skip to content

Traceability links in an area have gone unmaintained for two releases. What can an impact analysis built from them still be used for?

level: seniorimportance: should knowfreq 34%

answer

  1. Decay hits inclusions and omissions differently
  2. Old links still name real relationships
  3. Absence from the set proves nothing
  4. A floor for scope, never a ceiling
  5. Quote the record's age with the set

basics

~10 s

Use it as a floor, never a ceiling. What stale links include still names relationships that were genuinely real; what they omit proves nothing, because the record stopped following the product two releases ago.

solid answer

~50 s

Decay is asymmetric, and that asymmetry decides what the analysis is still worth. **Inclusions age well**: a traceability link written two releases ago recorded a relationship somebody knew was real, and unless the area was deleted it probably still holds, so the derived set is a legitimate starting list. **Omissions age badly**: everything built, moved or re-owned since then left no link behind, so absence from the set is evidence of nothing at all. The honest uses follow directly — widening scope, seeding a conversation with the people who own the surrounding areas, and cross-checking a set you reconstructed by reading. The dishonest ones are the mirror image: narrowing scope, arguing that an area is unaffected, or quoting the set as release evidence. Label the set with the record's age, and write back the links you had to reconstruct while walking.

code

pseudocode · 15 lines
pseudocode
set = walk(changedRequirement, record)
age = releasesSinceLinksMaintained(area)

if age >= 2:
    allow("widen scope with these items")
    allow("seed a conversation with the owning teams")
    allow("cross-check a set reconstructed by reading")
    deny ("narrow scope to only these items")
    deny ("claim an unlisted area is unaffected")
    deny ("quote as release evidence")
    label(set, "floor only - links last maintained " + age + " releases ago")
else:
    allow("normal derivation with a stated confidence band")

writeBack(linksReconstructedDuringWalk)

go deeper

for a junior

Know that a traceability record decays: work built since it was last maintained leaves no links behind it. A set derived from an old record is a starting point rather than an answer, and it has to be labelled as one.

for a middle

Explain the asymmetry — what an old record includes was once real, what it omits proves nothing — and be able to give one honest use and one dishonest use of a set derived that way.

for a senior

Show that you still produce the analysis, label it with the record's age, and write back the links you reconstructed while walking. Expect to be pressed on what you would refuse to let the set be used for, and why the refusal holds under schedule pressure.

for a principal

Own when a decayed record is repaired versus retired. A half-maintained record is worse than an honest statement that an area has no traceability and is verified another way, because it still looks authoritative. Make that call explicitly and tell the people who quote it which it is.

## How a record decays A traceability record is maintained by hand, at the moment the work is done. Skip that moment and the record does not go uniformly stale — it goes **partial in a specific direction**. Everything that existed when links were last written is still described. Everything built, moved, split, merged or re-owned since then is simply absent. Two releases of unmaintained links means two releases of new work the record has never heard of, sitting beside older work it describes accurately. This matters because both instinctive reactions are wrong. Declaring the record useless throws away real information; treating it as current produces a confident answer resting on a two-release blind spot. ## The asymmetry that decides every use | Property | Inclusions: what the set lists | Omissions: what it does not | | --- | --- | --- | | What they meant when written | A real, deliberate relationship | Nothing — absence was never recorded | | How they age | Slowly; the relationship usually persists | Instantly; new work is missing from day one | | What decay does to them | Occasionally leaves a stale entry behind | Silently hides everything recent | | What you may conclude | This is worth a look | Nothing whatsoever | An item present in a decayed set was connected to that requirement at some point, by somebody who knew. Unless the area was deleted, the relationship is probably still live, so the item is a legitimate candidate. An item absent from the set may be absent because it is unrelated, or because it was built last month. The record cannot tell you which, and that is not a gap you close by walking harder. So the rule is short: **the set is a floor, never a ceiling.** ## What it may honestly be used for - **Widening scope.** Add the items the set names to whatever you already intended to look at. Inclusions survive neglect, so this is free signal. - **Seeding a conversation.** Take the list to the people who own the surrounding areas and ask what is missing. A concrete list of eight items gets a far better answer than "does this change affect you?". - **Cross-checking your own reading.** If you reconstructed the impact set by reading the requirement, the old record is a second opinion; anything it names that you missed is worth pursuing. - **Sizing the repair.** How much you had to reconstruct by hand measures how far this area has slipped, and that measurement is the argument for funding the fix. ## What it may not be used for - **Narrowing scope.** Cutting planned work because an area is absent from the set rests entirely on omissions, which are the part decay destroys first and most completely. - **Claiming an area is unaffected.** The same error, stated more confidently, usually to somebody who cannot see how old the record is. - **Release evidence.** A claim that what was asked for has been verified needs a record that follows the product. A decayed one cannot support it, and a regulated context will not accept it. - **Comparing against an earlier analysis.** Two sets derived either side of a period of decay are not comparable, and any apparent shrinkage is an artefact of the neglect rather than a fact about the change. ## Working honestly with what you have Three practices keep a decayed analysis useful instead of misleading. **Label the set with the record's age.** Put it on the set itself, beside the items, because the set is what gets forwarded and the covering note is what gets dropped. *"The links behind this list were last maintained two releases ago; use it to widen, not to narrow."* One line, travelling with the thing it qualifies. **Repair what you reconstruct.** Producing the set forces you to work out by hand which criteria and cases genuinely belong to this requirement. Writing those links back costs minutes while the knowledge is still in your head, and it means the next analysis in this area starts from a record rather than from a reading. That is far cheaper than a repair project, and it targets exactly the areas people actually walk. **Decide whether the area is being repaired or retired.** A record half-maintained forever is the worst of the available options, because it looks authoritative and is not. If nobody will fund the upkeep for an area, say so plainly — this area has no usable traceability and is verified another way — so that nobody quotes it as though it did. That is an honest position. A decayed record quietly presented as evidence is not. The underlying discipline generalises well beyond traceability: state what your evidence can support, and refuse the uses it cannot, especially at the moment somebody wants a fast answer and the refusal is inconvenient.

  • What is the smallest repair worth making while you are already walking a neglected record?
    Write back the links you reconstructed by hand. Producing the set already forced you to work out which criteria and cases belong to this requirement, so recording them costs minutes while the knowledge is fresh and means the next analysis in that area starts from a record rather than a reading. Repairing the whole area is separate, larger work.
  • Why is it wrong to use a stale set to argue that an area is unaffected?
    Because that argument rests on absence, and absence is exactly what decay manufactures. Anything built or re-owned since the record was last touched is missing by construction, so an area can be entirely absent from the set and still be the one the change breaks. Inclusions survive neglect; omissions do not survive it at all.
  • How do you stop a stale analysis from being quoted later as though it were current?
    Stamp the record's age onto the set itself rather than into a covering note, because the set is what gets forwarded and the note is what gets dropped. One line saying the links behind it were last maintained two releases ago travels with the items and forces the next reader to discount it deliberately.

An out-of-date street map still shows you roads that exist; what it cannot tell you is that a new bypass is missing from it entirely.

saying these in an interview costs you the question

  • Treats a stale set as proof that an area is unaffected
  • Discards the analysis entirely because the record is old
  • Quotes a set built from neglected links as release evidence
  • Never states how old the links behind the set are
  • Refuses to proceed until the whole record has been repaired