[{"data":1,"prerenderedAt":64},["ShallowReactive",2],{"blog-pcba-first-pass-yield-en":3},{"title":4,"description":5,"date":6,"lastUpdated":6,"category":7,"image":8,"readingTime":9,"wordCount":10,"timeRequired":11,"htmlContent":12,"faq":13,"tags":29,"slug":33,"sourceLocale":34,"availableLocales":35,"jsonld":40},"PCBA First-Pass Yield: How to Calculate and Use It","Calculate first-attempt PCBA station yield using clear unit counts, retest rules and traceable results, and distinguish later passes from the initial outcome.","2026-10-27","testing",null,13,2407,"PT13M","\u003Cp>\u003Cstrong>First-attempt station test yield\u003C/strong> is the percentage of unique serialized boards that pass a specified test gate on their first completed attempt:\u003C/p>\n\u003Cp>\u003Cstrong>First-attempt station yield = (unique boards passing their first completed attempt / unique boards with a completed first attempt) x 100\u003C/strong>\u003C/p>\n\u003Cp>A board that fails and later passes a retest remains a first-attempt failure. So does a board that passes after repair. Later outcomes are useful, but they answer a different question.\u003C/p>\n\u003Cp>This metric is narrower than a process-wide first-pass yield. It describes one gate, such as in-circuit test or functional test, under a stated counting rule. For a broader improvement approach, see the \u003Ca href=\"/en/blog/yield-improvement-roadmap/\">yield improvement roadmap\u003C/a>. This guide focuses on defining, calculating, and using the metric without losing the original test history.\u003C/p>\n\u003Ch2>Define the counting boundary before calculating\u003C/h2>\n\u003Cp>Terms such as FPY, first-time yield, and retest yield are not used consistently. One organization may count a board as first-pass if it passes a retest before repair; another may count only the first station result. Neither label alone tells a reader which convention was used.\u003C/p>\n\u003Cp>For the metric in this article, specify:\u003C/p>\n\u003Cul>\n\u003Cli>\u003Cstrong>Unit:\u003C/strong> One uniquely identified board, not one test run, panel, defect, or serial-number scan.\u003C/li>\n\u003Cli>\u003Cstrong>Gate:\u003C/strong> The named station and test program or acceptance criteria being measured.\u003C/li>\n\u003Cli>\u003Cstrong>Attempt:\u003C/strong> A completed execution of the defined test sequence, with its recorded outcome.\u003C/li>\n\u003Cli>\u003Cstrong>Numerator:\u003C/strong> Boards that pass on their first completed attempt at that gate.\u003C/li>\n\u003Cli>\u003Cstrong>Denominator:\u003C/strong> Boards whose first attempt at that gate was completed during the selected cohort window.\u003C/li>\n\u003Cli>\u003Cstrong>Reporting snapshot:\u003C/strong> The date and time used to assess any later retest, repair, or unresolved status.\u003C/li>\n\u003C/ul>\n\u003Cp>A completed failure stays in the denominator even if the board is later repaired or its failure is found to be test-related. The metric records the first station outcome; it does not establish the cause of every failure.\u003C/p>\n\u003Cp>If testing occurs at panel level but the reported unit is a board, define how the panel result is assigned to individual board serials. Do not divide panel counts by board counts without a reliable mapping.\u003C/p>\n\u003Ch3>Decide how to handle exceptions\u003C/h3>\n\u003Cp>Publish exception rules with the yield. Otherwise, two reports with the same formula can produce different results.\u003C/p>\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Situation\u003C/th>\n\u003Cth>Treatment for this metric\u003C/th>\n\u003C/tr>\n\u003C/thead>\n\u003Ctbody>\u003Ctr>\n\u003Ctd>First completed attempt passes\u003C/td>\n\u003Ctd>Count the board once in the numerator and denominator.\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>First completed attempt fails\u003C/td>\n\u003Ctd>Count the board once in the denominator, even if a later test passes.\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>Retest passes without repair\u003C/td>\n\u003Ctd>Record as a later outcome; do not change the first result.\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>Pass after repair or rework\u003C/td>\n\u003Ctd>Record the action and later pass separately; do not count as a first-attempt pass.\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>Test aborts before completion\u003C/td>\n\u003Ctd>Preserve the abort record. Under a first-\u003Cem>completed\u003C/em>-attempt rule, the board enters the denominator when a qualifying attempt completes.\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>Completed test ends in error or inconclusive status\u003C/td>\n\u003Ctd>Define whether the station classifies this as a completed non-pass. If it does, include it in the denominator and identify it separately from product failures.\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>Board is on hold\u003C/td>\n\u003Ctd>If its first attempt was completed, retain that result. If no attempt was completed, report the board as pending outside the denominator.\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd>Invalid record or excluded unit\u003C/td>\n\u003Ctd>Apply a documented rule and retain the evidence, reason, and count. Do not remove unfavorable outcomes merely because they lower yield.\u003C/td>\n\u003C/tr>\n\u003C/tbody>\u003C/table>\n\u003Cp>Examples of possible exclusions include engineering samples or an invalid duplicate data record, if the parties agreed on those rules in advance. An exclusion is not the same as a board that failed and subsequently received another test.\u003C/p>\n\u003Cp>The distinction between an abort and a completed error matters. A station restart after an incomplete abort may produce the first \u003Cem>completed\u003C/em> attempt, but the earlier abort should remain visible. If a project instead wants to measure the first test \u003Cstrong>start\u003C/strong>, it should name and calculate that metric separately.\u003C/p>\n\u003Ch2>Worked example: 200 unique boards\u003C/h2>\n\u003Cp>Assume a fixed cohort of 200 unique serialized boards completes its first attempt at one test gate. At the reporting snapshot:\u003C/p>\n\u003Cul>\n\u003Cli>180 passed on the first attempt.\u003C/li>\n\u003Cli>20 failed on the first attempt.\u003C/li>\n\u003Cli>Of those 20 failures, 12 later passed retest without repair.\u003C/li>\n\u003Cli>Five later passed after repair.\u003C/li>\n\u003Cli>Three remained unresolved.\u003C/li>\n\u003C/ul>\n\u003Cp>The first-attempt station yield is:\u003C/p>\n\u003Cp>\u003Cstrong>180 / 200 x 100 = 90%\u003C/strong>\u003C/p>\n\u003Cp>The observed eventual test-pass percentage, using the same 200-board cohort and the stated snapshot, is:\u003C/p>\n\u003Cp>\u003Cstrong>(180 + 12 + 5) / 200 x 100 = 197 / 200 x 100 = 98.5%\u003C/strong>\u003C/p>\n\u003Cp>These figures are both useful, but they are not interchangeable. The \u003Cstrong>90%\u003C/strong> describes the first completed station result. The \u003Cstrong>98.5%\u003C/strong> includes later passes, including boards that received repair.\u003C/p>\n\u003Cp>Among the 20 initial failures, 12 passed retest without repair:\u003C/p>\n\u003Cp>\u003Cstrong>12 / 20 x 100 = 60%\u003C/strong>\u003C/p>\n\u003Cp>That does \u003Cstrong>not\u003C/strong> establish that 60% were false rejects. An intermittent board fault, connector seating, contact condition, test setup, handling, or another change could be involved. The count says what happened next, not why it happened.\u003C/p>\n\u003Cp>The three unresolved boards also matter. They remain in the 200-board first-attempt denominator, and they are not eventual passes at this snapshot. A later report may show a different eventual-pass percentage as cases close, while the original first-attempt results remain unchanged.\u003C/p>\n\u003Cp>Neither percentage is automatically a shipment yield, a field-quality measure, or a reliability prediction. Those require their own populations, criteria, and observation periods.\u003C/p>\n\u003Ch2>Separate boards, attempts, and defects\u003C/h2>\n\u003Cp>One board can generate several test attempts and several defect records. These are different units of measurement.\u003C/p>\n\u003Cp>For example, if one serialized board fails its first test, fails two retests, and then passes, the first-attempt calculation includes \u003Cstrong>one board and one first-attempt failure\u003C/strong>. An attempt-level report may show four test runs. If the investigation identifies two defects on the board, a defect report may show two defects.\u003C/p>\n\u003Cp>Mixing these counts creates misleading ratios. A numerator based on passing test runs divided by a denominator based on unique boards is not a valid first-attempt yield.\u003C/p>\n\u003Cp>A practical report can show all three views:\u003C/p>\n\u003Col>\n\u003Cli>\u003Cstrong>Unit view:\u003C/strong> First result and later disposition for each serial.\u003C/li>\n\u003Cli>\u003Cstrong>Attempt view:\u003C/strong> Every test execution, including aborts and retests.\u003C/li>\n\u003Cli>\u003Cstrong>Defect and action view:\u003C/strong> Diagnoses, no-trouble-found findings, rework, repair, and verification.\u003C/li>\n\u003C/ol>\n\u003Cp>“No trouble found” (NTF) should be recorded as a finding with its investigation scope, not used automatically as proof of a false first-test rejection. A fault may be intermittent, or the investigation may not reproduce the original conditions.\u003C/p>\n\u003Cp>Terminology for corrective work also needs agreement. In a PCBA report, use the applicable project definitions for rework and repair, define the action codes used and link each action to the affected serial and subsequent verification; do not hide both under an unexplained “retest” label.\u003C/p>\n\u003Ch2>Interpret station yield alongside the test method\u003C/h2>\n\u003Cp>A yield number is only as interpretable as its gate. An \u003Ca href=\"/en/blog/ict-testing/\">in-circuit test\u003C/a> and a \u003Ca href=\"/en/blog/pcb-functional-testing/\">functional test\u003C/a> exercise a board differently. Their results should not be compared as if they represent identical coverage or identical production populations.\u003C/p>\n\u003Cp>Record enough context to make a comparison meaningful:\u003C/p>\n\u003Cul>\n\u003Cli>Product and board revision.\u003C/li>\n\u003Cli>Gate name, station identifier, and test sequence.\u003C/li>\n\u003Cli>Test program or recipe revision and acceptance limits.\u003C/li>\n\u003Cli>Fixture or interface identification where relevant.\u003C/li>\n\u003Cli>Cohort window and reporting snapshot.\u003C/li>\n\u003Cli>Outcome categories, including errors, aborts, and holds.\u003C/li>\n\u003C/ul>\n\u003Cp>A change in yield may reflect a production change, but it can also coincide with a revised test program, fixture condition, product mix, or classification rule. Compare like with like before assigning a cause. If the test criteria change within a reporting period, show the results by version as well as any agreed aggregate.\u003C/p>\n\u003Cp>Inspection data can help investigation, but an inspection finding is not automatically a failed electrical test. Use the \u003Ca href=\"/en/blog/pcb-inspection/\">PCB inspection overview\u003C/a> and the guide to a \u003Ca href=\"/en/blog/pcb-assembly-inspection-report/\">PCB assembly inspection report\u003C/a> to decide which observations should be linked to the test record.\u003C/p>\n\u003Ch2>From stage yield to a rolled result\u003C/h2>\n\u003Cp>A station result is not automatically the yield for the whole assembly route. A board may pass one gate and fail another, or it may have been retested or repaired before reaching the gate being reported.\u003C/p>\n\u003Cp>Multiplying stage yields is valid only when those yields are \u003Cstrong>consistently defined, sequential, conditional rates for the appropriate cohort\u003C/strong>. For example, assume:\u003C/p>\n\u003Col>\n\u003Cli>Of 200 boards, 180 pass the first gate on their first attempt: \u003Cstrong>180/200 = 90%\u003C/strong>.\u003C/li>\n\u003Cli>Those same 180 first-pass boards proceed to the second gate without intervening retest or repair.\u003C/li>\n\u003Cli>Of those 180 boards, 171 pass the second gate on their first attempt: \u003Cstrong>171/180 = 95%\u003C/strong>.\u003C/li>\n\u003C/ol>\n\u003Cp>The proportion passing both gates under those conditions is:\u003C/p>\n\u003Cp>\u003Cstrong>(180 / 200) x (171 / 180) = 171 / 200 = 85.5%\u003C/strong>\u003C/p>\n\u003Cp>The second rate is conditional on membership in the first-pass group. No blanket assumption that the gates are independent is needed.\u003C/p>\n\u003Cp>By contrast, multiplying two separately reported station percentages can be wrong if they use different products, time windows, exclusions, or populations—including boards that reached the second gate after failing and recovering at the first.\u003C/p>\n\u003Cp>When in doubt, use the serial-level route history. Define the starting cohort and count the boards that pass \u003Cstrong>every required gate on the first completed attempt, without a completed-test retest or repair\u003C/strong>, under the agreed route boundary. Specify separately how earlier aborted starts are treated in the route metric. Show boards still awaiting a downstream gate separately; do not silently treat an incomplete route as a completed pass.\u003C/p>\n\u003Ch2>Protect the first result\u003C/h2>\n\u003Cp>A first-attempt metric depends on an immutable history of what happened first. A current “pass” status alone cannot distinguish a first-pass board from one that required several attempts and corrective work.\u003C/p>\n\u003Cp>For each board, retain or agree how to retrieve:\u003C/p>\n\u003Cul>\n\u003Cli>Unique board serial and relevant panel-to-board mapping.\u003C/li>\n\u003Cli>Gate, station, program revision, and test timestamps.\u003C/li>\n\u003Cli>The first start, any aborts, and the first completed outcome.\u003C/li>\n\u003Cli>Subsequent attempts and their outcomes, in order.\u003C/li>\n\u003Cli>Failure codes or measurements needed for investigation.\u003C/li>\n\u003Cli>Holds, dispositions, NTF findings, and corrective actions.\u003C/li>\n\u003Cli>Links between an action and the verification test that followed it.\u003C/li>\n\u003Cli>Authorized corrections to data, with the original record and reason preserved.\u003C/li>\n\u003C/ul>\n\u003Cp>A correction may be necessary for a genuinely misidentified serial or invalid record. It should be traceable rather than overwriting an unfavorable first result. Agree who can authorize corrections, how exclusions appear in the report, and how long records must be retained.\u003C/p>\n\u003Cp>This is a data and responsibility requirement, not an assumption that a supplier has a particular dashboard or automated traceability system. The method of collecting and exchanging the records should be agreed for the project.\u003C/p>\n\u003Ch2>Use the metric to investigate, not just score\u003C/h2>\n\u003Cp>A useful review starts with the first-attempt failures and asks where they concentrate. Group them by meaningful, controlled categories: test step, failure code, board revision, program version, station, fixture, or time period. Then inspect the associated serial histories.\u003C/p>\n\u003Cp>For retest passes, compare the conditions before and after the first failure. Was there an adjustment, handling event, fixture change, or different setup? For passes after repair, link the corrective action and verification result. For unresolved boards, identify the current hold reason and owner.\u003C/p>\n\u003Cp>A test issue and an assembly defect can require different actions. Avoid changing acceptance limits or routinely repeating tests merely to improve the reported percentage. Any test or process change should follow the project’s agreed review and validation process, with results compared under clearly identified versions.\u003C/p>\n\u003Cp>The count behind a percentage also matters. A percentage from a small cohort can change substantially when only a few board outcomes change. Report the numerator, denominator, and time window rather than a percentage alone.\u003C/p>\n\u003Ch2>What to put in the release package or RFQ\u003C/h2>\n\u003Cp>Engineers and procurement should agree on the measurement before production data is reported. A useful release or request for quotation includes:\u003C/p>\n\u003Cul>\n\u003Cli>\u003Cstrong>Product scope:\u003C/strong> Part numbers, board revisions, serialization method, and included production lots.\u003C/li>\n\u003Cli>\u003Cstrong>Test scope:\u003C/strong> Named gate, required program and acceptance criteria, and who provides and controls them.\u003C/li>\n\u003Cli>\u003Cstrong>Counting rule:\u003C/strong> First completed attempt, unique-unit denominator, pass definition, and cohort dates.\u003C/li>\n\u003Cli>\u003Cstrong>Exception rules:\u003C/strong> Aborts, completed errors, invalid data, holds, engineering units, and exclusions.\u003C/li>\n\u003Cli>\u003Cstrong>Later outcomes:\u003C/strong> Retest-without-repair, rework, repair, NTF, unresolved status, and snapshot cutoff.\u003C/li>\n\u003Cli>\u003Cstrong>Route definition:\u003C/strong> Any required multi-gate calculation and its starting cohort.\u003C/li>\n\u003Cli>\u003Cstrong>Data fields and format:\u003C/strong> Serial-level history, timestamps, action codes, summaries, and delivery method.\u003C/li>\n\u003Cli>\u003Cstrong>Responsibility:\u003C/strong> Who operates the test, reviews failures, authorizes dispositions and record corrections, and maintains each part of the traceability record.\u003C/li>\n\u003Cli>\u003Cstrong>Change control:\u003C/strong> How product, test-program, fixture, or acceptance-criteria revisions are identified and approved.\u003C/li>\n\u003Cli>\u003Cstrong>Commercial agreement:\u003C/strong> Required reporting frequency, retention period, confidentiality, and any separately agreed test or investigation work.\u003C/li>\n\u003C/ul>\n\u003Cp>The customer typically defines the product requirements and approves the applicable acceptance criteria; the assigned test operator records and executes the agreed process. The actual division can differ, so put it in the project agreement rather than assuming either party owns every investigation or disposition.\u003C/p>\n\u003Cp>If the project includes assembly, discuss the measurement and data handoff alongside the \u003Ca href=\"/en/products/smt-assembly/\">SMT assembly service\u003C/a>. Submit the applicable files and reporting requirements through \u003Ca href=\"/en/quote/\">Quote\u003C/a>, or use \u003Ca href=\"/en/contact/\">Contact\u003C/a> to clarify the intended test boundary and responsibilities. Any testing, records, or reports should be confirmed in the project agreement.\u003C/p>\n\u003Cdiv data-component=\"BlogQuickQuoteInline\">\u003C/div>\n\u003C!-- faq:start -->\n\u003Ch2>Frequently asked questions\u003C/h2>\n\u003Ch3>Is first-attempt station yield the same as process FPY?\u003C/h3>\n\u003Cp>Not necessarily. It measures one defined station. A process-wide result needs a defined route, cohort, and treatment of retest and repair across all required gates.\u003C/p>\n\u003Ch3>Does a retest pass count as a first pass?\u003C/h3>\n\u003Cp>Not under the definition used here. It is a later pass, even if no repair occurred. Report it separately.\u003C/p>\n\u003Ch3>Are test errors excluded from the denominator?\u003C/h3>\n\u003Cp>Not automatically. Define whether an error is a completed non-pass or an invalid attempt, preserve its record, and report exclusions with their reasons.\u003C/p>\n\u003Ch3>Does a retest pass prove the first failure was false?\u003C/h3>\n\u003Cp>No. It establishes that a later attempt passed. Investigation is needed to determine whether the original result involved the board, test setup, an intermittent condition, or another cause.\u003C/p>\n\u003Ch3>Can I multiply yields from two test stations?\u003C/h3>\n\u003Cp>Only if they are compatible sequential conditional yields for the appropriate cohort. Otherwise, count serialized boards that pass every required gate under the agreed first-pass rules.\u003C/p>\n\u003C!-- faq:end -->\n\u003Csection class=\"related-links\" aria-label=\"Related\">\u003Ch3>Related links\u003C/h3>\u003Cul>\u003Cli>\u003Ca href=\"/en/blog/yield-improvement-roadmap/\">Yield improvement roadmap: Controlling PCB manufacturing quality from raw materials to final test\u003C/a>\u003C/li>\u003Cli>\u003Ca href=\"/en/blog/ict-testing/\">In-Circuit Testing (ICT) for PCBA: Coverage and Fixture Design\u003C/a>\u003C/li>\u003Cli>\u003Ca href=\"/en/blog/pcb-functional-testing/\">PCB Functional Testing: Coverage, Fixtures, and Production Release\u003C/a>\u003C/li>\u003Cli>\u003Ca href=\"/en/blog/pcb-inspection/\">PCB Inspection Services: AOI, X-Ray, and Quality Control Solutions\u003C/a>\u003C/li>\u003Cli>\u003Ca href=\"/en/blog/pcb-assembly-inspection-report/\">PCB Assembly Inspection Report Format: Template and Checklist\u003C/a>\u003C/li>\u003Cli>\u003Ca href=\"/en/products/smt-assembly/\">SMT assembly service\u003C/a>\u003C/li>\u003Cli>\u003Ca href=\"/en/quote/\">Quote\u003C/a>\u003C/li>\u003C/ul>\u003C/section>",[14,17,20,23,26],{"question":15,"answer":16,"answerText":16},"Is first-attempt station yield the same as process FPY?","Not necessarily. It measures one defined station. A process-wide result needs a defined route, cohort, and treatment of retest and repair across all required gates.",{"question":18,"answer":19,"answerText":19},"Does a retest pass count as a first pass?","Not under the definition used here. It is a later pass, even if no repair occurred. Report it separately.",{"question":21,"answer":22,"answerText":22},"Are test errors excluded from the denominator?","Not automatically. Define whether an error is a completed non-pass or an invalid attempt, preserve its record, and report exclusions with their reasons.",{"question":24,"answer":25,"answerText":25},"Does a retest pass prove the first failure was false?","No. It establishes that a later attempt passed. Investigation is needed to determine whether the original result involved the board, test setup, an intermittent condition, or another cause.",{"question":27,"answer":28,"answerText":28},"Can I multiply yields from two test stations?","Only if they are compatible sequential conditional yields for the appropriate cohort. Otherwise, count serialized boards that pass every required gate under the agreed first-pass rules.",[30,31,32],"first pass yield","PCBA test","manufacturing yield","pcba-first-pass-yield","en",[34,36,37,38,39],"de","cn","fr","ru",{"blog":41,"breadcrumb":50},{"@context":42,"@type":43,"headline":4,"description":5,"url":44,"datePublished":6,"dateModified":6,"timeRequired":11,"keywords":45,"articleSection":7,"author":46,"publisher":49},"https://schema.org","BlogPosting","https://hilpcb.com/en/blog/pcba-first-pass-yield/","first pass yield, PCBA test, manufacturing yield",{"@type":47,"name":48},"Organization","HILPCB",{"@type":47,"name":48},{"@context":42,"@type":51,"itemListElement":52},"BreadcrumbList",[53,58,62],{"@type":54,"position":55,"name":56,"item":57},"ListItem",1,"Home","https://hilpcb.com/",{"@type":54,"position":59,"name":60,"item":61},2,"Blog","https://hilpcb.com/en/blog/",{"@type":54,"position":63,"name":33,"item":44},3,1794062104859]