A patent infringement analysis methodology is six stages, and the mapping everyone thinks of is the third.
Skipping the stages around it is what produces wrong answers. A chart built without construing the terms first changes the moment a court reads them differently. A chart with a gap, analysed without checking estoppel, reaches an equivalents conclusion that may not be available at all.
The order matters as much as the content. Verify before you construe, construe before you map, map before you consider equivalents, and check estoppel before you rely on them.
The six stages
| # | Stage | Output |
|---|---|---|
| 1 | Verify | Is the patent in force, and who owns it? |
| 2 | Construe | What do the disputed terms mean? |
| 3 | Map | Element by element, with cited evidence |
| 4 | Equivalents | Are the gaps insubstantial? |
| 5 | Estoppel | Is the equivalents argument available? |
| 6 | Conclude | A reasoned written outcome |
Stage one alone resolves a large share of questions. Nearly three in five US utility patents lapse before term, so a meaningful proportion of accused patents are already dead.
Stage 1: verify
| Check | Where | Time |
|---|---|---|
| Patent exists as cited | Google Patents | 2 min |
| In force | USPTO Patent Center | 3 min |
| Expiry date | Front page plus PTA | 5 min |
| Current owner | USPTO Assignment Search | 5 min |
| Pending continuations | Family view | 5 min |
| Terminal disclaimer | Front page | 2 min |
Do this before anything substantive. Twenty minutes, all free, and it occasionally ends the analysis outright.
| Ipiry Patent Survival Curve v1.0 | Rate |
|---|---|
| Survive the 3.5-year fee (2022 cohort) | 85.8% |
| Survive the 7.5-year fee (2018 cohort) | 64.6% |
| Reach full term (2014 cohort) | 41.4% |
| Abandoned before full term | 58.6% |
Computed from 27,273,654 USPTO maintenance fee records covering 8,262,336 US utility patents — see the patent survival curve.
Pending continuations change the risk assessment even where the granted claims are clearly avoided, because new claims can still be written toward your product.
Stage 2: construe
Decide what the words mean before comparing anything against them.
| Source | Priority |
|---|---|
| The claims themselves | Highest, including claim differentiation |
| The specification | Very high |
| The prosecution history | High |
| Extrinsic evidence | Lowest |
Identify only the terms that matter. A term where both readings produce the same outcome does not need construing, however interesting the linguistic question.
Note the transition word. "Comprising" is open, so extra features in the product are irrelevant. "Consisting of" is closed.
Record the construction you adopted and why. An analysis that does not state its construction cannot be checked or defended. See claim construction.
Stage 3: map
| Column | Contains |
|---|---|
| Element | The exact claim words, one element per row |
| Accused feature | The specific corresponding feature |
| Evidence | Source with a page or figure reference |
| Grade | Present / absent / arguable / unverified |
| Notes | Construction applied, alternatives considered |
One row per element, always. The discipline is what prevents overall-similarity reasoning.
Split at every limitation, not every clause. "Biased toward the inlet by a spring" contains two requirements and both must be present.
Grade "unverified" rather than assuming. An internal component you cannot observe is unverified, and saying so tells the reader exactly what evidence would settle it. See product mapping patent infringement.
Stage 4: equivalents
Only for elements not literally present.
| Test | Question |
|---|---|
| Function-way-result | Same function, substantially the same way, same result? |
| Insubstantial differences | Would a skilled person regard the difference as insubstantial? |
It applies element by element, not to the invention as a whole. You cannot argue the products are equivalent overall.
A wholly missing element cannot be bridged. Equivalents substitutes a different element for a claimed one; it does not excuse having none.
Do not conclude at this stage. The next one decides whether the argument is available.
Stage 5: estoppel
The stage most often skipped, and it frequently decides the case.
| Prosecution event | Effect |
|---|---|
| Narrowing amendment for patentability | Festo presumption — equivalents in the surrendered range barred |
| Amendment with no stated reason | Presumed patentability-related |
| Clear argument distinguishing prior art | Argument-based estoppel |
| Amendment for §112 clarity | Still counts |
| No amendment, no argument | No estoppel |
The three Festo rebuttals rarely succeed — unforeseeability, tangential relation, and some other reason.
This is decided from a public document. The file wrapper is free in Patent Center, and establishing estoppel needs no discovery, no experts and no trial. See prosecution history estoppel.
Stage 6: conclude
| Output | Purpose |
|---|---|
| Claim chart | The evidence |
| Reasoned summary | What it means |
| Construction stated | Makes the conclusion checkable |
| Limitations stated | What could not be verified |
| Written opinion of counsel | Willfulness defence |
State the construction you applied. A conclusion that depends on a reading of one term should say so, because a different construction changes it.
State what you could not verify. An honest gap is more useful than a confident guess and it protects the analysis.
A written opinion obtained before launch is the principal defence against enhanced damages, which can reach three times the award. It must predate the conduct to do its job. See freedom to operate.
Worked example: the stages changing the answer
A pressure valve claim, an accused competitor product.
| Stage | Finding | Running conclusion |
|---|---|---|
| 1. Verify | In force, 8 years left, owner changed twice | Proceed |
| 2. Construe | "During transition" disputed | Two readings possible |
| 3. Map | 5 of 6 elements present; sixth is magnetic, not spring | Not literal |
| 4. Equivalents | Function-way-result arguably satisfied | Possibly infringing |
| 5. Estoppel | Claim amended to "spring-biased" to overcome prior art | Barred |
| 6. Conclude | No literal infringement; equivalents estopped | Not infringing |
The answer reversed at stage five. An analysis stopping at stage four would have concluded possible infringement and recommended a licence.
The estoppel was found in a free public document in under an hour.
Stage two mattered too. Had "during transition" been construed broadly, the mapping at stage three would have differed on another element.
Analysing method claims
Method claims add a question the others do not.
| Extra step | Detail |
|---|---|
| Identify who performs each step | Add a performer column |
| One party performs all | Direct infringement |
| Steps split across parties | Divided infringement |
| Direction or control established? | Required to attribute |
| Joint enterprise? | Alternative basis |
Divided infringement is a real defence for architectures where a service performs some steps and a customer performs others.
Map the actor for every step, then ask whether one party directs or controls the others. An ordinary vendor-customer relationship generally does not meet that.
Screening versus full analysis
| Screening | Full analysis | |
|---|---|---|
| Time | A day | Substantially longer |
| Construction | Assumed | Argued |
| Mapping | Independent claims | All relevant claims |
| Equivalents | Noted | Analysed |
| Estoppel | Checked | Analysed |
| Output | Internal note | Written opinion |
| Who does it | Anyone competent | Counsel |
Screen first, always. Most candidates resolve at screening — expired patents, plainly avoided claims, wrong jurisdiction.
Full analysis is reserved for what survives, which is what keeps the cost proportionate.
Evidence quality
| Grade | Source |
|---|---|
| Strongest | The target's own technical documentation and filings |
| Strong | Regulatory submissions, published standards compliance |
| Good | Teardowns, independent testing, specification sheets |
| Moderate | Marketing material with specific technical claims |
| Weak | Screenshots, forum posts, inference |
| Unusable | Assumption stated as fact |
The target's own patent filings are the best source available. Companies describe their approach accurately in their applications, and those documents are public.
Marketing material is admissible and imprecise. A brochure saying the product uses adaptive control supports the element loosely; a service manual showing the control loop supports it properly.
Date every source. A product changes, and evidence from a 2022 manual may not describe the 2026 version.
Common failures
| Failure | Consequence |
|---|---|
| Comparing overall similarity | Wrong in both directions |
| Mapping against the abstract | Wrong scope entirely |
| Treating extra features as avoiding the claim | Wrong — "comprising" is open |
| Skipping construction | Conclusion collapses when terms are read |
| Skipping estoppel | Equivalents relied on when barred |
| Assuming unobservable elements | Unsupportable |
| Not checking in-force status | Analysing a dead patent |
Extra features never avoid infringement. Almost every claim uses "comprising", so a product containing everything claimed plus more still infringes. See infringe the patent.
When to stop
| Trigger | Action |
|---|---|
| Patent expired | Stop |
| An element clearly absent, estoppel bars equivalents | Stop |
| Wrong jurisdiction | Stop |
| Evidence gap that no public source can fill | Record and stop |
Stopping early is the point of screening. Continuing past a dispositive finding spends budget on certainty you already have.
Documenting the work
| Record | Why |
|---|---|
| Date of every source retrieved | Products change |
| Construction adopted, with reasoning | Makes it checkable |
| Claims analysed and claims skipped | Shows the scope of the work |
| Evidence gaps | Honest limitations |
| Who did the analysis | Privilege considerations |
Analyses done in anticipation of litigation may be privileged; the same work done routinely may not be. Involving counsel early affects that.
Where automation helps
| Stage | Automation |
|---|---|
| 1. Verify | Substantial — status and ownership lookups |
| 2. Construe | Limited |
| 3. Map | Substantial — first-draft charts |
| 4. Equivalents | Limited |
| 5. Estoppel | Retrieval helps; analysis does not |
| 6. Conclude | Human |
Mapping is where automation earns its place, and every cited source still needs verification. A fabricated or misread citation is worse than an empty row.
Construction and estoppel stay human because both involve predicting how a court will read words and apply doctrine.
Infringement analysis methodology: the checklist
- Verify in-force status first. It resolves a large share of questions for nothing.
- Check ownership and pending continuations before assessing risk.
- Construe the disputed terms before mapping, and record the construction.
- Note the transition word. "Comprising" means extra features never help.
- Map one element per row, splitting at every limitation.
- Cite a locatable source for every row.
- Grade honestly, using "unverified" rather than assuming.
- Apply equivalents element by element, never to the invention overall.
- Check the prosecution history before relying on equivalents. Estoppel is free to establish and frequently decisive.
- State the construction and the limitations in the conclusion, and obtain a written opinion where the risk is real.