FSIS Establishment Number Lookup: Company Mapping Guide

Executive Summary

An FSIS establishment number identifies a federal grant of inspection at a plant. It does not identify a company, a brand or a parent. The USDA Food Safety and Inspection Service (FSIS) describes its Meat, Poultry and Egg Product Inspection (MPI) Directory as a listing of establishments that produce meat, poultry and/or egg products regulated by FSIS ^[1], updated weekly with each edition replacing the last ^[1]. The FSIS page, last updated October 19, 2026, cites more than 7,100 federally inspected establishments ^[1]. Our tally of the October 19, 2026 CSV found 7,255 rows, one per plant record. A legal-research center summarizing FSIS data reported 6,413 establishments in August 2026 ^[2], and a commercial directory reports 7,592 facilities ^[3], a self-reported figure that should not be treated as an official count.

Three rules explain why a plant record is a lead and not proof of ownership. The number on a meat label is that of the establishment where the product was processed under inspection ^[4], while a distributor must be named with a qualifying phrase ^[5], so the branded company on a package can differ from the plant operator. The grant attaches to the applicant "whether tenant, subsidiary, or landlord" ^[6], and a change of ownership requires a new application ^[6]. FSIS also says its size variable "does not correlate highly with production volume" ^[7], and the size classes do not account for whether an establishment belongs to a larger corporate structure ^[8]. HACCP (Hazard Analysis and Critical Control Points) size class is therefore not a revenue screen.

Scope limits matter before any universe is counted. Cooperative Interstate Shipment (CIS) plants appear on a separate FSIS list ^[9], and thirty-one states run State MPI programs ^[10]. USDA's National Agricultural Statistics Service (NASS) counted 1,127 federally inspected slaughter plants and 1,796 under other inspection on January 1, 2026 ^[11]. Beneficial ownership reporting to FinCEN no longer applies to US-created companies ^[12], so operator names must be tied to entities through state registries, securities filings, Legal Entity Identifier data and trademark records.

The reference method is a plant-to-entity crosswalk with explicit evidence statuses, plant count and company count reported separately, and an unresolved rate defined as unresolved plant clusters divided by all plant clusters. One metadata discrepancy deserves attention: the Data.gov record lists a monthly update period and a modified date of January 22, 2025 ^[13], while its description and FSIS say weekly ^[13]. Record the edition date of the file you downloaded, not the catalog date.

Introduction and Background

A private equity origination analyst building a food-processing target universe usually starts with a question that sounds simple: which USDA-inspected plants belong in scope, and how many distinct companies do they represent? The FSIS directory is the obvious starting point because it is free, official and plant-level. It is also easy to misread. A count of rows is a count of plants. A count of distinct names is a count of name strings. Neither is a count of acquirable companies.

This report is written for origination teams that need a defensible crosswalk from FSIS establishment numbers to operating entities and, where evidence supports it, to parents. It answers the practical lookup questions behind queries such as "FSIS establishment number lookup," "FSIS establishment directory search" and "USDA establishment number lookup," then moves to the harder question of mapping plants to companies.

Docket, the publisher of this report, is a direct provider of sourced target research for private equity firms, delivered through its Triage, Scout and Audit agents in self-serve and managed forms. Use the cited independent sources to assess factual claims; the publisher’s service description is not independent evidence. The position taken here is deliberately conservative: a plant record is a lead, not proof of parent ownership, and it is certainly not evidence of a sale process.

On evidence handling, this edition reflects pages and files retrieved on October 20, 2026. Direct automated requests to some FSIS pages had returned HTTP 403 in an earlier research run on September 24, 2026, and no 2026 file had been inspected then. The directory page, both documentation files and the current CSV were subsequently retrieved through a rendering fetch service on October 20, 2026, and the field layout and tallies below come from that retrieval.

Identifiers and Entities: A Taxonomy

The central discipline is to keep six different things apart. Each answers a different question, and each has its own source of evidence.

• Establishment number: the identifier of a grant of inspection. FSIS says every number is preceded by a prefix indicating the type of grant ^[7].
• Facility (plant): the physical location. In the directory file, each row is one establishment: the October 19, 2026 edition had 7,255 rows and 7,255 unique establishment_id values. The Census Bureau uses the same logic: an establishment is a single physical location, and is not necessarily equivalent to a company or enterprise ^[14].
• Operator or licensee: the person or firm that applied for and holds the grant. Regulation requires the application from the person conducting operations, "whether tenant, subsidiary, or landlord" ^[6].
• Brand: the name consumers see on the pack. A label may show a distributor, and where the product is prepared by one company and distributed by another, a qualifying phrase such as "Prepared for" or "Distributed by" must precede the name ^[5].
• Legal entity: the registered company, with a state of formation, a registered agent and a status. State Secretary of State offices register domestic and foreign corporations ^[15].
• Parent: an entity that owns or controls the legal entity. The Census Bureau defines an enterprise as one or more establishments under common ownership or control ^[16].

What a package EST number does and does not say

For meat, the regulation allows the establishment number to appear inside the inspection legend or elsewhere with the prefix "EST", and it identifies the establishment in which the product was processed under inspection ^[4]. FSIS's labeling guide says the number is used to identify all containers of inspected products prepared in the plant ^[17]. It identifies a producing plant. It does not tell an analyst who owns the brand on the front of the pack, who contracts the production, or who ultimately controls the plant.

Contract manufacturing is the reason this gap is not rare. A University of Florida extension publication describes a contract packager, commonly called a co-packer, as a firm that often processes clients' products under a private label for a fee ^[18]. That source covers food manufacturing generally, so prevalence among FSIS plants is not quantified here.

One-to-many and many-to-one relationships

The relationships between plants and companies run in both directions. The text diagrams below use fictional placeholder names.

ONE-TO-MANY (one company, several plants and several numbers)

  Legal entity: Example Foods LLC (fictional)
      |-- M1111          Plant A, State X  (meat processing)
      |-- M2222+P2222    Plant B, State Y  (meat + poultry grants)
      |-- P3333          Plant C, State Z  (poultry slaughter)

  Three plant rows, four grant numbers, one company, if and only if the
  evidence ladder below supports the grouping.

MANY-TO-ONE (several operators, one site, or one name, several sites)

  Site: 100 Industrial Way, Town T (fictional)
      |-- M4444   Operator: Alpha Meats Inc.   (tenant)
      |-- P4445   Operator: Beta Poultry LLC   (tenant)
      Landlord: Gamma Realty LLC (no inspection grant)

  Two operators and one landlord share one address. Grouping by address
  would wrongly merge two companies. Grouping by name alone can also
  merge unrelated firms with the same name in different states.

The directory itself shows how common the problem is. In our tally of the October 19, 2026 file, 406 establishment names appear on more than one row (covering 1,387 rows). FSIS states that it lists the physical location of each establishment in the directory ^[1]. For this protocol, compare nonempty street values using street.strip().lower() and review repeated street strings separately from full-address, physical-site and ownership matches.

Number stability over time

An establishment number is not a stable company key. A 1992 FSIS directive states that when a company is sold, the number can transfer, but a new grant of inspection is issued to reflect the new ownership ^[19], and that the number remains with that operation as long as it stays in production. Current regulation likewise requires a new application when ownership or location changes ^[6]. A stable number can therefore sit across two different operators over time, so each row needs an "as of" edition date, and the grant date field (the most recent grant, not the original) is a review trigger.

Obtaining and Reading the FSIS Directory

The directory is published on a single FSIS page with CSV and PDF versions sorted by name or number. The page labels its numeric CSV "MPI Directory: Numerically by Establishment Number (CSV)", and separately offers a demographic supplement titled "Dataset: Establishment Demographic Data". FSIS also publishes a list of establishments inactive for three years or less ^[1] and a JSON application programming interface (API) at its MPI endpoint ^[20].

Update cadence and the Data.gov discrepancy

FSIS states that the data are updated weekly and that the current edition replaces all previous editions. Corrections to the data, it adds, appear in the next weekly update, which typically occurs on Mondays. The page and files we retrieved were dated October 19, 2026, a Monday.

The federal catalog record tells a slightly different story. The Data.gov harvest record for the directory carries an accrual periodicity of "R/P1M", which is a monthly cadence, and a modified date of 2025-01-22 ^[13]. The same record's description repeats the weekly claim ^[13]. We treat the FSIS page as authoritative for current cadence and the catalog fields as stale metadata. The practical rule for analysts: store the edition date printed on the FSIS page and file, never the catalog's modified date. The Data.gov dataset landing page itself did not return content when we requested it, so the catalog fields cited here come from the harvest record.

Schema as retrieved on October 19, 2026

The numeric CSV header contains 19 columns. The last ten, as printed in the header, are grant_date, activities, dbas, district, circuit, size, latitude, longitude, county and fips_code ^[21]. The first nine are establishment_id, establishment_number, establishment_name, duns_number, street, city, state, zip and phone. Selected field definitions from FSIS documentation:

• establishment_number: a prefix (M meat, P poultry, I imports, G eggs, V voluntary) plus digits. An establishment with several grants carries several numbers joined by a plus sign ^[7]. Some numbers carry a suffix letter.
• establishment_name: the name on the FSIS grant of inspection, which is not necessarily the brand or parent.
• dbas: the legal name used to carry out business activities ^[7]. It is the best alias field for brand matching, but see the quality notes below.
• duns_number: optional, as FSIS notes that the number is not required.
• activities: the inspection activities the establishment is approved for under its grant.
• size: the HACCP size class discussed next.

What "HACCP size" means

FSIS collects the size variable only for establishments under HACCP ^[1]. The classes originate in the 1996 HACCP final rule: very small establishments have fewer than 10 employees or annual sales under $2.5 million ^[22], large establishments have 500 or more employees ^[22], and small establishments fall between ^[22]. FSIS warns that the variable does not correlate highly with production volume. A March 2026 FSIS advance notice of proposed rulemaking explains that the classes are used to analyze regulatory impact and tailor assistance, and that they do not account for whether the establishment is part of a larger corporate structure ^[8]. It reports counts as of December 2025 of 2,961 very small, 2,847 small and 508 large establishments ^[8].

Two conclusions follow. First, a "Large" label means the site meets an employee threshold, not that the owner has a particular revenue. Second, a "Very Small" plant can belong to a large corporate group, because the class is assigned at the establishment level. Size class may be recorded as a descriptive attribute of a plant. It should never be a proxy for target revenue or enterprise size.

The demographic supplement

The supplement documentation describes more than 200 variables ^[23], including volume categories for processed products, exemption flags such as meat_exemption_custom_processing and activity flags based on the last 360 days, which can differ from the current activities field. The exemption flags describe exemptions applied for and granted to inspected establishments. They are not a list of exempt firms. Use the supplement for filters on species and volume band only after reading its documentation.

Scope Limits: Who Is Not in the Federal Directory

A target universe is only as good as its boundary. Four boundaries need explicit treatment before any filter is applied.

State-inspected and CIS plants

Establishments can choose federal or state inspection, and state programs must operate in a way that is at least as effective as federal requirements, according to the National Agricultural Law Center ^[2]. Thirty-one states currently operate State MPI programs, 11 of which also participate in CIS, according to an FSIS update dated September 18, 2026. The Law Center reported 10 CIS states as of August 2026 ^[2], so the participation count moved within weeks, or the two sources count on different bases. Interstate shipment through CIS is available to establishments individually selected for the program. A plant must apply through its state MPI agency, receive a state recommendation, undergo FSIS District Office review and acceptance, and meet the federal requirements for selection. Location in a participating state alone does not establish eligibility to ship interstate. FSIS publishes a separate list of participating establishments ^[24]. Our review of the directory documentation found no statement that ordinary state-only plants are in the main file, so the safe assumption is that they are absent and need a state-by-state source.

The scale of the omission is visible in NASS data. On January 1, 2026, NASS counted 1,127 plants slaughtering under federal inspection and 1,796 under other inspection, for 2,923 in total ^[11]. These figures cover slaughter plants only, so they do not size the processing universe, but they show that a federal-only view of slaughter excludes a majority of plant locations by count. Plant count is not volume. NASS reports that in 2025 the 11 largest federally inspected plants accounted for 47 percent of cattle slaughtered ^[11]. Location counts and volume shares therefore tell different stories.

Exempt operations

Retail and custom operations are handled as exemptions in the regulations. Custom operations, for example, remain subject to the requirement that establishments be maintained and operated in accordance with sanitation provisions ^[25]. Poultry has its own exemptions, and processing more than 20,000 poultry is a disqualifier for several of them ^[26]. As far as the FSIS documentation shows, an operator holding only an exemption has no grant of inspection and therefore no establishment number to look up.

Product jurisdiction

Use the FSIS decision chart as a guide, including its species and product exclusions and subsequent conditions; meat percentages alone are not a definitive jurisdiction determination. For a definitive amenability determination, submit the product formula and processing procedures through askFSIS, as directed in the chart. FDA product jurisdiction alone does not establish directory absence: some animals under FDA slaughter jurisdiction receive voluntary FSIS inspection ^[27]. The directory also covers locations where FSIS provides other inspection services ^[7]. For this protocol, check the grant and activity fields before excluding a facility, including game or exotic-animal operations. Facilities regulated exclusively and throughout by USDA are exempt from FDA food facility registration ^[28], while an FDA question-and-answer document notes that a facility that is not regulated exclusively by USDA would not be exempt ^[29]. FDA food-facility registration (FFR) information is not public; a cross-check requires lawfully obtained company-provided records or authorized access, with that provenance recorded. The public FEI identifier lookup is separate from FFR access ^[30]. For ordinary public research, use inspection and recall data ^[31] alongside corporate records.

Activity filters

Activity strings in the file include entries such as meat slaughter, poultry slaughter, meat processing, poultry processing and egg product. Because the activities field shows approved activities and the supplement uses a 360-day look-back, a plant may be approved for an activity it did not perform recently. Define the thesis first, then apply filters, and document which field and edition drove each exclusion.

From Plant to Company: The Crosswalk Method

The reference asset is an establishment-to-entity crosswalk. It converts a weekly plant file into a controlled list of entities, each with an evidence status. The steps below are ordered so that lookups come before groupings and groupings come before counts.

Step 1: Define the thesis and boundaries

Write the universe definition as four fields: product activity, geography, federal inspection scope (federal grants only, or federal plus CIS plus state lists) and company type. Record what is out of scope, such as exempt operators, FDA-only foods and unlisted state plants.

Step 2: Download, date and archive

Store the file URL, retrieval date, edition date and header row for each pull. Keep the raw file unmodified, and archive the supplement and the inactive list from the same week.

Step 3: Normalize numbers and addresses

Apply four rules:

• Split, but keep the original: break composite numbers on the plus sign into grant components and retain the raw string.
• Preserve prefixes and suffix letters: a suffix denotes a distinct establishment.
• Standardize without overwriting: normalize street suffixes and unit designators, never the raw address.
• Flag shared locations: mark clusters where several numbers share one address.

Step 4: Cluster plants without forcing a company match

FSIS states that each dataset row represents one regulated establishment ^[7]. For this protocol, keep each establishment_id record as the starting plant unit. Use shared normalized names, addresses and DUNS values only to generate review candidates. Merge records only when documented evidence shows that they represent the same physical operation; record the evidence and retain every original identifier. Keep company groupings separate from plant reconciliation. Calculate the unresolved denominator from the in-scope plant units remaining after that review.

Step 5: Cross-check operators against independent evidence

The ladder below ranks the evidence and states what each rung proves. Table 1 then summarizes the lookup routes available to a team, and Table 2 sets out the evidence statuses used by the crosswalk.

• State registries: confirm the legal name, status and registered agent. NASS provides a route to each state's corporate registration and name database ^[15].
• Securities filings: a registrant's list of subsidiaries must give each subsidiary's jurisdiction of organization ^[32], but registrants may omit subsidiaries that are not significant in the aggregate ^[32]. Absence from an Exhibit 21 list is not evidence of non-ownership.
• Legal Entity Identifier (LEI) data: the Global LEI Foundation's Level 2 data describe direct and ultimate accounting-consolidating parents ^[33]. For this protocol, retain relationship type, dates and reporting exceptions; an absent parent relationship must not be treated as proof of independence. Require separate transaction or corporate evidence before asserting equity ownership, beneficial ownership or sponsor control.
• Trademarks: the USPTO records assignments that meet minimum formal requirements, without verifying title ^[34]. A recorded trademark owner is evidence of a brand owner, not of the plant operator.
• Federal registration: the General Services Administration (GSA) states that the Unique Entity ID (UEI) is the authoritative identifier for those doing business with the federal government ^[35]. The public entity API exposes names, UEIs and registration details ^[36]. The agency stopped using the DUNS number for identifying entities in April 2022 ^[37], which is one reason FSIS's optional DUNS field can be stale.
• Company-provided or authorized FDA registration records: the "Parent Company Name" field identifies the company that owns the registered facility, if different from the facility name ^[38]. FFR information and its parent-company fields are not a public lookup source ^[30]. For this protocol, use only lawfully supplied company records or authorized access and retain the provenance; the field guide explains a field, not public access.

FinCEN states that companies created in the United States are no longer reporting companies ^[12], and in August 2026 announced it would delete previously reported information by US persons ^[39].

Table 1 compares lookup routes. The practical reading is that the FSIS file and API supply the plant layer from public data, and every route that claims a company layer depends on the evidence ladder in the next table. The commercial directory differs from the official file in both count and update claim, and neither difference is explained on the page, so its figure serves only as a pointer for reconciliation. The last row is shown for completeness because the publisher offers sourced target research; it adds process, not new public data.

Table 2 sets out the statuses. The key design choice is that "Group unverified" and "Conflict" are first-class states, not failures to be hidden. A grouping that rests only on a shared name is a hypothesis. Conflicts are retained with both source URLs and dates, because a later document, such as a new state filing after a transaction, may resolve them.

Step 6: Crosswalk CSV template

Use one row per plant cluster, with the following columns. The sample row is fictional and labeled as such (Hypothetical Example).

fsis_number,source_edition,plant_name,raw_address,activity,size_class,proposed_legal_entity,parent_evidence_url_date,match_rule,confidence,exclusion_reason
M0000+P0000,2026-10-19,Example Foods Plant A,"1 Example Rd, Town T, ST 00000",Meat Processing; Poultry Processing,Small,Example Foods LLC (ST),"https://example.invalid/registry 2026-10-20",R2 registry name+state+address,Verified,

Columns worth explaining:

• fsis_number and source_edition: the raw composite string and the edition date of the pull.
• plant_name and raw_address: values as on the grant record, never the cleaned versions.
• size_class: a descriptor only, never a financial metric.
• parent_evidence_url_date: the document URL and date, so a reviewer can re-check it.
• match_rule and confidence: the rule code that produced the match and one of the evidence statuses in Table 2.
• exclusion_reason: populated when the plant is out of the universe.

Step 7: Weekly change log

Because each edition replaces the last, history exists only if the analyst saves it. Compare each new edition with the prior one by establishment_id and log six event types:

• Additions: new establishment_id values, routed to the lookup queue.
• Removals: missing values, matched against the inactive list.
• Name and dbas changes: possible rebrand or operator change.
• Activity changes: possible scope drift against the thesis.
• Size-class changes: descriptor updates only.
• Grant-date changes: possible ownership event.

A removal may mean inactivity, and FSIS publishes an inactive list for establishments inactive three years or less, so match removals against that list before treating a plant as gone. A grant-date change may indicate an ownership event, since a change of ownership requires a new application ^[6]. It is a trigger for review, not a conclusion. Route every uncertain match, every conflict and every changed operator to a human review queue.

Data Analysis and Evidence

This section reports measurements from the October 19, 2026 numeric CSV, which we parsed ourselves, alongside official figures. Our tallies are descriptive counts of the file as retrieved on October 20, 2026. They are not FSIS statistics.

Table 3 places four different counts side by side: a legal-research center's figure of 6,413, the FSIS statement, our file tally and a commercial figure. The file tally of 7,255 is consistent with the FSIS page’s lower bound of "more than 7,100" ^[1]; the Law Center and commercial totals require separate comparison of dates and scope. The Law Center reports 6,413, the FSIS statement says "more than 7,100", our tally of rows is 7,255, and the commercial directory reports 7,592. The differences may reflect timing, inclusion of inactive or voluntary records, or different counting rules, but none of the sources explains the gap, so an analyst should cite the edition and counting rule, not the headline number. The size-class counts show a second timing issue: our file's classified rows (3,000 + 2,923 + 514 = 6,437) exceed the December 2025 FSIS figures (2,961 + 2,847 + 508 = 6,316), consistent with different dates and a different base. Of the 7,255 rows, 818 have an "N / A" or blank size class, so size filtering silently drops about 11 percent of plant records unless handled explicitly.

The dbas field is useful but noisy. Of 2,404 non-empty values, placeholders such as "n/a" and "none" account for 117, and many values are multi-name strings separated by semicolons. Another 101 distinct dbas values appear on more than one row, together covering 390 rows. Those shared values are candidate company groupings and also a hazard if they record past as well as current names.

Market-structure context

Concentration data describe firms, not plants, and illustrate why the plant-to-firm step matters. USDA's Economic Research Service (ERS) reported that the four largest firms handled 85 percent of steer and heifer purchases and 67 percent of hog purchases in the data it analyzed ^[40], compared with 36 percent for the four largest beef packers in 1980 ^[40]. A related ERS report put national broiler four-firm concentration at 53 percent ^[41]. NASS reports that 786 plants slaughtered hogs in 2025 ^[11]. Hundreds of plant records can roll up to a handful of companies in the largest segments, while in other segments each plant may be its own company. Plant counts alone cannot reveal which segment is which.

Reporting rules for counts

Four reporting rules keep the numbers honest:

• Separate quantities: report plants and companies as separate counts.
• Reconcile before counting: start with establishment_id records, consistent with the documented row unit ^[7]. For this protocol, review potential duplicates and merge only records documented to represent the same physical operation. Assign an evidence status to each resulting plant unit, then count separately verified company groupings.
• Unresolved rate: unresolved plant clusters divided by all plant clusters.
• No implied scale: neither number implies market share or addressable revenue, and no denominator beyond the clusters in the defined universe should be introduced.

(Hypothetical Example) Suppose a thesis-filtered pull yields 100 establishment_id records, using the documented establishment-level row unit ^[7]. Shared names, addresses and DUNS values generate review candidates. Assume documented review establishes that 18 records duplicate physical operations already represented, leaving 82 reconciled plant units. Of those, 52 reach Verified status and map to 31 distinct legal entities; 30 remain unresolved. The mapped-company count is 31, the reconciled plant count is 82 and the unresolved rate is 30 divided by 82, or 36.6 percent. These assumptions are illustrative only and do not describe any real universe.

Implications and Future Directions

For origination teams, the main implication is that the FSIS directory is best treated as a plant sensor, refreshed weekly, feeding a slower and more evidence-intensive entity layer. The weekly cadence favors automated diffing; the entity layer favors documented human judgment. Ownership events can be detected as candidate signals (a changed name, dbas value or grant date), but a signal at the plant level says nothing about whether a sale process exists. Any outreach decision belongs after entity verification, and the compliance and approval steps for outreach sit outside the scope of this report.

Deal activity in the wider sector is increasingly shaped by buyers that are themselves sponsor-owned. Kroll's spring 2026 industry insights describe food and beverage M&A as predominantly driven by strategic buyers, including companies primarily owned by private equity investors ^[42]. That report covers food and beverage broadly, not meat, so it is context and not a market-size input for this universe. Its relevance is indirect: where a platform buys several plants, the plant records of a target may later roll up under a new parent, and the crosswalk needs a mechanism to represent a change of parent without losing history.

Three developments are worth monitoring. First, FSIS's 2026 notice on establishment size definitions suggests the classes may change, so any dataset keyed on size class should store the definition version. Second, the beneficial ownership landscape has shifted: FinCEN's August 2026 final rule ends reporting for US companies ^[39], which changes reporting coverage. BOI access was already restricted before that change: FinCEN states that general business or commercial use is not authorized ^[43]. BOI is not a public origination lookup source, and authorized recipients remain subject to purpose and disclosure limits ^[44]. Use public state registries, filings and LEI data for ordinary research. Third, Agriculture and Agri-Food Canada links the USDA directory from its poultry and egg market-information page ^[45], which suggests continuing use by researchers outside the United States who need a plant-level view of North American protein supply.

Limitations: no Census County Business Patterns table for the meat manufacturing industries was opened, so no firm-versus-establishment count for the industry is reported. The NAICS (North American Industry Classification System) definitions reviewed come from the 2012 Economic Census reference pages, which describe animal slaughtering other than poultry ^[46] and poultry slaughtering and processing ^[47]; confirm current titles before mapping codes. An earlier Congressional Research Service product referred to 6,900 federally regulated establishments ^[48], a figure that predates the October 2026 page.

Frequently Asked Questions (FAQs)

How to do an FSIS establishment number lookup

Download the numeric CSV from the FSIS MPI Directory page or query the API, then filter on establishment_number. Search the full composite string, since one row can hold several numbers joined by a plus sign, and watch for suffix letters.

Does the establishment number identify the parent company?

No. It identifies the plant that processed the product under inspection ^[4]. The brand on the label may belong to a distributor named with a qualifying phrase ^[5], and the grant belongs to the applicant operating the plant ^[6]. For a meat and poultry plant owner lookup, treat the grant holder as the operator lead, and note that parentage needs separate evidence from registries, filings, LEI data or trademark records.

How often is the directory updated?

FSIS says weekly, with each current edition replacing earlier editions. The Data.gov catalog record lists a monthly period and a January 2025 modified date ^[13], which conflicts with the FSIS page. Use the FSIS edition date.

Can one company have many establishment numbers?

Yes. Multiple grants at one plant appear as one composite number, and one company can operate several plants. The reverse also occurs: tenants and landlords can share a site ^[6]. Cluster plants first, then attach clusters to entities.

Does HACCP size show how big a target is?

No. The classes ignore corporate affiliation ^[8] and track employees, not volume.

How to map food processing companies for private equity

Define the thesis, pull the FSIS file, cluster plants, resolve operators using the evidence ladder, and report plant clusters, mapped companies and the unresolved rate separately. For broader food manufacturing, FDA-regulated plants need other lists.

Is the Protein Intel directory an official source?

No. It is a commercial directory that describes its data as updated daily and shows a count of 7,592 facilities. Use the FSIS file as the system of record and treat the commercial count as an unverified comparison point.

How to identify acquisition targets in meat and poultry processing

To identify food manufacturing acquisition targets or build a food processing market map for private equity, a plant record can place a site in a defined universe. Whether the company behind it is a suitable target depends on entity resolution, ownership and scale evidence beyond size class. Treat the directory as an input to screening, not a ranking.

Conclusion

The FSIS MPI Directory is a reliable, official, weekly-refreshed plant layer for US meat, poultry and egg products. It is not a company file. The establishment number identifies a producing plant under a grant of inspection, the grant belongs to an applicant, and labels can show a different company. The October 19, 2026 edition, as we retrieved it, contained 7,255 plant records, a count consistent with the FSIS page’s lower bound of "more than 7,100" ^[1]. The commercial directory’s 7,592 is a separate self-reported total requiring reconciliation of date and scope.

A defensible target universe therefore depends on discipline rather than on the file alone. Keep plant count and company count separate. Record the edition date and the schema for every pull. Use HACCP size only as a descriptor. Mark every grouping with an evidence status, retain conflicts, and compute the unresolved rate as unresolved plant clusters over all plant clusters. Finally, check grant and activity records before applying scope exclusions, and remember that a plant record never proves a sale process. FDA jurisdiction alone is not an exclusion rule: some animals under FDA slaughter jurisdiction receive voluntary FSIS inspection ^[27]. Teams that need verification of a particular site or operator should rely on primary documents and consult qualified legal counsel for questions of ownership and control.