emphasis; a structure points to the component-led emphasis.
What the Answer Points To
A fourth organization seen in the 32, claim-dominant, is chosen by no question. It is a matter of how much of a document is claims, which is settled when an application is written, not by the idea.
The Organization the Explanation Fits
Interface-centered walkthrough; toward the problem-led end; mechanism-led with a procedure-led emphasis; mechanism-led with a component-led emphasis; or cannot tell, with what is absent.
What That Organization's Detailed Description Walked Through
Mechanism-led (20 of 32): a walk through the figures one at a time, carrying the parts, what each does, the steps and the data; procedure-led follows the method's flowchart step by step, component-led follows the system's block diagram part by part. Problem-led (4 of 32): an extended account of the setting before and around the mechanism, with definitions and comparisons with other approaches. Interface-centered (6 of 32): the mechanism alongside a walk through the screens, each with what it shows, what the person does and what the system does in response.
Open Points
Anything the material could not answer, and anything about the result you want to raise with a registered patent practitioner.
The Four Organizations, in 32 Applications
The groups were formed from a profile of each application's content and structure, and they differ in what the content is made of, not in which sections appear: the same section order runs through all four. Element shares below are medians, in this reading.
| Organization | Applications | What the detailed description walked through |
|---|---|---|
| Mechanism-led | 20 of 32 | A walk through the figures, one at a time. Structure, behavior, procedure and data were 39% of elements, in this reading. Background is short. Procedure-led: the method's flowchart, step by step, each step with its inputs, conditions and alternatives. Component-led: the system's block diagram, part by part, each with what it is, what it holds and what it is configured to do. |
| Problem-led | 4 of 32 | An extended account of the setting before and around the mechanism, with definitions and comparisons with other approaches. Problem and context were 20% of elements, against 5% in the other groups, in this reading. 3 of the 4 have no summary section. |
| Interface-centered | 6 of 32 | The mechanism alongside a walk through the screens: what each shows, what the person does, what the system does in response. People and interface were 7% of elements, against 0% to 3% in the other groups, in this reading. All 6 have interface figures among their drawings, against 6 of the other 26, and all 6 have a summary. |
| Claim-dominant | 2 of 32 | A large claim set with a short description that restates each claim element with its alternatives and scope language. Claims were 37% of elements, in this reading. Two documents, and no question here selects it. |
Section presence across the 32: a title, a detailed description, claims and an abstract in all 32; a brief description of the drawings in 32 of 32; a background in 31 of 32; a cross-reference to related applications in 27 of 32; a summary in 25 of 32. Beyond the tendencies in the table, the sample shows no property of the idea that decides which sections appear, so the worksheet asks nothing about sections.
One Distinct Kind, Two Ends of a Scale
The four groups were formed by three independent grouping runs, and a group was kept only where at least two runs agreed. Each group was then tested by removing its members and grouping what remained, by a run that had never seen the four groups.
That is why question 1 is a yes-or-no question and question 2 is a question of degree with a starting threshold. The procedure-led and component-led emphases are softer still: the split came from one of the three original runs and recurred in two of the three removal runs that kept the mechanism-led documents. It was never a majority, which is why the worksheet offers them as emphases inside one organization and not as organizations of their own.
Question 1, Answered Two Ways
Two answers recorded in the first held-out set show where the line falls. Both come from the provisional alone, and both matched the later application. Families are identified by the sample's own reference numbers.
- Family F06, yes. A translation tool in which a translator edits a displayed text fragment in an editing region of the interface. That edit is what sends the text back to be split into sub-fragments, and selecting a sub-fragment is what feeds it to the suggestion engines again. The recorded answer reasoned that leaving the screens out would leave a hole in the account of the mechanism, because the edit happens there. The suggestion engines behind it would work without a screen, but they are what the edit acts on. Its later application was interface-centered.
- Family F16, no. A system that measures activity on kiosk devices, with a reporting interface where a user can review a multi-device view. The provisional also describes a software module that builds the same view for automated analysis, so the screen is a way out of the mechanism, not a step of it, and leaving it out leaves the account of collection, learning, detection and remediation whole. Its later application was mechanism-led, and the answer named the component-led emphasis.
What the Questions Cannot See
- An explanation written after the material. An application can become problem-led when the setting is written later. In the first held-out set, one problem-led application followed a provisional that spent 16% of its content on the setting, in this reading, below the starting point. No question about the earlier material can choose problem-led there.
- A borderline concept. In the second set's one miss, the material spent a fifth or more of its length on the setting, but the recorded answer found no concept the mechanism needed explained first, and noted its own uncertainty on exactly that point.
- The threshold itself. In the 32 families, the 4 provisionals followed by problem-led applications spent 22% to 37% of their content on the problem, the setting and the field's terms, and 18 of the 20 followed by mechanism-led applications spent 18% or less, in this reading. The one-fifth starting point sits between those, and it is a starting point, not a finding.
- Claim-dominant applications. Nothing in an idea or in a provisional selects one. A held-out family whose application was claim-dominant would count as a miss.
What This Rests On
- 32 applications in one technology area, software. It is not a survey of US practice, and the organizations are observed patterns, not types anyone is asked to choose from.
- 16 held-out families in two sets of 8. Each set's result turns on one or two families.
- Element shares describe this reading of the documents, not a property fixed in the filings.
Related
- What patent applications contain, the section order and the four organizations as findings
- How we read these filings, the sampling, the definitions and the limits
- What real provisional applications contain, the same reading applied to the provisionals
- Idea Disclosure Template, a plain-language form for the material these questions read
- Showing possession of what you claim
Scope. This worksheet describes how a body of technical material is organized, against organizations observed in 32 US software applications. It does not say how any application ought to be organized or drafted, what to file, or whether anything is patentable, and it gives no verdict on any filing.
Disclaimer. This worksheet is educational and descriptive. It and any output are not a substitute for advice from a registered patent practitioner. Use of this worksheet does not constitute legal services or legal, patent, patentability, drafting, filing-strategy or deadline advice, and does not create an attorney-client relationship with Obviously Not. Decisions about what to file, when, and how an application is drafted belong with a registered patent practitioner.