Home / Synonyms / anomaly vs discrepancyNo. 0085

anomaly vs discrepancy

Anomaly and discrepancy both flag that something is off, measured against different things. An anomaly breaks a pattern: one reading that does not fit the behaviour of everything around it — unusual, and possibly a real discovery. A discrepancy breaks a match: two figures that should agree and don't — usually an error waiting to be reconciled. Anomaly needs a pattern; discrepancy needs two records.

Quick rule
breaks the expected pattern, cause unknownanomaly
two things that should match but don'tdiscrepancy
anomaly

Fig. 1aanomaly.

/əˈnɑːməli//əˈnɒməli/·noun
vs
discrepancy

Fig. 1bdiscrepancy.

/dɪsˈkrepənsi//dɪsˈkrepənsi/·noun

Ask what the odd thing is being compared against. An anomaly stands against a background of expected behaviour — everything else pulses together and one light lags, so the deviation is from a norm, not from another number. A discrepancy stands against a counterpart: two clocks that should read alike, a ledger and a bank statement, a quoted price and an invoice. The distinction matters because the two carry different values. An anomaly may be a fault or may be the interesting thing in the data; a discrepancy almost always means someone has to find out which of the two records is wrong.

01What each means

anomaly

An anomaly is the data point that refuses to behave: the warm winter in a cooling century, the orbit that drifts off its predicted path, the payment that matches no invoice. From the Greek anōmalos, 'uneven' — not (an-) even (homalos). What matters is the background: an anomaly exists only against a field of agreement, and science watches that spot closely, because anomalies are where current theories quietly begin to fail.

discrepancy

A discrepancy is a mismatch where agreement was expected — the gap between two accounts that ought to tally. The figures in a report show a discrepancy; a witness's story has discrepancies with the evidence. The word does not merely note that two things differ, as a disparity in wealth does; it implies they were supposed to be the same, so the difference is a problem demanding explanation. A single unreconciled line can flag a discrepancy in the whole.

02At a glance
anomalydiscrepancy
Measured againsta pattern or expected behavioura counterpart that should match
Needsa background of normalitytwo records, figures or accounts
Usually meanssomething unusual — cause unknownsomething doesn't add up — likely an error
Valuemay be a fault or a findingalmost always a problem to reconcile
Natural verbdetect, investigate, explain an anomalyreconcile, resolve, account for a discrepancy
Adjectiveanomalousdiscrepant
03How to remember

The straggling light and the two clocks. The anomaly is one light half a beat behind thirty-four others: nothing to compare it with except how the field behaves, and it might be a fault or the most interesting thing on the panel. The discrepancy is the pair of clocks that should read alike and don't: two specific things that were meant to tally. Ask what it is being judged against. A pattern: anomaly. Another record: discrepancy.

04In use

anomaly

  • iThe satellite recorded a temperature anomaly nobody could explain.
  • iiOne anomaly in the fossil record forced the whole timeline to be redrawn.
  • iiiTheir unusually low error rate is an anomaly among comparable labs.

discrepancy

  • iAuditors found a discrepancy between the invoice and the delivery note.
  • iiThere's a discrepancy in the two witnesses' accounts of the evening.
  • iiiEvery discrepancy has to be accounted for before the books close.

Data work keeps both words busy and keeps them apart: anomaly detection scans for readings that break the expected pattern, while reconciliation hunts discrepancies between records that should agree. The value judgement differs too — teams chase anomalies hoping some are real signals, and chase discrepancies hoping all of them are clerical. Discrepancy's finer separation from disparity has its own page in the 'Related' list below; anomaly's from outlier is settled on the anomaly word page.

05Asked
What is the difference between an anomaly and a discrepancy?
Each is measured against something different. An anomaly deviates from a pattern: it stands out against a background of things behaving normally, and nothing but that background says it is wrong. A discrepancy is a failure to match between two specific things meant to agree — a ledger and a bank statement, two witnesses, two clocks. An anomaly needs a norm to break; a discrepancy needs a counterpart to disagree with.
In data analysis, when is it an anomaly and when is it a discrepancy?
Anomaly is the word when a value breaks the pattern the data usually follows — a spike in server traffic at 3 a.m., a sensor reading nothing else supports. Discrepancy is the word when two sources that should carry the same number don't: the warehouse count against the system count, the bank feed against the ledger. Anomalies are investigated to learn what happened; discrepancies are reconciled to find which record is wrong.
Can something be both an anomaly and a discrepancy?
Often, at different stages of looking — and the scenes above sit at those two stages. A revenue figure that breaks the trend is an anomaly, the light burning half a beat behind a field that otherwise agrees. Trace it, find the sales system and the accounts showing different totals, and you now have the two clocks as well: a discrepancy between specific records. One names how it looked; the other names what was wrong.
Is a discrepancy an anomaly?
Not automatically, though people use the words loosely. A discrepancy always implies a pair — two records that were supposed to agree — while an anomaly implies a background of normal behaviour it stands out against. A neat test sits in the prepositions: you can say what a discrepancy is between, but an anomaly is only ever against the trend, the norm, the expected. Where both apply, they are describing different aspects of the same trouble.
Is an anomaly always a problem?
No, and that is where the two words differ in value. A discrepancy is nearly always unwelcome: something is wrong and must be reconciled. An anomaly may be a fault, but it may equally be the most valuable thing in the dataset — unexplained readings have opened up new planets, new particles and new diseases. Scientists chase anomalies hoping some survive explanation; accountants chase discrepancies hoping none do.
Related