Norva

Exact or Broad Search: Which Query Should You Try First?

Start broad when memory or metadata is uncertain; start exact only when the phrase and the interface’s matching rules are both known.

In short: Start broad when you remember only part of a title, spelling is uncertain, or alternate titles may exist. Start exact only when you know the full phrase and the interface’s exact-match syntax. After the first query, let the result shape decide: narrow a large relevant set; broaden or repair a zero-result query; verify a single result with independent metadata.

Exact and broad are not rival philosophies. They are tools for different levels of certainty and different result sets.

Define exact and broad for the interface

A broad search may match words across a title or several indexed fields, regardless of precise phrase order. An exact search may require the same words in the same order or a specific field value.

Those meanings vary by product. The Library of Congress Search Help uses quotation marks for exact phrases in its search, and its Advanced Search guide distinguishes “contains all,” “contains any,” “exact phrase,” and “starts with.” Do not transfer that syntax to Norva unless current guidance confirms it.

Use the certainty-result search quadrant

Choose a starting action:

Memory certaintyExpected result sizeStart withGoal
LowUnknownBroad distinctive fragmentDiscover stored wording
HighLargeExact or fielded titleConfirm a known phrase
HighSmallBroad title plus contextAvoid brittle overconstraint
LowLargeBroad fragment, then one filterLearn and narrow

After the query, reclassify based on the actual results rather than defending the initial strategy.

Start broad in four situations

Use a broad query when:

Choose the rarest reliable word through the partial-title method. Broad does not mean vague: “moon” may be broad, while a distinctive surname or unusual title word can be both short and informative.

Start exact in limited situations

An exact query is useful when:

Exact matching can fail when the stored title includes a subtitle, alternate spelling, localised form, edition label, or metadata error. Treat zero results as a signal to broaden, not proof the item is absent.

Follow the broad-to-exact two-pass

Pass 1: discovery

Run one broad, high-information term. Inspect titles, years, people, types, and sources in the first results. Identify how the catalogue spells and structures the likely record.

Pass 2: confirmation

Use the discovered title form plus one verified context clue. If exact syntax is supported, apply it only now. Confirm the chosen record using year, creator, synopsis, series placement, or version.

This sequence lets the catalogue teach you its stored wording before you demand an exact match.

Respond to each result shape

For repeated names, use the similar-title workflow. For huge result sets, use the large-catalogue search plan.

Record whether a change helped

QueryExact/broadOne changeRelevant candidatesMissed expected item?Next action

Precision is the share of results that appear relevant; recall is whether expected items are included. You do not need formal metrics for one search, but the concepts clarify the trade-off.

Norva can organise compatible authorised sources, while its searchable fields and matching behaviour can evolve. Observe current results and consult support for product-specific syntax.

Common mistakes and limitations

Broad queries cannot recover absent aliases, and exact queries cannot correct wrong metadata. Search strategy and catalogue quality remain connected.

Frequently asked questions

Is exact search always faster?

It can reduce results, but only when the stored phrase and syntax match. A failed exact query often costs more attempts than a well-chosen broad fragment.

Should I use quotes around every full title?

Only when the interface documents quotation marks as exact syntax and the title form is certain. Otherwise, quotes may be ignored or treated as literal characters.

What if broad and exact searches show the same results?

The interface may not support a distinct exact mode, or the result set may already be small. Confirm current guidance and focus on contextual verification.

Your next step

See how Norva works

Sources