In This Article
- Executive Summary
- Where Unit Conflicts Come From
- What GMP Rules Expect of Units
- Conversions That Are Not Arithmetic
- Building a Single Unit Master
- Finding the Conflicts You Already Have
- Interface Design So Units Travel With Values
- Who Owns the Unit Master Under Change Control
- A First Pass in Four Steps
- Conclusion
- For Further Reading
- References & Sources
Executive Summary
In most pharma and biotech companies, the same material is measured in different units by different systems. ERP buys and stocks it in kilograms or drums. MES dispenses it in grams, often adjusted for potency. LIMS reports its assay as a percentage, its potency in international units per milligram, and its bioburden in CFU per gram. Each system is right on its own terms. The errors happen at the boundaries, when a number crosses from one system to another and its unit, basis, or decimal convention does not cross with it.
The fix is not a new system. Unit of measure standardization in pharma comes down to three things: one controlled unit master with explicit conversion rules, interfaces that refuse to send a number without its coded unit, and a named owner who manages that master under change control. UCUM, the Unified Code for Units of Measure, gives you a ready reference vocabulary, and it also tells you which conversions should never be automated.
This article covers where cross-system unit conflicts come from, what GMP rules say about units and data exchange, the conversions that are not simple arithmetic, how to structure a unit master and its conversion layer, how to find the conflicts already in your data with a profiling pass, how to design interfaces so units travel with values, and who should own the result.
Where Unit Conflicts Come From
Unit conflicts are rarely the result of carelessness. They come from the fact that ERP, MES, and LIMS were each designed around a different question, and each one picked the unit that answers its own question best. Understanding that is the first step, because it tells you that the goal is not to force every system onto one unit. The goal is to make every translation between units explicit, controlled, and checked.
Each System Measures for a Different Job
ERP answers commercial and logistics questions. How much do we hold, how much do we owe the supplier, what is it worth, and when do we reorder? Its base unit for a material is usually the unit it is bought and valued in: kilograms, liters, drums, bags, or each. Alternative units (a drum equals 25 kg, a case equals 12 bottles) are stored as conversion factors on the material record. Some ERPs also support potency units, where a quantity is expressed as the amount of active content rather than the physical amount. Oracle’s JD Edwards documentation, for example, explains that the potency stored on the lot record is used for conversion, and warns that companies storing inventory in potent units must know the physical size of that inventory.1
MES answers execution questions. How much of this lot do I weigh into this batch, on which scale, and did the operator get it right? Its working unit is whatever the scale and the master batch record use, often grams or milligrams. For actives and some excipients, the target weight changes from lot to lot because it is adjusted for assay, water content, or potency.
LIMS answers quality questions. Does this lot meet its specification? Its units are the units of the test method: percent of label claim, mg per gram, ppm, IU per mg, CFU per gram, EU per mL. Many of these are not quantities of material at all. They are concentrations, ratios, or assay-defined activities.
Put those three side by side and the conflict is built in. A single lot of an active ingredient can be received as 50 kg in ERP, released by LIMS with an assay of 98.7 percent on the anhydrous basis and a water content of 1.2 percent, and dispensed in MES as a potency-adjusted target of some number of grams. Every one of those numbers is correct. The question is whether every system that uses them knows exactly what they mean.
Five Types of Cross-System Unit Conflict
Most unit problems between LIMS, ERP, and MES fall into one of five types. Naming them helps, because each one needs a different control.
| Conflict type | What it looks like | Typical consequence |
|---|---|---|
| Scale | The same dimension at a different prefix: mg against g, µg against mg, mL against L. | Errors of exactly 1,000 times. Inventory that looks a thousand times too large, or a dispense target that cannot be met. |
| Dimension | Units that measure different things: IU against mg, units of activity against mass, CFU against grams. | A conversion that has no general answer is applied anyway, with a factor taken from the wrong substance or the wrong reference standard. |
| Presentation | Amount per container against amount per volume or per dose: per vial against per mL, per tablet against per gram. | Quantities that differ by the fill volume or tablet weight, often found only at reconciliation or labeling. |
| Basis | The same unit symbol with a different meaning: as-is against anhydrous, free base against salt, percent w/w against percent of label claim. | Potency adjustments that double-count or miss a correction, and results that compare cleanly with the wrong limit. |
| Notation | Different text for the same unit, or different decimal conventions: mcg, ug, and µg; IU and U; 1,250 meaning one thousand two hundred fifty in one locale and one and a quarter in another. | Mapping failures, unreadable values, or values parsed to the wrong number with no error raised. |
The first three are mostly visible if someone looks. The last two are the dangerous ones, because the number and the unit symbol can look correct while the meaning is wrong.
Notation Problems Are Data Problems, Not Typing Problems
Notation sounds trivial until you look at what the systems store. In UCUM, the micro prefix is written as a lowercase “u” in the case-sensitive form, so micrograms is “ug,” and UCUM states that its case-insensitive symbols are not compatible with its case-sensitive ones.2 Meanwhile, clinical safety guidance pushes in a different direction for human readers. ISMP’s list of error-prone abbreviations records that the symbol µg has been misread as mg and recommends writing mcg instead.3 A company can easily end up with mcg in its labeling and batch record templates, ug in a LIMS that follows a coded vocabulary, and µg in an instrument export. All three mean the same thing. A naive interface treats them as three different units, or worse, maps one of them to mg.
Decimal conventions add a second layer. The General Conference on Weights and Measures resolved in 2003 that the decimal marker “shall be either the point on the line or the comma on the line,” and reaffirmed that in grouped digits “neither dots nor commas are ever inserted in the spaces between groups.”4 In practice, many systems do insert commas or dots as thousands separators, and they do it according to the local settings of the machine. Microsoft’s own support documentation notes that Excel uses the separators set in the operating system’s regional settings unless a user overrides them.5 ISMP adds a note that commas are used to separate digits in the United States, while in some other countries commas are used in place of decimal points.3 When a global company moves data through spreadsheets, CSV extracts, or file-based interfaces between sites, a value of 1,250 can mean two very different quantities depending on which computer wrote it and which one read it.
Why These Conflicts Survive Validation
If unit conflicts are this common, why do validated systems still carry them? There are three reasons we see repeatedly.
- Each system is validated alone. The LIMS validation proves LIMS calculates correctly in LIMS units. The MES validation proves MES weighs correctly in MES units. The interface between them is often tested with a few happy-path messages for a few materials, all of which use the same units on both sides.
- Test data is too clean. Test materials are created fresh for validation, with tidy units. The legacy material master, with fifteen years of free-text unit entries, is never in the test set.
- Units are configured, not coded. Unit tables and conversion factors are configuration data. They change after go-live through master data requests, often outside the change control path that governs the system itself.
Our earlier article on LIMS master data under change control covers how units, rounding, and specifications drift inside LIMS. This article looks at the problem one level up: what happens between systems, and how to build the conversion layer that connects them.
What GMP Rules Expect of Units
No GMP regulation tells you to use UCUM, or any other unit vocabulary. What the regulations do require is an accurate statement of quantities, a defined calculation when quantities vary, and checks on data that moves into, out of, and between computerized systems. Read together, those requirements describe a unit master and a checked conversion layer, even though they never use those words.
Master Records Must State Units Accurately and Consistently
Under 21 CFR 211.186(b)(4), master production and control records must include “An accurate statement of the weight or measure of each component, using the same weight system (metric, avoirdupois, or apothecary) for each component.”6 The rule was written for paper records, but its intent applies directly to systems: one recipe, one measurement system, no mixing.
The charge-in rules in 21 CFR 211.101 add two points that matter here. First, the batch “shall be formulated with the intent to provide not less than 100 percent of the labeled or established amount of active ingredient.” That is the regulatory reason potency-adjusted dispensing exists. Second, the second-person check on each dispensed container must confirm that “The weight or measure is correct as stated in the batch production records.”7 If the batch record and the scale disagree about the unit, that check cannot be done correctly.
For active pharmaceutical ingredients, ICH Q7 is more direct. Section 6.41 says master production instructions should include “An accurate statement of the quantity or ratio of each raw material or intermediate to be used, including the unit of measure,” and that where the quantity is not fixed, the calculation for each batch size or rate of production should be included.8 Section 6.60 says laboratory control records should include a record of all calculations performed in connection with the test, “including, for example, units of measure, conversion factors, and equivalency factors.”8 Conversion factors are named explicitly as part of the record.
Data Moving Between Systems Must Be Checked
21 CFR 211.68(b) requires that “Input to and output from the computer or related system of formulas or other records or data shall be checked for accuracy,” with the degree and frequency of checking based on the complexity and reliability of the system.9 A conversion factor applied inside an interface is exactly the kind of output that needs that check.
EU GMP Annex 11 says the same thing in system-to-system terms. Clause 5 states that “Computerised systems exchanging data electronically with other systems should include appropriate built-in checks for the correct and secure entry and processing of data, in order to minimize the risks.” Clause 4.8, on migration, asks validation to include “checks that data are not altered in value and/or meaning during this migration process.”10 The phrase “value and/or meaning” is the important one for units. A quantity of 500 that arrives as 500 but is now read as milligrams instead of grams has kept its value and lost its meaning.
PIC/S guidance on data management and integrity (PI 041-1) goes further. Its section on data transfer says “Interfaces should be assessed and addressed during validation to ensure the correct and complete transfer of data,” and it lists, among the risks to check, that data may be inadvertently lost, amended, or transcribed incorrectly during transfer. It also suggests that a data flow map can help in understanding the risks of interfaced systems.11 That data flow map is the starting point for the profiling pass described later in this article.
What the rules add up to. Regulators do not prescribe a unit vocabulary. They expect that every quantity is stated with its unit, that any conversion or calculation is defined and recorded, and that data crossing a system boundary is checked so that its meaning, not only its digits, arrives intact. A controlled unit master with documented conversion rules is the most direct way to show all three.
Conversions That Are Not Arithmetic
Most unit conversions are fixed and universal. A gram is 1,000 milligrams everywhere, for every material, forever. Those conversions are safe to automate, and a unit master should hold them once and apply them everywhere. The trouble comes from conversions that look like arithmetic but depend on the substance, the reference standard, the lot, or the container. Those are where most serious cross-system errors start, and they need a different kind of rule.
International Units Depend on the Substance and the Standard
The international unit is not a unit of mass. For biological medicines, WHO establishes international biological reference standards through collaborative studies, and where appropriate a value is assigned, for example a potency in International Units. For other reference standards, no unitage is assigned at all.14 The IU for one substance has nothing to do with the IU for another.
Vitamins show how far apart the factors can be, even for substances most people think of as simple. The NIH Office of Dietary Supplements gives 1 mcg of vitamin D as equal to 40 IU.12 For vitamin E, the same office gives 1 IU of the natural form as equal to 0.67 mg of alpha-tocopherol, and 1 IU of the synthetic form as equal to 0.45 mg.13 So even within one vitamin, the conversion depends on which chemical form is in the drum. For biologics, enzymes, and heparin-type products, the relationship between activity and mass is established by assay against a reference standard and may change from lot to lot.
UCUM treats this correctly. It classes the international unit as an arbitrary unit, and defines arbitrary units as “units whose meaning entirely depends on the measurement procedure (assay).” It goes on to state that since version 1.7 of the code, “it is no longer possible to convert or compare arbitrary units with any other arbitrary unit.”2 UCUM also keeps the international unit ([iU]) separate from the enzyme unit (U, defined as a micromole per minute) and from the USP unit ([USP’U]).2 Many plant systems use “U” and “IU” as if they were the same thing. In a coded vocabulary they are not.
The practical rule follows from this. A unit master should never hold a general IU to mg factor. If a conversion between activity and mass is needed, it belongs to the material (and often to the lot), it comes from an approved source such as a specification, a certificate of analysis, or a LIMS result, and it is recorded as data with its source, not as a unit setting.
Per-Vial, Per-mL, and Per-Dose Are Different Quantities
Parenteral products carry at least two strengths: the amount in the container and the concentration. FDA’s 2022 final guidance on container labels and carton labeling explains that overdoses have occurred when a user saw the concentration (such as 10 mg/mL) but missed the net quantity in the container, and recommends that for small-volume injection products the quantity per total volume be the primary expression, followed by the quantity per mL in parentheses, for example “500 mg/10 mL (50 mg/mL).”15 The guidance is nonbinding, but the problem it describes is the same one that appears in systems.
Inside a company, the per-vial amount usually lives in ERP and labeling (it drives the product description and the inventory unit), while the per-mL concentration lives in LIMS (it is what the assay measures) and in MES (it drives the fill and compounding calculation). The two are linked by the fill volume, which is itself master data owned by yet another group. If the fill volume changes, or a new presentation is added, three systems need to agree on a new relationship. A per-vial value sent to a field that expects a per-mL value is wrong by a factor equal to the fill volume, which can be small enough to look plausible.
The same problem appears in solid doses (mg per tablet against mg per gram of blend) and in combination products where the dose is per actuation or per device. In each case the rule is the same: the presentation is part of the unit. “mg” alone is not enough. The unit master needs to distinguish “mg per vial,” “mg per mL,” and “mg per tablet” as different expressions, and the conversion between them must name the presentation attribute it depends on.
Potency and Basis Change the Meaning of a Kilogram
A kilogram of active ingredient with an assay of 97 percent on the as-is basis is not the same quantity of active as a kilogram at 99 percent. Potency-adjusted dispensing exists to meet the 100 percent formulation requirement in 211.101(a), and it is one of the most common places for unit and basis errors to enter.7
The sequence typically looks like this. LIMS reports the assay, and often water content or residual solvent, on a stated basis. That result is passed to ERP, which stores a potency on the lot. MES reads the potency (from LIMS directly, or from ERP) and calculates a target weight. Each step is a chance for the basis to be lost. If LIMS reports assay on the anhydrous basis and MES treats it as the as-is basis, the correction for water is missed. If both the LIMS result and the MES formula correct for water, it is counted twice. Neither error is visible in the number itself.
The JD Edwards documentation is useful here because it shows how an ERP handles this at the inventory level. It notes that commitments can be brought out of balance when the primary unit of measure for an item is not a potent unit but commitments exist in a potent unit.1 Whatever ERP you run, the point is the same: a potency value is a conversion factor, and it needs the same control as any other conversion factor. That means a stated basis, a stated source, and a clear statement of which system applies it.
Decimal and Locale Conversions Happen Without Anyone Deciding
The last category is the one nobody designs. When a value is written to text, in a CSV extract, a flat file, a PDF certificate, or a spreadsheet, the software writing it applies the local decimal and thousands conventions. When it is read back, the software reading it applies its own. If the two settings differ, the value changes. No conversion rule was configured and no one approved it, yet the number in the receiving system is different from the one that was sent.
This is why the interface rules later in this article insist on a machine format for numbers (a decimal string with a point as the decimal marker and no grouping separators) and never a display format. Display formatting belongs at the screen and the printed report, not in the data passing between systems.
Building a Single Unit Master
A unit master is a controlled list of every unit your GxP systems are allowed to use, with one agreed code for each, a defined meaning, and a set of approved conversion rules between them. It is small compared with a material master. Its value comes from being the one place where units are defined, so that every system’s local unit codes can be mapped to it, and every conversion can be traced back to it.
Use UCUM as the Reference Vocabulary
There is no reason to invent a code system. UCUM was built for unambiguous electronic exchange of quantities, it covers SI and customary units, it handles prefixes in a predictable way, and, as shown above, it already distinguishes arbitrary units that must not be converted.2 It is also the unit vocabulary preferred in HL7 FHIR, which matters if your quality data ever feeds clinical, safety, or regulatory systems.16
Two other standards are worth knowing. ISO 11240:2012, one of the ISO IDMP standards for identifying medicinal products, sets rules for the coded representation of units of measurement when exchanging information about quantitative product characteristics such as strength. It also provides structures for mapping between different unit vocabularies, and its scope is limited to data interchange between computer applications. ISO lists the 2012 edition as last confirmed in 2020.17 In the EU, EMA’s Referentials Management Service publishes controlled lists, including units of measurement, that comply with ISO 11240 and are used across the European regulatory network.18 If you submit IDMP data to EMA, your unit master will need a mapping to those lists anyway, so it makes sense to build that mapping in from the start.
Using UCUM as the reference does not mean every screen has to show UCUM codes. UCUM is the internal code. Each unit record can also carry a display label (such as mcg for printed batch records, in line with ISMP advice) and the local code each system uses. The display label is for people. The UCUM code is for systems.
What a Unit Record Holds
| Field | Purpose | Example |
|---|---|---|
| Reference code | The one agreed code, from UCUM where UCUM has one. | ug |
| Display label | What people see on screens, records, and labels. | mcg |
| Dimension | What the unit measures. Conversions are only allowed within a dimension. | Mass |
| Arbitrary flag | Marks assay-defined units that must never be converted by a general rule. | Yes for [iU]; No for mg |
| Presentation or basis qualifier | Records whether the unit is per container, per volume, per dose, or on a stated basis. | per vial; anhydrous basis |
| Local codes by system | The code each system stores for this unit, so mappings are explicit. | ERP: UG; LIMS: µg; MES: mcg |
| Status and effective date | Whether the unit may be used in new records, and since when. | Active from a set date |
| Change record reference | The change control that created or last changed the entry. | Change number |
The local codes in the example are illustrative. The point is that the mapping from each system’s own code to the reference code is written down in one place and approved, rather than assumed by each interface developer.
Four Kinds of Conversion Rule
The most important design decision in a unit master is to separate conversion rules by where their factor comes from. We use four categories.
Universal Conversions
Same factor for every material and every lot: g to mg, L to mL, kg to g. Held once in the unit master, automated everywhere, tested once and then regression-tested on change.
Material-Specific Conversions
Factor depends on the material or presentation: drum to kg for a given pack, mg per tablet to mg per gram for a given tablet weight, vial to mL for a given fill. Held on the material or product record, with an approved source.
Lot-Specific Conversions
Factor depends on the lot: potency, assay, water content, activity in IU per mg. Comes from an approved LIMS result or certificate of analysis. Treated as transactional quality data with a stated basis, not as unit configuration.
Conversions That Must Not Exist
Any general rule between dimensions, such as IU to mg, CFU to grams, or one substance’s unit to another’s. Listed explicitly so that no system or interface can create one by configuration.
The prohibited category is the one most companies skip, and it is the one that prevents the worst errors. If your unit master says plainly that there is no general conversion from [iU] to mg, any request to configure one has to come through change control with a material-specific justification.
Decide Where Each Conversion Is Done
Every conversion should be done in exactly one place. That sounds obvious, yet in practice the same potency correction is often applied in LIMS (as a calculated result), again in an ERP batch characteristic, and again in the MES dispense formula. The rule we recommend is:
- The system that owns the data sends it in its own unit, coded against the unit master. LIMS sends assay as LIMS reports it, with the basis stated.
- The system that uses the data converts it, using a rule from the unit master, and records the conversion in its own record. MES converts potency into a target weight and records the factor used.
- The interface in between does not convert anything. It checks the unit code against the unit master, checks that the receiving field accepts that unit, and rejects the message if not.
This keeps conversions visible in a GMP record, where 211.68(b) and ICH Q7 6.60 expect them to be, rather than hidden in middleware mapping logic that QA never reviews.98 Some companies prefer a central conversion service that every system calls. That can work well, as long as the service is validated, its rules come from the unit master, and the calling system records the result and the rule version it used.
Finding the Conflicts You Already Have
A unit master prevents new conflicts. It does not tell you how many are already in your data. For that you need a profiling pass: a structured, read-only look at the unit fields and paired values across your systems. It does not need special tools. A database query client, extracts from each system, and a spreadsheet or notebook are enough for a first pass.
Map the Unit-Bearing Fields
Start from the data flow map PIC/S recommends for interfaced systems.11 For each interface between ERP, MES, and LIMS, list every field that carries a quantity, the field that carries its unit (if any), and whether the unit is stored per record or assumed from configuration. Fields with an assumed unit are your first risk list.
Extract the Distinct Unit Strings
For each system, pull every distinct value stored in every unit field, with a count of records using it. Expect variants of the same unit (mg, MG, Mg, mg. with a period), free text, blanks, and a few values nobody can explain.
Map Every String to the Reference Code
Map each distinct string to a UCUM code, or mark it unmapped. Record the decision and who made it. Unmapped strings and strings that map to more than one meaning (U used for both enzyme units and international units, for example) go to the subject matter experts for a ruling.
Compare Paired Values Across Systems
For the same material, lot, or batch, compare the quantity each system holds after converting both to the reference unit. Received quantity in ERP against the sum of dispensed quantities in MES. Potency in ERP against the released LIMS result. Fill volume in ERP against the MES recipe value. Each mismatch is a candidate conflict.
Classify, Assess, and Decide
Classify each conflict by the five types above, assess whether it has affected any GMP decision (a dispense, a release, a stock count, a label), and decide on correction, deviation, or no action with a documented reason.
Signals That Point to a Unit Conflict
Paired comparisons generate a lot of mismatches, most of them for ordinary reasons such as timing, open transactions, or yield losses. These patterns are the ones that point to a unit problem rather than a process one.
| Signal in the data | Likely cause | What to check next |
|---|---|---|
| Ratio between paired values close to 1,000, 1,000,000, or 0.001 | Scale conflict (prefix) | The unit fields on both records and the interface mapping |
| Ratio close to a fill volume, tablet weight, or pack size | Presentation conflict | Whether one system holds per container and the other per volume or per dose |
| Ratio close to the lot’s potency, assay, or (1 minus water content) | Basis conflict or double correction | The basis stated on the LIMS result and the formula in MES or ERP |
| Ratio of 100 or 1,000 even though both records show the same unit, or a value with a different number of decimal places after transfer | Decimal or locale conflict | The file format, the locale settings of sending and receiving systems |
| Activity values with a mass unit, or mass values with an activity unit | Dimension conflict | Whether a general IU to mass rule exists anywhere in configuration |
| Blank or free-text unit on a GMP quantity field | Unit assumed, not stored | How the receiving system decides which unit to use |
Scope the First Pass Carefully
A full profiling pass across every material is a large piece of work. A first pass does not need to be complete. Pick the materials where a unit error would do the most harm: actives and potency-adjusted materials, biologics measured in activity units, controlled substances (where inventory reconciliation is legally required), and any product with more than one presentation. Then pick the interfaces those materials pass through. That usually gives a list of a few dozen materials and a handful of interfaces, which is enough to show whether the problem is isolated or systemic.
Before you correct anything. A unit conflict found in historical data may have affected released batches, stock records, or reported results. Correcting the master data without first assessing impact can erase the evidence of what happened. Treat confirmed conflicts in GMP data as potential deviations, assess impact on past decisions first, and correct under change control.
Interface Design So Units Travel With Values
Most cross-system unit errors happen because a number was sent without its unit, and the receiving system assumed one. The interface design rule that prevents this is simple to state: a quantity is a value and a coded unit together, and neither is ever sent without the other.
Do Not Trust the Message Standard to Enforce It
Common message standards allow units, but they do not always require them. In the current MESA B2MML schema for ISA-95 messages, the quantity structure contains a quantity string that must be present and a unit of measure element that is optional. The unit itself is a code, with optional attributes that can name the code list it comes from.19 A message can therefore be fully valid against the schema and still carry a bare number. Your interface specification has to make the unit mandatory, and has to name the code list, because the standard leaves both to you.
HL7 FHIR, used for clinical and increasingly for laboratory data, takes a similar approach with more guidance. Its Quantity data type carries a value, a human-readable unit, a system that identifies the code system, and a coded unit. The specification says that if the unit can be coded in UCUM and a code is provided, “it SHOULD be a UCUM code,” and that if a UCUM unit is provided in the code, “a canonical value can be generated for purposes of comparison between quantities.” It also warns that the human-readable unit text “cannot be assumed” to contain a valid UCUM unit.16 That last point applies to every system: the display text is not the unit. The code is.
The Message Rules
For any interface carrying GMP quantities between LIMS, ERP, and MES, we recommend the following as minimum requirements in the interface specification:
- Value as a decimal string in machine format. Point as the decimal marker, no thousands separators, no locale formatting. Avoid converting to binary floating point in transit, since that can change the last digits of a decimal value.
- Unit as a code from the unit master. With the code system named (for example, UCUM). Free text is not accepted.
- Basis and presentation where they apply. As-is or anhydrous, per container or per volume, free base or salt. Sent as coded qualifiers, not buried in a comment.
- Source and version for any lot-specific factor. A potency value carries the LIMS result or certificate it came from.
- No default units. A missing unit is an error, never an invitation for the receiving system to assume one.
Reject, Do Not Guess
The receiving side matters as much as the sending side. When a message arrives, the receiving system or the interface layer should check that the unit code exists in the unit master, that its dimension matches what the target field expects, and that any needed conversion rule exists and is not prohibited. If any check fails, the message should be held for review, not loaded with a guessed unit, and not converted with a fallback factor.
This is the practical meaning of Annex 11’s built-in checks for “the correct and secure entry and processing of data.”10 It is also what PIC/S has in mind when it asks for interfaces to be assessed during validation to ensure correct and complete transfer.11 A held message is visible and fixable. A guessed unit is invisible until it reaches a batch, a stock count, or a certificate.
An interface specification is not a control. NASA lost the Mars Climate Orbiter in 1999 because ground software produced thruster data in pound-seconds where the software interface specification required newton-seconds. The investigation board found that the data was required to be in metric units by existing interface documentation, and that the trajectory modelers assumed it was.20 The requirement was written down. Nothing checked it. A pharma interface that specifies grams in a document but accepts a bare number has the same weakness.
Test the Interface With Units, Not Just Messages
Interface validation usually proves that messages flow. To prove that units flow, add test cases built around the five conflict types:
- Every unit the interface is allowed to carry, at least once, including the unusual ones.
- A message with a missing unit, an unknown unit, and a unit of the wrong dimension, each of which must be rejected.
- Values that would change under a different locale, such as 1,250 and 1.250, sent from a system with each regional setting your sites use.
- An activity unit sent to a mass field, which must be rejected.
- A per-container value and a per-volume value for the same product, each going to the correct field.
- Potency values on each basis your LIMS reports, with the MES calculation checked by hand against an independent calculation.
- Very large and very small values, to check that no digits are lost or added in transit.
Keep these tests as a regression suite. Run them again whenever the unit master, an interface mapping, or a system version changes.
Who Owns the Unit Master Under Change Control
A unit master without an owner becomes another uncontrolled table within a year. Ownership has to be explicit, and it has to cross the IT, quality, and operations boundary, because no single function can judge every change on its own.
Roles
| Role | Owns | Typical home |
|---|---|---|
| Unit master owner | The list itself: entries, codes, conversion rules, mappings, and the review cycle. Assesses every change request first. | Data governance or master data team, with a named individual |
| Quality approver | Approval of any change that affects GMP calculations, specifications, or records. Decides whether impact on past data needs investigation. | QA |
| Subject matter experts | Technical correctness of conversions: analytical scientists for assay and activity units, process engineers for dispensing and fill, supply chain for pack units. | QC, manufacturing science, supply chain |
| System owners | The local unit tables and mappings in each system, and making sure they match the unit master. | ERP, MES, and LIMS system owners |
| Regulatory affairs | Confirming that strength and quantity expressions match what is filed, and flagging when a unit change touches a submission. | Regulatory CMC |
This model is different from LIMS master data ownership, where QA approves specification content and a LIMS administrator configures it. The unit master spans all three systems, so its owner has to have authority across them. In smaller companies that is often the person who already owns material master data in ERP, with QA as approver. What matters is that the role is named, written into a procedure, and given time to do the work.
What Counts as a Change
The following should all go through change control, because each one can change the meaning of GMP data:
- Adding, retiring, or redefining a unit entry.
- Adding or changing a fixed conversion rule.
- Adding or changing a material-specific factor, such as a pack size, fill volume, or tablet weight used in conversion.
- Changing how a local system code maps to the reference code.
- Changing the system or step where a conversion is done.
- Moving a conversion out of the prohibited category for a specific material.
Lot-specific factors such as potency and assay are different. They are quality data produced under the laboratory’s own controls, and they flow through the interfaces as results with a basis and a source. They do not need a change request for each lot. The rule that says how they are used does.
Impact Assessment Questions
When a change request comes in, the unit master owner and QA should be able to answer these questions before approval:
- Which systems hold this unit or rule, and which interfaces carry it?
- Which materials, products, specifications, and recipes use it?
- Does the change alter how any existing record is read, or only new records?
- Do historical records need to stay readable under the old definition? (They almost always do. Retire entries rather than redefine them.)
- Does it touch a filed strength, label, or specification?
- Which regression tests from the interface suite need to run?
Under 211.68(b), changes to master production and control records in computer systems must be made only by authorized personnel.9 A unit master that drives those records falls within that expectation. Access to change it, in the master and in each system, should be limited to the roles above and visible in each system’s audit trail.
Periodic Review
Even with good change control, drift creeps in through new materials, system upgrades, and site integrations. Run a short version of the profiling pass on a regular cycle: distinct unit strings by system, unmapped values, and paired-value mismatches for high-risk materials. Report the results with the system periodic reviews, so the evidence that units are under control is in the place an inspector would look for it.
A First Pass in Four Steps
For a company that has never looked at units across systems, here is a practical way to start without committing to a large program.
Pick the Riskiest Twenty
Select around twenty materials and products where a unit error would matter most: potency-adjusted actives, activity-unit biologics, controlled substances, and multi-presentation products. List the interfaces they pass through.
Profile and Compare
Extract unit strings and paired values for those materials from ERP, MES, and LIMS. Map strings to UCUM, compare paired values, and classify every mismatch by conflict type.
Draft the Unit Master From What You Found
Build the first version of the unit master from the units in actual use, mapped to UCUM, with the four rule categories filled in. Include the prohibited conversions you found or nearly found.
Name the Owner and Fix the Worst Interface
Write the ownership and change control procedure, name the owner, and choose the one interface with the most conflicts. Rework its specification so units are mandatory and coded, add reject-not-guess checks, and run the unit test suite. Use it as the pattern for the rest.
What good looks like after the first pass. You can show an inspector a controlled list of units with UCUM codes, a documented mapping from every system’s local codes, an explicit list of conversions that are not allowed, a named owner, and at least one GMP interface that rejects a quantity without a valid unit. You also have evidence, from the profiling pass, of how many conflicts existed and what was done about each one.
Conclusion
Units of measure are easy to overlook because every system handles its own units well. The risk is in the space between systems, where a number moves and its meaning is supposed to move with it. Scale errors, activity units treated as mass, per-vial values read as per-mL, potency applied twice, and decimal commas read as thousands separators all come from the same root cause: a quantity that crossed a boundary without its full definition. The controls are not exotic. One unit master with UCUM codes and four kinds of conversion rule. Interfaces that treat value and unit as a single item and reject anything incomplete. A profiling pass to find what is already wrong. A named owner and a change control path that covers the whole set. Together, those give you a clear answer to the questions that GMP rules already ask about accurate statements of quantity, defined calculations, and checked data transfer.
Sakara Digital works with pharma and biotech organizations on master data, system integration, and data quality across LIMS, ERP, and MES. If you are planning an integration, a system migration, or a site consolidation and want an independent view of where your unit conflicts are likely to be, we are happy to have that conversation.
For Further Reading
For Further Reading
- LIMS Master Data Under Change Control: Keeping Specifications, Methods, and Units in Step
- Integration Patterns Between LIMS and MES That Avoid Two Versions of Truth
- Data Quality Regression Testing After System Migrations: A Practical Test Design
- Reference Data Governance in Pharma: UNII, MedDRA, SNOMED, and Version Drift
- Supplier Data Quality: Getting Clean Certificates of Analysis Into Your Systems
References & Sources
- Oracle. “Conversion of Units of Measure for Potent Units.” JD Edwards EnterpriseOne Applications 9.2 documentation. https://docs.oracle.com/en/applications/jd-edwards/supply-chain-manufacturing/9.2/eoash/conversion-of-units-of-measure-for-potent-units.html
- Regenstrief Institute and the UCUM Organization. “The Unified Code for Units of Measure” (specification, including Section 3.2 on arbitrary units). https://ucum.org/ucum
- Institute for Safe Medication Practices. “ISMP List of Error-Prone Abbreviations, Symbols, and Dose Designations.” 2024. https://online.ecri.org/hubfs/ISMP/Resources/ISMP_ErrorProneAbbreviation_List.pdf
- Bureau International des Poids et Mesures. “Resolution 10 of the 22nd CGPM (2003): Symbol for the Decimal Marker.” https://www.bipm.org/en/-/resolution-cgpm-22-10
- Microsoft Support. “Change the Character Used to Separate Thousands or Decimals.” https://support.microsoft.com/en-us/office/change-the-character-used-to-separate-thousands-or-decimals-c093b545-71cb-4903-b205-aebb9837bd1e
- U.S. Code of Federal Regulations. 21 CFR 211.186, Master Production and Control Records. eCFR. https://www.ecfr.gov/current/title-21/chapter-I/subchapter-C/part-211/section-211.186
- U.S. Code of Federal Regulations. 21 CFR 211.101, Charge-in of Components. eCFR. https://www.ecfr.gov/current/title-21/chapter-I/subchapter-C/part-211/section-211.101
- International Council for Harmonisation. “ICH Q7: Good Manufacturing Practice Guide for Active Pharmaceutical Ingredients,” Sections 6.41 and 6.60. https://database.ich.org/sites/default/files/Q7%20Guideline.pdf
- U.S. Code of Federal Regulations. 21 CFR 211.68, Automatic, Mechanical, and Electronic Equipment. eCFR. https://www.ecfr.gov/current/title-21/chapter-I/subchapter-C/part-211/section-211.68
- European Commission. “EudraLex Volume 4, Annex 11: Computerised Systems.” 2011. https://health.ec.europa.eu/system/files/2016-11/annex11_01-2011_en_0.pdf
- PIC/S. “PI 041-1: Good Practices for Data Management and Integrity in Regulated GMP/GDP Environments,” Sections 9.1 and 9.4. July 1, 2021. https://picscheme.org/docview/4234
- National Institutes of Health, Office of Dietary Supplements. “Vitamin D: Fact Sheet for Health Professionals.” https://ods.od.nih.gov/factsheets/VitaminD-HealthProfessional/
- National Institutes of Health, Office of Dietary Supplements. “Vitamin E: Fact Sheet for Health Professionals.” https://ods.od.nih.gov/factsheets/VitaminE-HealthProfessional/
- World Health Organization. “Providing International Biological Reference Standards.” https://www.who.int/activities/providing-international-biological-reference-standards
- U.S. Food and Drug Administration. “Safety Considerations for Container Labels and Carton Labeling Design to Minimize Medication Errors: Guidance for Industry.” May 2022. https://www.fda.gov/media/158522/download
- HL7 International. “FHIR R4 Data Types: Quantity.” https://hl7.org/fhir/R4/datatypes.html
- International Organization for Standardization. “ISO 11240:2012 Health Informatics: Identification of Medicinal Products, Data Elements and Structures for the Unique Identification and Exchange of Units of Measurement.” https://www.iso.org/standard/55033.html
- European Medicines Agency. “Referentials Management Service (RMS).” https://www.ema.europa.eu/en/human-regulatory-overview/research-development/data-medicines-iso-idmp-standards-overview/substance-product-organisation-referential-spor-master-data/referentials-management-service-rms
- MESA International. “B2MML Common Schema (B2MML-Common.xsd), Version 0701,” QuantityValueType and UnitOfMeasureType definitions. 2023. https://github.com/MESAInternational/B2MML-BatchML/blob/master/Schema/B2MML-Common.xsd
- NASA Mars Climate Orbiter Mishap Investigation Board. “Phase I Report.” November 10, 1999. https://llis.nasa.gov/llis_lib/pdf/1009464main1_0641-mr.pdf








Your perspective matters—join the conversation.