Does the Patent Firm Change the Outcome?
The answer
Less than almost anything else that was measured. Across five outcomes, the firm accounts for at most 2.2% of the variation, counting firm-and-client pairings as the firm. The examiner, the client and the technology each account for more, and between 60% and 74% sits with the individual application and is not explained by any of them.
This page shows where the variation in past outcomes sits. It is not a verdict on any firm or on the value of patent counsel. It measures grant, abandonment, timing and continuation practice, not the scope or quality of what issued. “The firm” here is the USPTO correspondence account, a stand-in whose errors push the firm’s measured share toward zero.
Where the variation in an outcome sits
Each level is estimated net of the others, so these are not overlapping explanations of the same thing. This is the grant outcome; the table below carries all five.
Share of the variation in whether an application is granted
Bars run from zero. These are separately estimated components and are not parts of one whole, so they are not stacked and do not sum to 100%.
The application itself, unexplained
66.3%
The examiner
9.0%
The technology center
8.1%
The art unit
5.3%
The client
3.5%
The firm
1.5%
Two thirds of the variation is not explained by the firm, the client, the examiner or the technology. It sits with the individual application: what was claimed, what the prior art was, how the exchange went. From an exploratory analysis: this measure comes from the firm review, which had no plan written before it ran.
The firm's share, across every outcome measured
The same component, read for each outcome. Bars run from zero on the same scale as the chart above.
Granted within 36 months
2.2%
A continuing application after a restriction
1.7%
Granted
1.5%
Abandoned
1.4%
Restriction requirement
0.4%
The firm's largest share of any outcome is 2.2%, for whether a grant arrives within 36 months. This is the upper figure of two: the firm alone carries at most 1.4%, and including specific firm-and-client pairings it is 2.2%. A reader asking whether the firm matters means both, so the page uses the larger. From an exploratory analysis: this measure comes from the firm review, which had no plan written before it ran.
| Outcome | Unexplained | Examiner | Technology center | Art unit | Client | Firm |
|---|---|---|---|---|---|---|
| Restriction requirement | 64.1% | 10.4% | 16.1% | 3.9% | 1.1% | 0.4% |
| Abandoned | 73.7% | 4.6% | 4.4% | 2.7% | 4.8% | 1.4% |
| Granted | 66.3% | 9.0% | 8.1% | 5.3% | 3.5% | 1.5% |
| Granted within 36 months | 60.1% | 11.4% | 8.7% | 10.8% | 2.7% | 2.2% |
| A continuing application after a restriction | 65.2% | 1.7% | 1.0% | 0.8% | 16.6% | 1.7% |
Accounts still differ more than chance
A small share of the variation is not the same as none. Across 441 accounts large enough to measure, outcomes spread wider than they would if every account behaved identically.
How much wider accounts spread than chance alone
The bottom bar is what identical behaviour would produce. Bars run from zero, scaled to the largest.
Grant rate
3.8 times
Abandonment
4.8 times
What chance alone would give
1.0 times
About four times chance, and most of that is the clients an account files for rather than the account itself. The two findings sit together: accounts do differ, and the difference is mostly not about the firm. From an exploratory analysis: this measure comes from the firm review, which had no plan written before it ran.
Applications per registered practitioner
An account's applications divided by the practitioners of record on them. The median account files 95 applications per practitioner. At the top of that ranking the number stops describing one person's work and starts describing a roster.
Accounts above and below 600 applications per practitioner
515 accounts with a registered-practitioner count. Bars run from zero as a share of those accounts.
600 or more applications per practitioner
49 accounts
under 600 applications per practitioner
466 accounts
Two bins rather than a histogram. A five-bin version put 105 accounts in its top two bins against 131 in the top quarter of the same ranking, and subtracting one from the other would have isolated 23 accounts, so it was refused. From an exploratory analysis: this measure comes from the firm review, which had no plan written before it ran.
Continuation practice, by quarter of that ratio
Share of an account's US applications that are continuations. Bars run from zero.
highest quarter: US continuations
11.3%
third quarter: US continuations
15.8%
second quarter: US continuations
23.8%
leanest quarter: US continuations
29.2%
This is the sharpest gradient on the page: the lowest quarter of the ratio files continuations at nearly three times the rate of the highest. Whatever the ratio is picking up, it goes with how an account uses continuations. From an exploratory analysis: this measure comes from the firm review, which had no plan written before it ran.
Grant rate, by the same quarters
Raw share granted by the data date. Bars run from zero.
highest quarter: granted
53.8%
third quarter: granted
53.6%
second quarter: granted
53.6%
leanest quarter: granted
53.6%
Flat, within two tenths of a point across every quarter. The ratio separates accounts on how they practise and not on whether their applications are granted. From an exploratory analysis: this measure comes from the firm review, which had no plan written before it ran.
Grant against what the same work would predict
Raw grant rates are not comparable across accounts, because recent filings have not resolved. These compare each account with what its own mix of art units and filing years would predict, in percentage points, and report the median account in each group.
Grant against expected, by what an account looks like
Median account in each group, in percentage points against its own expected rate. Bars show the size of the gap with the sign in the number. Bars run from zero.
one client (in-house, single-client or likely captive) (159 accounts)
+1.6 pts
none of these (187 accounts)
-0.6 pts
small/micro-heavy (35%+ small or micro entities) (129 accounts)
-0.7 pts
roster of record (600+ applications per practitioner) (40 accounts)
-1.2 pts
Every group's median sits within about two points of its own expectation. Whatever distinguishes these kinds of account, it is not showing up as a grant-rate advantage. A planned description: the grouping and the comparison were planned before they ran.
The same, by how many applications the account files
Median account in each size group, in percentage points against expected. Bars run from zero.
1,000 to 1,999 (297 accounts)
+0.2 pts
2,000 to 4,999 (161 accounts)
-0.7 pts
5,000 or more (57 accounts)
-0.7 pts
Size does not separate them either. A planned description: the grouping and the comparison were planned before they ran.
Solo practices against other accounts
Solo practices file mostly for small and micro entities. Compared on the same art units, years and entity status, solo practices as a group are granted about as often as large accounts. The smallest solo practices are the exception, and these figures do not say why.
Grant against expected, solo practices and others
Against a reference group of accounts filing 1,000 applications or more. Bars show the size of the gap with the sign in the number, scaled to seven points. Bars run from zero.
Solo practices, on art unit and year
-6.4 pts
Solo practices, also on entity status
-1.0 pts
Other accounts under 100, also on entity status
-0.6 pts
The first two bars are the same accounts measured twice. Compared only on art unit and year, solo practices are granted 6.4 points less often than large accounts. Adding entity status to the comparison cells leaves 1.0 points, with an interval that includes zero. Three quarters of solo filings are for small or micro entities, and that is most of what the first bar was measuring. Other accounts under 100 applications do not differ either, so what remains goes with a solo practice's low volume rather than with small accounts in general. A planned description: the grouping and the comparison were planned before they ran.
Solo practices by volume, grant against expected
Compared on art unit, year and entity status. Bars in rose differ from the reference; grey bars do not. Scaled to twelve points. Bars run from zero.
1 to 9 (715 accounts)
-11.0 pts
10 to 49 (859 accounts)
-4.2 pts
50 to 249 (650 accounts)
-0.1 pts
250 or more (190 accounts)
-0.2 pts
The group result hides a split. The smallest solo practices are granted 11.0 points less often than large accounts even with entity status held, and practices filing 50 or more do not differ at all. These figures do not say why. Volume, experience, the kind of work taken on and the resources behind a reply are all consistent with this and none of them is tested here. A planned description: the grouping and the comparison were planned before they ran.
| Band | Accounts | Gap (points) | 95% interval | Differs |
|---|---|---|---|---|
| 1 to 9 | 715 | -11.0 | -14.2 to -8.2 | yes |
| 10 to 49 | 859 | -4.2 | -5.9 to -2.4 | yes |
| 50 to 249 | 650 | -0.1 | -1.5 to 1.4 | no |
| 250 or more | 190 | -0.2 | -2.1 to 1.9 | no |
This page counts correspondence accounts, not people. The solo practitioner page counts registered practitioners. The two use different units and their totals should not be paired.
Our own counts from the USPTO record. No firm, account, client or practitioner is named or ranked anywhere on this page. These describe what happened across past applications, they are associations rather than causes, and none of this is legal advice.
The figures on this page are available as a CSV: /data/does-the-patent-firm-change-the-outcome.csv. It is generated from the tables and charts above, so it carries the same numbers and cannot fall out of step with them.
More in this cluster
How long does a patent take?
About two to three years for the typical filer, and far longer in biotechnology. Measured over an equal window, grants got slower rather than faster. Most of the pendency is a single stage.
Read the Answer
How many rejections?
Half of granted applications saw exactly one rejection before allowance and one in five saw none. What the rejections are about, and which grounds come back in the next action.
Read the Answer
After a final rejection
Three in five continue with a request for continued examination. Seven in ten are granted within four years, one RCE is the commonest route, and almost nothing is still open at four years.
Read the Answer
How applications end
Abandonment is nearly always silence: a reply period runs out. Express withdrawal is under two percent. Whether the family continues afterwards, and why it usually does not.
Read the Answer
Does who files change it?
Small and micro entities are granted less often than large ones. More than half of that gap goes with technology mix and who files the application, and a gap remains after both. Examiners in the same art unit also differ.
Read the Answer
After a restriction requirement
Whether you ever see one depends almost entirely on your field: 52.9% of biotechnology applications against 2.5% in computer architecture. Most restricted applications are patented without a continuing application, and in a hand-read sample the examiner had rejoined the withdrawn claims in a quarter of those.
Read the Answer
Related
What Happens During Prosecution
Five questions answered from 1.3 million US utility applications rather than from folklore. How long it takes, how many rejections to expect, what follows a final rejection, how applications end without a patent, and whether who files changes the outcome.
Read the Cluster
Who practices patent law alone
24,913 practitioners list a firm of one. Following their filings instead of their roster entry, 2,521 actually file as a solo practice, and the top 10% of those make 68.2% of all solo filings.
Read the Analysis
Does who files change it?
Small and micro entities are granted less often than large ones. More than half of that gap goes with technology mix and who files the application, and a gap remains after both. Examiners in the same art unit also differ.
Read the Answer