A counting machine accuracy test during Factory Acceptance Testing (FAT) should show that the equipment repeatedly delivers the programmed quantity under defined, representative conditions. One successful bottle is not enough. A useful protocol controls the product, challenges the intended speed range, records every container result, and applies acceptance criteria approved before testing begins.
This guide gives pharmaceutical production, quality and engineering teams a practical structure for a tablet or capsule counting machine FAT. It can be adapted to a semi-automatic counter, a multi-channel electronic counter or an automatic pill counting machine. The final protocol must still follow the project URS, quality procedures, risk assessment and agreed contractual requirements.
What Should a Counting Machine Accuracy Test Verify?
The test should verify both accuracy-whether the observed quantity matches the programmed target-and repeatability-whether that result remains stable across repeated runs and intended operating conditions.
| Test item | What it demonstrates | Evidence to retain |
|---|---|---|
| Individual container count | Each bottle receives the target quantity within the approved tolerance. | Target and verified actual count for every bottle. |
| Repeatability | The machine produces consistent results over repeated runs. | Run number, container number and pass/fail result. |
| Speed challenge | Counting remains controlled across the intended operating range. | Actual speed setting and results at each condition. |
| Product handling | Feeding, singulation and discharge do not create bridges, overlaps, damage or jams. | Observations, alarms, rejects and stoppages. |
| Controls and interlocks | No-bottle/no-count, alarm, reject or stop functions behave as specified. | Challenge result and system response. |
| Data integrity | Results are attributable, legible, contemporaneous, original and accurate. | Signed records, audit trail or electronic report where applicable. |

Automatic pill counting equipment suitable for a structured FAT accuracy and repeatability challenge.
Freeze the Test Conditions Before FAT Starts
Ambiguity creates disputes. Approve the method and blank recording forms before anyone sees the results. At minimum, identify:
- Machine model, serial number and software version
- Product name, batch/sample ID and physical characteristics
- Tablet or capsule size, shape, coating and dust level
- Target count per container
- Bottle size, material and neck diameter
- Low, nominal and high test speeds
- Feeder, vibration and sensor settings
- Warm-up and cleaning status
- Reference count method and calibrated equipment
- Number of runs and bottles per run
- Individual and run-level acceptance rules
- Retest and deviation procedure
How to Choose the FAT Sample Size
There is no universal number of bottles that is correct for every counting machine. Sample size should be justified by risk, target count, product variability, machine technology, speed range, available material, customer requirements and the confidence needed from the demonstration.
A practical risk-based structure
- Define the test unit. Usually one filled container is one independent result. Record every container rather than only a pooled total.
- Define the conditions. Use at least nominal production speed. Add low and high qualified speeds when speed can affect feeding or detection.
- Set bottles per condition and repeated runs. Distribute observations across conditions so that a single long run does not mask start-up or changeover effects.
- State the rationale. Link the chosen design to project risk and the approved acceptance rule.
Optional zero-failure demonstration
When the protocol uses an attribute rule such as "no miscounted containers," the familiar zero-failure approximation below can help frame sample size. It does not replace a company sampling standard or process capability study.
| Demonstration objective | Minimum zero-failure sample* | Interpretation |
|---|---|---|
| 90% reliability at 90% confidence | 22 containers | Useful for a lower-risk preliminary demonstration. |
| 95% reliability at 95% confidence | 59 containers | A commonly understood statistical benchmark. |
| 99% reliability at 95% confidence | 299 containers | Stronger evidence, but often material-intensive for FAT. |
*Rounded up. These examples assume independent trials, constant performance and zero observed failures. They do not prove long-term process capability or replace qualification at the final site.

Representative counting system used to challenge accuracy across defined operating speeds.
Step-by-Step Counting Machine FAT Test Method
- Verify readiness. Confirm installation, utilities, guarding, cleaning, calibration status, approved drawings and applicable pre-FAT checks. Record the machine configuration.
- Condition and identify the sample. Confirm product quantity and physical condition. Record any substitute material and its justification.
- Set the target count and baseline parameters. Record the recipe, target quantity, feeder/vibration setting, sensor setting and container format.
- Establish the reference count. Verify each filled container with an approved independent method-for example, a controlled manual count by trained personnel or a qualified reference system. Avoid using the machine under test as its own reference.
- Run the low-speed condition. After the defined stabilization period, collect the approved number of consecutive containers. Record rejects, alarms, stops and observations.
- Run the nominal-speed condition. Change only the planned speed-related setting unless the protocol specifically permits other adjustments. Repeat the approved sample plan.
- Run the high qualified speed. Challenge the upper intended condition without exceeding the approved operating range. Repeat the sample plan.
- Calculate and review results. Evaluate every container, each condition and the complete test. Do not average away a failing individual result.
- Close deviations. Document the symptom, investigation, correction and authorized retest. Preserve original data.
Counting Error Calculation and Result Reporting
Agree on the calculation before execution. For each container:
Count Error (%) = [(Actual Count − Target Count) ÷ Target Count] × 100
If the target is 100 tablets and the verified actual count is 99, the result is −1 tablet and −1.00%. If the actual count is 101, the result is +1 tablet and +1.00%.
Report undercounts and overcounts separately. A total of 500 tablets across five bottles does not prove that all five bottles contained 100; results of 99, 100, 100, 100 and 101 would incorrectly appear perfect if only the pooled total were considered.
| Run | Speed | Bottle | Target | Actual | Error (units) | Error (%) | Result |
|---|---|---|---|---|---|---|---|
| 1 | Nominal: ___ | 001 | 100 | 100 | 0 | 0.00% | Pass |
| 1 | Nominal: ___ | 002 | 100 | ___ | ___ | ___ | ___ |
| 1 | Nominal: ___ | 003 | 100 | ___ | ___ | ___ | ___ |
Define Acceptance Criteria Before Testing
The FAT protocol should state which criteria apply to individual containers, each speed condition and the overall test. The actual numerical limits must come from the approved URS, supplier/customer agreement, quality system and risk assessment-not from a generic article.
| Criterion | Protocol wording to complete before FAT |
|---|---|
| Individual count | Each tested container shall contain ___ units / remain within ___ units or ___% of target. |
| Undercount rule | Undercounts are [not permitted / permitted within the approved limit of ___]. |
| Overcount rule | Overcounts are [not permitted / permitted within the approved limit of ___]. |
| Repeatability | All ___ runs at each condition shall meet the individual and run-level criteria. |
| Speed range | The machine shall pass at ___, ___ and ___ [bottles/min or units/min]. |
| Operational events | No unexplained jam, uncontrolled discharge or critical alarm; specified reject/interlock challenges shall pass. |
| Record completeness | All raw data, calculations, observations, deviations and approvals shall be complete and traceable. |
How to Investigate a Failed Counting Accuracy Test
A failed result identifies a symptom, not automatically a defective sensor. Hold the affected test, preserve the original data and investigate one variable at a time.
| Observed symptom | Likely factors | Confirmation check | Potential response |
|---|---|---|---|
| Counts vary between repeat runs | Unstable hopper level, bridging, inconsistent vibration or mixed product condition | Repeat under unchanged settings while observing product singulation and hopper level | Stabilize feed, define refill level, optimize approved feeder settings |
| Accuracy declines at higher speed | Insufficient separation, sensor saturation, discharge overlap or timing mismatch | Compare low, nominal and high-speed video/results | Adjust within the qualified range; reassess maximum approved speed |
| Transparent or reflective product miscounts | Optical contrast, orientation, reflection or overlapping units | Inspect sensor/vision signal using representative samples | Optimize lighting/sensitivity or evaluate vision counting technology |
| Intermittent undercounts | Dust, blocked channel, static adhesion, rebound or bottle timing | Inspect the detection and discharge path; verify rejected units and bottle position | Clean, control static/dust, correct separation or timing |
| Drift after extended operation | Material buildup, component heating, loosening or wear | Compare beginning, middle and end-of-run results | Define cleaning/inspection frequency; service affected components |
| Reference recount disagrees | Manual verification error or unqualified reference method | Use an independent second verification under controlled conditions | Correct the measurement system before judging the machine |
For a deeper troubleshooting reference, review factors that affect capsule counting machine accuracy and these calibration tips for high-speed vision counters.

Inspection and configuration checks should be recorded alongside counting results.
What to Include in the FAT Report
The final package should allow a reviewer who was not present to reconstruct what was tested, how it was verified and why it passed or failed.
- Approved protocol and revision
- Machine ID and configuration
- Product/sample identification
- Container and target count
- Pre-test readiness checks
- Reference method and calibration status
- All operating parameters
- Individual raw count results
- Signed calculation and review
- Alarms, rejects and observations
- Deviation and investigation records
- Approved retest results
- Open-item/punch-list status
- Final acceptance signatures
Common FAT Protocol Mistakes to Avoid
- Testing only one easy condition: it gives no evidence about performance at the normal or upper intended speed.
- Reporting only an average: it can hide individual undercounts and overcounts.
- Changing several variables together: it prevents a meaningful root-cause conclusion.
- Using an undefined or self-referencing count method: the reference measurement must be independent and controlled.
- Retesting without a documented cause: repeated attempts until a pass create biased evidence.
- Ignoring the sample difference: a smooth placebo may not represent a dusty, static or transparent commercial product.
- Claiming FAT proves long-term capability: extended performance under site conditions requires the appropriate qualification or process study.
FAQs About Counting Machine Accuracy Testing
What is the purpose of a counting machine accuracy test during FAT?
It verifies that the configured machine repeatedly produces the required count under defined factory conditions before shipment. The test should evaluate individual container accuracy, repeatability, intended speed conditions, operational events and record completeness.
How many samples are required for a counting machine FAT?
No universal sample size applies to every project. Select and justify the number using product and process risk, required confidence, machine technology, speed conditions, available sample and approved customer/quality requirements. A zero-failure statistical calculation may support the rationale when its assumptions fit.
Should the counter be tested at different speeds?
Yes, when speed variation is part of intended use or can affect product separation and detection. Test representative low, nominal and high qualified conditions while controlling other variables.
Can average count be used for acceptance?
Average results may be reported as supplementary information, but they should not conceal a failing bottle. Individual-container criteria are normally needed because an undercount and overcount can cancel in a pooled total.
What should happen after a failed result?
Preserve the original data, open the required deviation record, identify the symptom, investigate material/method/machine/environment factors, document any correction, and perform only an authorized retest under predefined conditions.
Does passing FAT complete validation?
No. FAT is factory evidence under the stated test conditions. SAT, qualification and process validation may still be required according to the validation plan and quality system.
Prepare a Better Counting Machine FAT
Share the product dimensions and characteristics, target count, bottle specification, required output and draft acceptance criteria. SED Pharma can review the application and discuss a suitable counting equipment configuration and FAT test approach.
Request a FAT Test Plan
