Our Data Methodology: How We Verify 762 Pandas Across 87 Locations

Key Fact: PandaCommon maintains data on 762+ giant pandas across 87 locations — and every data point is verified against a hierarchical system of authoritative sources before publication. The research methodology combines the International Studbook (the global registry of captive pandas), official Chinese government and facility announcements, peer-reviewed scientific literature, zoo records, and verified news reports. When sources conflict — as they regularly do — a structured resolution protocol determines which data is published and which is flagged as uncertain. This article explains the verification framework that makes PandaCommon a trusted source in an information ecosystem where panda data is scattered across languages, institutions, and decades.

Key Takeaways

1. Data is verified against a four-tier source hierarchy — Tier A (studbook, official announcements) through Tier D (social media, community submissions).

2. Conflicting information is resolved through a structured protocol — not by editorial preference but by systematic source ranking.

3. Transparency about uncertainty is a data quality feature — acknowledging what we do not know is as important as recording what we do.

Why Panda Data Is Hard

The same panda can be known by multiple names: a Chinese name, an English translation, a nickname given by keepers, a studbook number, a media-given moniker. Its birth date may be recorded differently in the studbook, in the facility’s own records, and in news reports. Its parentage — the foundation of genetic management — can be ambiguous when records from the 1980s and 1990s were maintained on paper in different formats across different institutions.

Consider a typical panda record. The International Studbook lists the panda’s parents as Male A and Female B. The facility’s own records list Mother C. A news report from the panda’s birth year lists Male D as the father. Which is correct? The studbook is authoritative — but the studbook itself may have been corrected in a later edition. The facility records the mother’s identity at the time of birth — but the mother may have been misidentified.

This is the challenge of panda data: not a shortage of information, but a scattering of information across sources that do not always agree.

Our Source Hierarchy

To manage this complexity consistently, PandaCommon organizes sources into a four-tier hierarchy based on authority and reliability.

Tier	Source Type	Examples	Authority Level	Update Speed	Used For
A	Official registries and government announcements	International Giant Panda Studbook, CWCA official announcements, facility birth/death/transfer notifications	Highest — primary authoritative source	Hours to days	Core identity data (studbook, parentage, birth date)
B	Peer-reviewed science and direct-access media	Scientific journal publications, major Chinese media with facility access (Xinhua, China Daily), zoo annual reports	High — independently verifiable	Weeks to months	Behavioral and biological data, research findings
C	Verified news and secondary reports	Credible news outlets, zoo press releases, conservation organization reports	Moderate — may lack direct facility access	Days to weeks	Context, history, non-critical updates
D	Community and social media	Public submissions, social media posts by keepers or researchers, unofficial forums	Low — treated as leads requiring Tier A/B confirmation	Variable	Leads for investigation, corrections, supplementary information

No data point is published on the basis of a Tier D source alone. Tier D submissions are investigated, cross-referenced against Tier A or B sources, and published only when confirmed.

How Verification Works

The verification process follows a structured workflow that applies the same standards to every record, regardless of which panda or facility is involved.

Discovery. A new data point enters the system — a birth announcement, a studbook update, a community correction, a scientific paper. The source tier is identified immediately.

Cross-reference. The new data point is compared against existing records. If it confirms existing data, no further action is needed. If it conflicts, the conflict is logged and investigated.

Conflict check. If the data comes from a Tier A source, it is accepted as authoritative unless it contradicts another Tier A source — a rare but not impossible scenario. If the data comes from a Tier B or C source, it is cross-checked against the nearest Tier A source before publication.

Source ranking. For conflicting Tier A sources — for example, a studbook that lists one birth date and a facility announcement that lists another — the most recent authoritative update is accepted, and the discrepancy is documented.

Publication. Verified data is published. Unresolved conflicts are noted with an explanatory flag. Data that cannot be verified against at least a Tier C source is not published.

Handling Conflicting Information

Conflicts between sources are not bugs in the system — they are inevitable features of a data landscape spanning decades, languages, and institutions. Our resolution protocol handles the most common conflict types:

Birth date conflicts. The studbook record is authoritative for captive-born pandas. For wild-born pandas captured before comprehensive studbook records (primarily pre-1990s), the estimated age at capture is used, and the record is clearly marked as “estimated.” If a later scientific paper provides more precise age determination (for example, through dental cementum analysis), the scientific data supersedes the original estimate.

Parentage conflicts. When the studbook and facility records disagree on parentage, the studbook is authoritative — but the discrepancy is noted. Parentage records from the era before genetic confirmation (pre-2000s) may contain errors that are progressively corrected as DNA analysis is applied to historical lineages.

Name conflicts. Chinese name, English transliteration, nickname, and studbook number are all recorded as alternative identifiers. The authoritative name is the Chinese name as recorded in the studbook; alternative names are listed as aliases.

Transfer history conflicts. Transfer announcements from the exporting and importing facility are cross-referenced. If dates differ by more than a few days, the discrepancy is noted. Transfer records from the 1990s and earlier that rely on single-source documentation are flagged as having higher uncertainty.

Data Fields We Verify

Not all data fields receive the same verification intensity. Some are critical for identification; others are contextual.

Field	Verification Standard	Primary Source	Typical Accuracy	Notes
Studbook number	Must match official studbook	International Studbook	100% (verified)	The anchor identifier for all records
Chinese name	Cross-referenced with studbook and facility	Studbook + facility records	99%+	Occasional alternative characters
Birth date	Studbook for captive; estimated for wild captured	Studbook + capture records	95%+ (captive), 70% (wild estimated)	Estimated wild dates clearly marked
Parentage	Studbook authority; genetic data when available	Studbook + genetic studies	95%+	Pre-2000 records may have higher uncertainty
Death date	Facility announcement + studbook update	Official notification	99%+	Occasionally delayed reporting
Location/transfers	Cross-referenced announcement chain	Export + import facility records	95%+	Historical transfers less documented
Sex	Studbook + facility records	Studbook	99%+	Rare errors in early records
Body story	Multiple source cross-reference	Tier A/B sources	90%+	Contextual narrative, not primary data

Error Detection

We operate multiple automated and manual error detection systems:

Duplicate detection. The system cross-references studbook numbers, names (Chinese characters, Pinyin, English), birth dates, parent identities, and known locations to identify potential duplicate records. Approximately 0.5-1% of records are flagged as potential duplicates during initial processing.

Timeline anomalies. A birth date that precedes a mother’s birth by fewer than 3 years, a transfer that occurs before a previous transfer was completed, or a death date that conflicts with a later birth record — these anomalies trigger automated alerts for manual investigation.

Logical consistency checks. A panda cannot have two mothers. A death cannot precede a birth. A transfer cannot originate from a facility that the panda had not been reported at. These logical rules catch errors that source-level verification might miss.

Update Schedule

Data Type	Update Frequency	Typical Lag After Event	Primary Trigger
Births	Continuous	24-48 hours after official announcement	Facility or studbook notification
Deaths	Continuous	24-72 hours after official confirmation	Facility notification
Transfers	Continuous	24-48 hours after both export and import confirmed	Dual-source confirmation
Studbook corrections	Monthly batch	1-4 weeks after new studbook edition	Studbook publication cycle
Behavioral/biological data	As published	Weeks to months after scientific publication	New research papers
Historical records	Ongoing	Variable	Community contributions, archival research

Why Some Records Remain Uncertain

Transparency about uncertainty is a feature, not a flaw. Several categories of records carry inherent uncertainty:

Pre-studbook wild captures. Pandas brought into captivity before the International Studbook system was comprehensive (pre-1980s) often have uncertain birth years, unknown parentage, and incomplete early histories. These are marked as “estimated” or “unknown.”

Historical records with single-source documentation. A panda transferred in 1992 with only the exporting facility’s log as evidence — the importing facility’s records may have been lost. This record is marked as having “limited verification.”

Pandas known only from field DNA. The Fourth National Survey identified individual pandas through fecal DNA without ever seeing or naming them. These pandas are recorded as field identifications with studbook-like identifiers but without the full record of captive pandas.

Building a Living Panda Database

PandaCommon is not a static archive. It is continuously updated as new pandas are born, as records are corrected, as scientific understanding advances. Every birth announcement from a zoo, every studbook update, every community-submitted correction passes through the verification framework described here before it appears on the site.

The result is a database that is never complete — but is always improving. The 762 pandas, 87 locations, and thousands of data points represent not a finished product but a commitment: to maintain the most accurate, transparent, and trustworthy public panda knowledge base available.

Frequently Asked Questions

Why do official sources sometimes disagree?

Official sources can disagree for several reasons: record-keeping errors that were later corrected, different documentation standards between institutions, timing differences (a birth announced on the actual date versus the date the announcement was made), and, very rarely, transcription errors in the studbook itself. Our conflict resolution protocol handles each type systematically.

Why is some historical information missing?

Panda record-keeping was not standardized until the International Studbook system was established. Records from the 1960s-1980s, when many wild-caught pandas entered captivity, were maintained on paper in institutional formats that varied widely. Some records were simply not preserved. The gaps are not evidence of secrecy — they are evidence of a pre-digital era when global data standards did not exist.

How do you verify wild panda population counts?

Wild panda population estimates come primarily from the National Giant Panda Surveys (four surveys conducted between 1974 and 2014), which use fecal DNA analysis to identify individual pandas. PandaCommon reports the survey data with its official date and methodology, noting that population estimates are snapshots in time, not real-time counts.

Can I submit a correction to a panda’s record?

Yes. A submission form is available on each panda’s page. Submissions are logged, investigated against Tier A or B sources, and — if verified — incorporated with acknowledgment. We prioritize corrections that include source documentation (links to official announcements, studbook references, or published research).

How does PandaCommon handle AI-generated misinformation?

We do not accept AI-generated content as a source. The proliferation of AI-written panda articles and social media posts that fabricate or hallucinate data is a growing challenge. Our Tier A/B source requirement screens out the vast majority of AI-generated misinformation. When suspected AI-generated data is identified in submissions, it is rejected without further investigation.

Your Turn

The 762 pandas in this database are more than entries in a spreadsheet. Each record represents a life, a lineage, and a contribution to the survival of a species. The verification framework that supports these records is not bureaucratic overhead — it is the infrastructure of trust. Every time you read a panda’s story on this site, you are reading data that has passed through a structured verification process designed to catch errors, resolve conflicts, and acknowledge uncertainty. Read our explanation of how camera traps and AI recognition identify wild pandas to understand the detection layer, and our article on infrared camera monitoring of wild pandas to see the field infrastructure that generates much of the data we verify.