This document discusses methods for valuing and determining standard essential patents (SEPs). It notes that while many patents are declared as SEPs, studies show that only 20-28% are actually essential. Determining essentiality is an expensive and subjective process. The document proposes using a data-driven approach combining semantic analysis of patent claims and standards documents with characteristics like inventor participation to predict SEP essentiality in a more objective manner. This allows identifying potential SEP portfolios and their probabilistic essentiality likelihoods.
How to valuate and determine standard essential patents
1. Intellectual Property
Analytics
How to Valuate and Determine Standard
Essential Patents
Dr. Tim Pohlmann
IPlytics GmbH
Ohlauer Strasse 43, 10999, Berlin
www.iplytics.com
5. Standards and innovation
“[…standardization is the joint development
of innovative technologies …]”
Baron et al. (2014): Standards, Consortia and Innovation, International Journal of
Industrial Organization, 36, pp.22-35.
6. Yearly technology contributions by 3GPP members
0
20.000
40.000
60.000
80.000
100.000
120.000
2010 2011 2012 2013 2014 2015 2016 2017 2018
3GPP contributions to 2G, 3G, 4G or 5G per year
3GPP contributions
7. Yearly technology contributions by IEEE members
0
1.000
2.000
3.000
4.000
5.000
6.000
7.000
8.000
9.000
2010 2011 2012 2013 2014 2015 2016 2017 2018
IEEE 802 WG contributions per year
IEEE 802 contributions
9. The licensing of SEPs
pooled
8%
not-pooled
92%
Pooled vs. not pooled SEPs➢In total, over 200 independent patent owners have
declared at least 100 SEPs
➢Only 8% of the SEPs are licensed through patent pools
➢The number of licensors has been growing
➢Consequently cross-licensing is complex, often
inefficient and subject to conflicts
➢Accordingly, the number of SEP litigation is increasing
11. What share of declared SEPs is actually essential?
“…studies indicates that only 20-28% of
patent families declared essential are
actually essential…”
EU Study (2017): Landscaping Standard Essential Patents. European Commission
DG GROW Unit F.5
12. How much effort is it to check essentiality?
“To identify if one of our own declared SEPs
is actually essential we need up to 3 full
time experts over at least one week...”
Patrick Hofkens, Director IPR Policy at Ericsson
13. Average costs associated to essentiality checks?
EU Study (2017): Landscaping Standard Essential Patents. European Commission DG GROW
Unit F.5 as of data delivered by the PA Consultant Group
➢ SEP claim chart reports
cost around 7,500 EUR
per patent
➢ With about 160k (as of
today) declared LTE SEPs
costs would sum up to
1,2 billion EUR
14. How accurate are essentiality checks?
Keith Mallinson "Do not Count on Accuracy in Third-Party Patent-Essentiality
Determinations " http://www.ip.finance
➢Two claim chart based SEP studies
differ in portfolio estimations by
factor 17
➢Disagreement on LTE SEPs:
Regression shows extremely weak
correlation between two studies’
results (R2=0.0008)
15. SEPs often relate to multiple standard specifications
Ø 6,84
➢SEPs‘ claims may be essential to several
standards specification
➢SEPs are declared to on average 6,84
standard specifications
17. n:m combinations of patents and standards
281,699 worldwide SEPs
1,927,854 combinations
of declared SEPs and standards
documents
declared to Ø 6,84 standards specifications
x
=
18. Royalty share
Patent owner’s SEP portfolio for LTE
Total number of worldwide SEPs for LTE
= Royalty Share
Numerator: Identifying the actual number
of SEPs is expensive and may only in some
cases - depending on the SEP portfolio
size – be economically feasible.
Denominator: Identifying the actual
number of all SEPs related to e.g. LTE
is economically not feasible.
19. Essentiality checks are subjective
“…determining overall essentiality across
thousands of patent families … with
standards such as 4G LTE is subjective and
potentially unreliable…”
Keith Mallinson, http://www.ip.finance, 2017
20. Objective essentiality are data driven
patent claims
standards
documents
standards
contributions
standards WG
attendance
SEP
declarations
➢A data driven SEP essentiality check is
objective and not biased
➢Semantic models used in big data
frameworks allow comparing millions of
document combinations humans will never
be able to analyze
➢A probabilistic model estimates the
likelihood of SEP essentiality and makes use
of SEPs claim charts to verify results
21. Data source
90 M
Patent
Documents
280.000
SEP
declarations
4 M
Standards /
Contributions
Worldwide Patents (USA, Europe, Korea, Japan, China, etc.)
• INPADOC family identifier
• IPC/CPC classifications
• Legal status analysis (granted/active/expired)
• Reassignment information
Declared Standard Essential Patents
• Patent and standards document number
• Licensing commitments (e.g. FRAND)
• Patent Pools
Standards Documents
• 2,5 M standards documents (Titl., Abstr.., supporting company)
• 1,5 M standards contributions (Titl., Abstr.., supporting company)
• Standards release and version update
22. The Universe of potential SEPs
37,380 declared
SEPs for 5G
168,863 patents in
the potential 5G
Universe
o Contributing companies
o Similar IPC/CPC
o Similar prio. date time span
23. Characteristics of essentiality
168,863 patents in
the potential 5G
Universe
1. Patent’s claims are semantically similar to corresponding proposed
Tdoc
2. Patent’s claims are semantically similar to corresponding standard
document
3. Patent’s inventors (name, surname, affiliation) participated at 5G
corresponding RAN meeting
4. Patent’s applicant/assignee contributed at 5G corresponding WG
5. Patent’s inventors (name, surname, affiliation) supported the
corresponding 5G specification
6. Patent’s IPC/CPC overlap with core IPC/CPCs in declared SEPs
7. Patent’s prio. date overlaps with core prio. date range of declared
SEPs
8. Patent has been cited by declared SEPs (excluding self-citations)
25. Semantic analysis of patent claims and standards
➢Word vector matrix (term-document
matrix) where each row corresponds
to a term (of the documents of
interest), and each column
corresponds to a document.
➢Each element (m,n) in the matrix
corresponds to the frequency that the
term m occurs in document n.
➢We apply Log Entropy as local and
global term weighting.
26. Semantic analysis of patent claims and standards
➢ Dimensionality is reduced by deleting all but
the k largest values on this diagonal, together
with the corresponding columns in the other
two matrices.
➢ This truncation process is used to generate a
k-dimensional vector space. Both terms and
documents are represented by k-dimensional
vectors in this vector space.
➢ The relatedness of any two objects
represented in the space is reflected by the
proximity of their representation vectors, in
our case: cosine measure.
28. Prediction model
logit(x) = logit(a + b1*x1 + b2*x2 + … bk*xk) = Prob(Essentiality given x1...xk)
➢ The likelihood function picks values for b1 through bk that provide the best
possible match between the model’s predictions and the actual data on
verified essential patents
➢ The weights (z values) are chosen by the computer to provide the best
possible fit between the “predicted” probability of essentiality
29. Prediction model
Predicted: SEP Verified Coef. Std. Err. z P>z [95% Conf. Interval]
inventor's meeting attendance at
RAN Meeting 0.2143 0.0185 3.58*** 0.0010 0.1781 0.2506
applicant/assignee's standard
proposals at relevant RAN Meeting 0.0016 0.0002 4.67*** 0.0010 -0.0021 -0.0011
inventor's support TS documents
0.0152 0.0045 3.36*** 0.0010 0.0063 0.0241
cited by other SEP count
0.0103 0.0075 1.37* 0.1710 -0.0251 0.0045
patent's prio. date overlap with
declared SEPs 0.6195 0.0795 1.79*** 0.0010 0.4637 0.7754
patent's CPC/IPC overlap with
declared SEPs 0.0383 0.0772 0.52 0.6010 -0.1081 0.1867
standard document-claim similarity
score 1.4153 0.1540 4.19*** 0.0010 1.1134 1.7172
proposal document-claim similarity
score 2.1700 0.1218 7.82*** 0.0010 1.9313 2.4087