Holarch

Guide · MBSE practice

Product line engineering and variant management in MBSE

Most systems are built in more than one version. Product line engineering models the shared design once and manages the differences on purpose: a feature model says what can vary, configurations say what each product chose, and every requirement, component and test says which products it belongs to. This guide explains the method (the 150% model, feature models, valid configurations, effectivity, and documents, verification and baselines per product), then works it through on a CubeSat built in two versions.

On this page
  1. What product line engineering is
    1. When a product line pays off
  2. The 150% model
  3. Feature models
  4. Configurations and their validity
  5. Binding variability to the model
  6. Parameters per configuration
  7. Effectivity by serial number and date
  8. Documents, verification and baselines per configuration
  9. Common failure modes
  10. Worked example: EO-1 Flight and EO-1 Lite
    1. The feature model
    2. The two configurations and their verdicts
    3. What belongs to which product
    4. A value that differs: 5.4 Daily Data Volume
    5. A part change by serial number
    6. Verification per product
    7. One specification per product
    8. The EO-1 Flight PDR baseline
    9. What the checks find
  11. Product lines in Holarch
    1. Variability in SysML v2
  12. Frequently asked questions
    1. What is a 150% model?
    2. What is the difference between a variant and a version?
    3. When should I use effectivity instead of a feature?
  13. Try it in Holarch
  14. Exercise: find and fix four product-line faults
    1. Plant the faults
    2. Find them as a reviewer
    3. What the checks do not catch
  15. Where to go next

What product line engineering is

A product line is a family of products that share most of one design and differ at a few known points. A launch customer's spacecraft and a cheaper commercial version, a vehicle sold with three battery sizes, a radar with a domestic and an export variant, block upgrades of a ground station: each is one design with variants. Product line engineering (PLE) is the discipline of building and maintaining that family as one engineered asset rather than as a set of separate projects.

The alternative is clone and own: copy the first product's model, documents and test plans, then edit the copy for the second product. It is fast at first and expensive later. Every fix has to be made in each copy, nobody can say which differences are intended, and the copies drift apart.

PLE replaces the copies with three things:

  • One shared model that holds every requirement, function, component, interface and test of every product.
  • An explicit description of the variability: the choices that differ between products and the rules between them.
  • A binding from each variable item to the choices it depends on, so that any one product can be derived from the shared model on demand.

ISO/IEC 26550 describes a reference model for software and systems product line engineering. It separates domain engineering, the work on the shared assets and their variability, from application engineering, the work of deriving and delivering one product from them. ISO/IEC 26580 adds methods and tool capabilities for the feature-based form of PLE this guide follows. Product variants also live inside configuration management: ISO/IEC/IEEE 15288 makes configuration identification and baselines part of its configuration management process, and SAE EIA-649 treats effectivity, the units a configuration applies to, as part of configuration identification.

When a product line pays off

PLE has a set-up cost: a feature model to agree and a binding to maintain. It pays off when:

  • two or more products share most of their requirements, functions and architecture;
  • a change to the shared part should reach every product;
  • each product still needs its own specification, verification record and baselines;
  • new variants are expected, so the differences need to be described once and reused.

It does not pay off for one product that changes over time: units built in sequence with part changes need only effectivity.

The 150% model

The shared model of a product line is often called the 150% model. It holds the union of every product: both radios when one product has an S-band radio and another an X-band radio, the propulsion module some products may carry, the encryption requirement only one customer asked for.

Each product is a 100% view of it: the items that belong to that product, with its values, derived rather than copied. That gives the method its main property: a change to a shared item is made once and is in every product immediately, while a change to a variable item reaches only the products it belongs to.

Two kinds of difference have to be represented in the 150% model:

  • Presence: an item is in some products and not in others (the X-band radio, the encryption requirement, the X-band link test).
  • Value: an item is in every product but with a different value in some (a daily data volume of 3 GB in one product and 1 GB in another, a mass, a test result).

A third difference, time (which units or build dates), is handled with effectivity.

Feature models

A feature is a characteristic of the product that a customer, a market or an engineering decision can choose: "X-band Downlink", "Encrypted Downlink", "Propulsion". A feature model arranges the features in a tree and adds the rules that say which combinations are allowed. The notation goes back to feature-oriented domain analysis (FODA) in the early 1990s and is now the most common way to describe variability in product lines.

Each feature, and each group of child features, carries one of a small set of rules:

RuleMeaningExample
MandatoryEvery product that has the parent has this feature.Every spacecraft has a downlink.
OptionalChosen per product.Encryption, propulsion.
Alternative (choose one)Exactly one child of the group per product.The downlink is S-band or X-band, never both and never neither.
Or (choose at least one)One or more children per product.A sensor suite of visible, infrared or both.
RequiresChoosing this feature forces another.Encryption requires the X-band radio, whose modem supports it.
ExcludesThe two features never appear together.A deployable antenna excludes the stowed-volume option.
ConstraintAny other condition, written as a logical expression over features.Propulsion and not the small battery.

Name features after the product, not the documents or the parts: "X-band Downlink", not "Section 5 option B" or "Radio P/N 4471". One feature usually maps to several components, requirements and tests; keep the feature tree and the component tree separate and let the binding connect them. Write each rule into the model as soon as it is known.

Configurations and their validity

A configuration is one product: a named selection of features, such as "EO-1 Lite: S-band downlink, no encryption, no propulsion". The configuration is what a customer buys, what a specification is written for, what a test campaign verifies and what a baseline freezes.

A configuration's effective selection is larger than the features someone ticked. It also includes the parents of each chosen feature, the Mandatory children of each chosen feature, and the Mandatory features at the root. An item bound to a parent feature is therefore in every product that chose any of its children.

A configuration is valid when its effective selection follows every rule of the feature model. A check of every configuration gives one of three answers:

  • Valid: every rule holds.
  • Breaks a rule: for example, a product chose encryption without the X-band radio it requires.
  • Incomplete: a choice is missing, for example no downlink band in a choose-one group.

Validity is the cheapest check in PLE: an invalid configuration is a product that cannot be built as specified, found when it is defined rather than at integration. Check the feature model too: a feature no configuration selects is a future option or dead weight.

The number of possible configurations grows quickly: ten independent optional features allow 1,024 combinations. Real product lines define a handful of configurations deliberately and treat the rest of the space as not offered.

Binding variability to the model

The feature model says what can vary. The binding says which model items depend on which choices. Each variable item carries a presence condition: a logical expression over features (and, rarely, configurations) that is true for the products that include the item. Items without one are in every product, and in a well-built product line that is most of the model.

The binding has to cover every kind of item, or the 100% views will be inconsistent:

  • Requirements: a requirement that only some customers asked for is bound to the feature they chose. A requirement every product has, but with a different value, is not bound; it gets a value per configuration instead.
  • Functions: usually shared; one varies only when a product does something the others do not.
  • Components and interfaces: the X-band radio, its data interface and its RF link are all bound to the same feature. An interface whose two ends are never in one product is a binding error.
  • Tests and verification: a test of the X-band link belongs only to products with the X-band radio. A test both products run stays shared and records one result per product.
  • Relationships: a satisfies or verifies link usually follows its ends. Bind a relationship separately only when its two ends are in several products together but the link holds in only some.

Write presence conditions in terms of features, not configurations. "X-band Downlink" stays correct when a third product with an X-band radio is added; "EO-1 Flight" silently leaves the new product out. Name a configuration in a condition only for a genuinely one-off item, such as a customer-specific label.

Parameters per configuration

Many differences are values, not presence: a data volume, a mass allocation, a design life. A separate requirement per product for each value would multiply items and split their links.

The better pattern is one item with a base value and an own value per configuration where it differs. The item keeps one identity (one number, one set of links, one history), and each product's view shows its own value. A configuration without its own value uses the base.

Give an own value only where the value really differs, so a change to the base still reaches every other product. And check the text around the value: rationales, test thresholds and derived requirements often quote the number, and each quotation needs the same treatment or the product's view contradicts itself. The worked example has one.

Effectivity by serial number and date

Effectivity says from which unit or build date an item applies. It answers a different question from the presence condition:

Presence condition (Applies To)Effectivity
QuestionWhich products include the item?Which units, or which build dates, does it apply to?
Written asAn expression over features and configurationsSerial number ranges such as 1-20, 25, 40-, and a date range
Typical useOptions, variants, customer choicesPart changes during production, block upgrades, retrofits
Empty meansEvery productEvery unit, always

The two combine: an item is in a view when it belongs to the product and is in effect for the unit or date being looked at. A part change from serial number 2 on is modelled as two items, the old part with effectivity "SN 1" and the new part with "SN 2 and later", both with the same presence condition. That keeps both designs in the model, so the as-built record of unit 1 stays accurate after unit 2 changes.

A configuration has an effectivity of its own: the serial numbers built to it. It records which units are which product; it does not filter the items. Two configurations whose serial ranges overlap usually point to a numbering mistake.

Documents, verification and baselines per configuration

The point of the 150% model is that each product still gets everything a single-product project would have:

  • Documents per configuration. A specification generated for one product leaves out the sections that do not apply, shows that product's values, and says on its title page which configuration (and, if relevant, which serial number or date) it describes. Section numbers in such a document are view numbers: they close up the gaps left by removed sections, while the stored numbers in the model stay the same for every product.
  • Verification per configuration. "Verified" means verified for one product. A requirement shared by two products can be verified for one and still open for the other, so results are recorded per configuration and the verification matrix is read per configuration.
  • Baselines per configuration. A product's design review freezes that product. A configuration baseline records the whole product line labelled with the configuration under review, so it can be compared with a later state of the same product or with another product at the same moment.
  • Change control per configuration. A change request should say which products it changes. A change to a shared item changes every product; a change to one configuration's own value changes only that one.

Common failure modes

Failure modeWhat goes wrongBetter
Clone and own as a stopgapThe copy becomes permanent; fixes reach one product and not the other.Start the second product as a configuration of the first model.
Features named after documents or partsThe feature model becomes a parts list, and every design change rewrites it.Name features after customer-visible choices; let the binding map them to parts.
Presence conditions that name configurationsA new product silently misses items bound to "CFG-1".Bind to features; name a configuration only for one-off items.
Copies instead of own valuesTwo "Daily Data Volume" requirements, each with half the links.One requirement with a base value and an own value per configuration.
Stale quoted valuesA rationale or threshold still quotes the base value in a product that has its own.Review each product's view, not only the 150% model.
Rules kept outside the modelA configuration that cannot be built passes review.Write requires, excludes and constraints in the feature model; check every configuration.
Interfaces bound inconsistentlyAn interface whose ends are never in one product, or a component without its interface.Bind a component, its interfaces and its tests to the same feature; run the consistency checks.
Verification counted for the lineA requirement shows "verified" because one product passed.Read the verification matrix per configuration.

Worked example: EO-1 Flight and EO-1 Lite

The example is EO-1, a fictional 6U CubeSat that images farmland, modelled in Holarch as an LML project. It is the same model as the end-to-end MBSE example. The demo builds EO-1 as two products:

  • EO-1 Flight, the spacecraft for the launch customer, whose contract asks for full scenes delivered in one pass and encrypted in transit;
  • EO-1 Lite, a lower-cost spacecraft for commercial imagery buyers, with the S-band downlink of the original design.

The trade study guide compares S-band and X-band for the downlink and picks S-band on cost and ground-station support. That decision holds for the commercial product. Its sensitivity analysis showed X-band winning when data rate weighs heavily, which is the launch customer's case; a product line keeps both answers.

The feature model

On the Control & Reviews → Product Line page, EO-1 has five features:

FeatureKindChildrenRulesMeaning
F-1 DownlinkMandatoryChoose oneEvery product has exactly one downlink band.
F-2 S-band DownlinkOptional (child of F-1)2.2 GHz at 10 Mbps through the partner ground network.
F-3 X-band DownlinkOptional (child of F-1)8 GHz at 100 Mbps, for same-pass delivery of full scenes.
F-4 Encrypted DownlinkOptionalrequires F-3Payload data encrypted in transit.
F-5 PropulsionOptionalrequires F-3A cold-gas thruster for orbit maintenance. It fits only beside the X-band radio. Studied; no product has it yet.

The choose-one group is the heart of the model: the two radios are alternatives, so no product can have both or neither. The two requires rules record engineering facts: the encryption runs in the X-band modem, and the thruster fits only in the layout that the X-band radio frees up.

The two configurations and their verdicts

ConfigurationStatusSerial numbersSelectsVerdict
CFG-1 EO-1 FlightReleased1–2F-3 X-band Downlink, F-4 Encrypted DownlinkValid
CFG-2 EO-1 LiteDraft3 and laterF-2 S-band DownlinkValid

Both configurations also contain F-1 Downlink, as a Mandatory root feature and as the parent of their chosen band; the grid marks it as implied. To see the rules at work, tick Encrypted Downlink in the CFG-2 column. The verdict changes to "Breaks: Encrypted Downlink requires X-band Downlink." Encryption on the Lite product would be a design change, not a configuration change. Undo the tick with ⌘Z / Ctrl+Z.

What belongs to which product

Most of EO-1 has an empty Applies To: the structure, power system, imager, command link, flight software and most requirements are in both products. The items that differ are bound to features, never to a configuration:

Applies ToItemsIn
F-2 S-band Downlink1.5.2 S-band Transmitter; requirement 5.1 S-band Downlink RateEO-1 Lite
F-3 X-band Downlink1.5.3 X-band Payload Radio; interfaces IF-08.1 X-band Data Interface and IF-12.1 X-band Downlink; requirement 5.5 X-band Downlink Rate; test case TC-21 X-band Link TestEO-1 Flight
F-4 Encrypted DownlinkRequirement 4.4 Downlink EncryptionEO-1 Flight

The X-band radio is bound together with everything that depends on it: its two interfaces, the requirement it satisfies and the test that verifies that requirement. It also performs the same downlink operation as the S-band transmitter, and in each product only one of the two radios is present, so the function is shared while its performer varies.

The binding is not complete, and a review of EO-1 Flight's view shows where. Search that view for "S-band": requirement 5.3 Downlink Link Margin ("The S-band downlink shall close with at least 3 dB link margin…"), the test case TC-10 S-band Link Test and the interface IF-12 S-band Downlink have no Applies To, so they are in EO-1 Flight too, a product without an S-band radio. Model Checks cannot see this: the items are consistent as links, and only their text names the variant. Each needs a decision: bind it to F-2 and add the X-band counterpart where one is missing (5.3 has none; IF-12.1 and TC-21 already exist), or reword it so it holds for both bands. Searching each product's view for the names of the variants it does not have is a cheap review step for any product line.

Because the binding uses features, a future third product with an X-band radio and no encryption would pick up the radio, its interfaces and TC-21 automatically, and leave out 4.4.

A value that differs: 5.4 Daily Data Volume

Requirement 5.4 Daily Data Volume is in both products, so it is not bound. Its base text reads "The ground segment shall receive at least 3 GB of payload data each day." EO-1 Lite's commercial buyers need less, so CFG-2 has its own value: "at least 1 GB". Choose View as ▾ → CFG-2 EO-1 Lite and open 5.4: the description reads 1 GB and carries a CFG-2 value chip. In All configurations the field is marked Varies. 5.4 keeps one number, one set of links and one history.

The numbers behind it are simple. At 10 Mbps an 8-minute S-band pass returns 4.8 gigabits, or 0.6 GB; 3 GB needs five passes a day and 1 GB needs two. At 100 Mbps one X-band pass returns 6 GB, which is why EO-1 Flight delivers a full day of scenes in a single pass.

Now look at 5.1 S-band Downlink Rate, which applies only to EO-1 Lite. Its rationale reads "Clears one day of collection (3 GB, Req 5.4) in five 8 minute passes." In EO-1 Lite, 5.4 asks for 1 GB, so the rationale quotes a value the only product it belongs to does not use. Nothing is wrong with the rate itself; the justification is stale. This is the "stale quoted values" failure mode above, and it is found only by reading the product's own view. The fix is to give 5.1's rationale the Lite reasoning (two passes for 1 GB, with margin for missed passes), or to state the rate's real driver.

A part change by serial number

Open 1.3.2 Battery Pack. It has two heater characteristics: Thermostat Heater Power (2 W, effectivity SN 1) and Software-Controlled Heater Power (3 W, SN 2 and later). The heater changed from the second spacecraft on. Both are in every product: the change is about units, not options.

The serial ranges are where the two concepts meet. EO-1 Flight is serials 1–2, so the change happens inside one configuration: unit 1 flies the thermostat heater, unit 2 the software-controlled one. EO-1 Lite starts at serial 3 and has only the new heater. With View as ▾ → Effectivity ▸ → Serial Number… set to 1, the view shows unit 1 as built; set to 2, it shows the second spacecraft. The new heater draws 1 W more, which is the kind of value the change impact guide follows into the power budget.

Verification per product

The test case TC-11 UHF Command Test runs on both products, and holds one result per configuration: Passed in EO-1 Flight (500 of 500 commands within 0.4 s of their tags, on flatsat FS-1) and In Progress in EO-1 Lite (312 of 500 commands sent, on flatsat FS-3). The requirements it verifies are therefore verified for one product and open for the other.

Verification → RVTM shows both readings. In All configurations, a Configuration column group gives one rolled-up status per configuration, with "—" where a requirement does not apply (5.5 X-band Downlink Rate has no status for EO-1 Lite). In a configuration view the RVTM lists only that product's requirements and results, so "verified" means verified for that product. TC-21 X-band Link Test, still Not Run, appears only in EO-1 Flight's matrix.

To record new results for one product, choose it in View as ▾ and use Run Suite in Verification → Test Center: it says "Recording results for CFG-2." and saves the results as that configuration's values.

One specification per product

The requirements live in one document, SRS-001 EO-1 System Requirements Specification. To deliver it for EO-1 Lite, open it, choose Deliver ▾ → Specification (Word), set Configuration to CFG-2 EO-1 Lite and Effectivity to Any unit, and export. The Lite specification leaves out 4.4 Downlink Encryption and 5.5 X-band Downlink Rate, shows 5.4 with 1 GB, and names the configuration on the title page, in the footer and in the file name.

The EO-1 Flight specification leaves out 5.1 S-band Downlink Rate instead, so the requirements after it in section 5 are numbered one lower in that document, and the file says "Numbering as shown in CFG-1." The model's numbers do not change, so a citation of one product's requirement should name the configuration with the number.

Exported with All configurations and Show applicability on, the same document is the 150% specification: every section, with "Applies to: CFG-1" under the sections only EO-1 Flight has.

The EO-1 Flight PDR baseline

The demo holds the configuration baseline EO-1 Flight PDR, taken for EO-1 Flight's Preliminary Design Review, in Control & Reviews → Baselines & Versions. It records the whole product line labelled with CFG-1, made with Baseline a Configuration…. Comparing it with the current model by configuration shows what changed in EO-1 Flight since PDR, and nothing that changed only in EO-1 Lite.

CFG-1 is Released and CFG-2 still Draft: the Lite product is still being defined. With configuration control on, a configuration that is Released and in the latest baseline is controlled, so its own values change only through a change request. A change request's Configurations card names the products it changes, and a configuration's Owner must approve requests that change it.

What the checks find

Verification → Model Checks runs the Product line rule set in All configurations. On EO-1 the rules report no errors, and ProductLine.4 warns that Propulsion is selected by no configuration. Here that is intended: F-5 is a studied option, kept in the feature model so the rule "requires X-band Downlink" is recorded before anyone offers it. In a real program the warning is a prompt to decide, with the decision recorded: offer it, keep it as a study, or delete it.

Product lines in Holarch

Holarch implements the method in one model: the Product Line page holds the feature tree and the configurations with their verdicts, the inspector's Product Line group holds Applies To and effectivity, and View as ▾ shows and edits every page as one configuration. Documents, the RVTM, the Test Center, search (appliesto:CFG-2, varies:yes) and Model Checks follow the view.

Product lines are part of the Individual and Team plans. Baselines per configuration, comparing by configuration, configuration owners and the configurations a change request changes are part of the Team plan.

Variability in SysML v2

SysML v2 has variability in the language: a variation marks a point where products differ, its variant members are the allowed choices, an assert constraint rules combinations out, and a configuration is a usage that binds one variant to each point. In a SysML v2 project Holarch reads these directly: the Product Line page shows the variation points and the configurations with a verdict, and View as ▾ shows the model as one configuration. The Hybrid Car demo's CarLine, with the points pack and charger and the rule wallboxNeedsLargePack, is the example. See SysML v2 Variants and Configurations.

Frequently asked questions

What is a 150% model?

The shared model of a product line, holding the union of every product's items. Each product is a 100% view of it, derived by keeping the items whose presence condition is true for that product and showing that product's values. The name says only that the model is larger than any single product.

What is the difference between a variant and a version?

Variants exist side by side: EO-1 Flight and EO-1 Lite are both current products. Versions follow each other in time: the design at PDR, then at CDR. Variants are managed with features and configurations, versions with baselines, and units built over time with effectivity. A configuration baseline combines the first two.

When should I use effectivity instead of a feature?

When the difference is about which units were built when, not about what a customer chose. A part that changed from serial number 2 on is effectivity. A part that some customers order and others do not is a feature, even if the first customer to order it happens to get the later serial numbers.

Try it in Holarch

  1. Create the demo with Manage Projects → Project Files ▾ → New Demo Project → CubeSat (Guided systems engineering).
  2. Open Control & Reviews → Product Line. Read the feature tree and the two configuration columns; both verdicts read Valid.
  3. In the CFG-2 EO-1 Lite column, tick Encrypted Downlink. Read the verdict, then press ⌘Z / Ctrl+Z.
  4. Open requirement 4.4 Downlink Encryption and look at the inspector's Product Line group: Applies To is F-4, and the line under it names the one configuration it applies to.
  5. Choose View as ▾ → CFG-2 EO-1 Lite. Open 5.4 Daily Data Volume and find the CFG-2 value chip; then open 5.1 S-band Downlink Rate and read its rationale against it.
  6. Open Verification → RVTM in the same view, then click Leave in the banner and compare the Configuration column group in All configurations.
  7. Open 1.3.2 Battery Pack and its two heater characteristics. Choose View as ▾ → Effectivity ▸ → Serial Number… and enter 1, then 2.
  8. Open SRS-001, choose Deliver ▾ → Specification (Word), set Configuration to CFG-2 EO-1 Lite and export.
  9. Open Verification → Model Checks in All configurations and find the ProductLine.4 warning for Propulsion.

Product lines need the Individual or Team plan; the configuration baseline and configuration owners need Team. The full reference is in Product Lines and Configurations, including Applies To, editing in a configuration view, effectivity and documents, RVTM and verification.

Exercise: find and fix four product-line faults

A product line rarely fails all at once. It fails through small edits that each look reasonable: a rule added in a hurry, a serial range typed one unit too low, a test bound to the wrong feature, a rule change that quietly breaks a product already released. This exercise plants four such faults in the EO-1 product line and has you find each one with the checks a reviewer would run, then fix it. It also shows where the checks stop, because knowing what a tool does not check is part of using it well.

You need: a fresh copy of the CubeSat EO-1 demo and a plan with product lines (Individual or Team, or an academic or class account; on the Free plan the Product Line page is read only). Work in pairs if you can: one person plants the faults while the other looks away, then the other finds them. Alone, plant all four, close the project, and come back to it as a reviewer.

Plant the faults

  1. Conflicting feature rules. Open Control & Reviews → Product Line. In the F-5 Propulsion row's ⋯ menu choose Rules…, tick F-3 X-band Downlink under Excludes and click Apply. Propulsion now both requires and excludes the X-band downlink.
  2. Overlapping effectivity. In the CFG-2 EO-1 Lite column's ⋯ menu choose Effectivity… and change Serial Numbers from 3- to 2-, as if the Lite build had been pulled forward one unit.
  3. Missing verification for one configuration. Open the test case TC-10 S-band Link Test and, in the inspector's Product Line group, set Applies To to F-3, as if the link test belonged to the X-band radio.
  4. A change that invalidates a released configuration. Back on the Product Line page, open Rules… on F-4 Encrypted Downlink, tick F-5 Propulsion under Requires and click Apply, as if encryption hardware now needed the propulsion bay.

Find them as a reviewer

Use only what a reviewer would use: the verdict row and Check Configurations on the Product Line page, Verification → Model Checks (click ↻ Run Checks and look at the Product line rule set), and the RVTM, both in All configurations and with View as ▾ set to each product. Write down each fault you find, what showed it, and which product it affects, before you open the answers.

Show where each fault shows up, and the fix
FaultWhere Holarch shows itFix
Change that breaks a released configurationThe verdict of CFG-1 EO-1 Flight, which is Released, reads "Breaks: Encrypted Downlink requires Propulsion." Check Configurations lists it, and Model Checks reports it as ProductLine.3, an error.Remove Propulsion from F-4's Requires. Holarch did not refuse the edit: the rule lives on the feature, not on the released configuration, so the verdict is what catches it.
Overlapping effectivityProductLine.7 warns on both configurations ("Shares serial numbers with …"), and the Effectivity dialog says which range overlaps which. SN 2 is now claimed by EO-1 Flight and EO-1 Lite.Set CFG-2's serial numbers back to 3-.
Missing verification for one configurationProductLine.6 warns on TC-10: it is never in one configuration with 5.1 S-band Downlink Rate, which applies only to F-2. With View as ▾ → CFG-2 EO-1 Lite, the RVTM shows 5.1 without a test case; with CFG-1 it does not list 5.1 at all.Set TC-10's Applies To to F-2: it tests the S-band radio, so it belongs to the products that have one. Clearing it, as the demo ships, also works.
Conflicting feature rulesNothing turns red: no configuration selects Propulsion, so no verdict changes, and ProductLine.4 already warned that no configuration selects it. To test the rules, click + New Configuration, select Propulsion with X-band ("Breaks: Propulsion excludes X-band Downlink."), then with S-band ("Breaks: Propulsion requires X-band Downlink."). No valid product can include Propulsion.Remove X-band from Propulsion's Excludes, then delete the test configuration.

What the checks do not catch

  • Rules are checked through configurations. Holarch checks each configuration against the feature rules; it does not analyse the rules on their own, so a contradiction stays silent until a configuration selects the feature. A short-lived Draft configuration is the way to test a new rule.
  • Effectivity overlap is by serial number. ProductLine.7 compares the serial numbers of configurations. Two configurations whose build dates overlap are not reported.
  • A released configuration is not frozen against rule changes. With configuration control on, a configuration that is Released and in the latest baseline changes its own selection and values only through a change request. The feature rules it is checked against are separate items, so run Check Configurations whenever a change touches features.
  • Verification gaps per product show in the RVTM. ProductLine.6 finds a test bound so that it never meets its requirement. A requirement that has no test case in one product for another reason shows up in the RVTM of that product's view, so read it per configuration before a review.

Before you finish, run the checks once more: the verdicts should read Valid again, and the Product line rules should report only the ProductLine.4 warning for Propulsion that the demo has on purpose.

Where to go next