Database Search Technique: The Nine Steps
Nine numbered tips, printed as one flat list. Six of them are a procedure you run in order; the other three are troubleshooting you enter only when the results come back wrong.
The nine tips as the source prints them
Working a database is the mechanical half of the literature search. The judgement half — which concepts you are actually looking for, and which resource is likely to hold work on them — is settled before you type anything, and the six rules the source gives for phrasing a query sit on literature search strategies. This page covers the nine numbered tips for operating a database once you are inside one.
Your concepts are not arbitrary either. The source's stated position elsewhere is that the key concepts in your conceptual framework direct both your search for literature and your reading, which makes step one a translation job rather than a brainstorm — see the conceptual and theoretical framework.
Two things are worth noticing before you use them. The list mixes instructions with questions: steps 5, 6 and 7 are phrased as questions about your result set, not as actions. And the sequence never returns to the top — nothing tells you to run the revised search again, although two of the three branches obviously require it.
Why the list is not a sequence
- Select concepts
- Identify keywords
- Select database
- Run the search
- Evaluate results
- Save the set
- Locate full text
Read that way, the procedure is seven moves with a diagnostic loop wedged between evaluation and saving. Steps 6 and 7 change the query, which means going back to run the search again and evaluating a second time. Expect several passes before the set is worth saving.
Steps 1 to 4: building the search and reading what comes back
The four steps that always run
Select the concepts, then identify keywords
The order in the source is deliberate — concepts first, keywords second. A concept is the idea you are researching; a keyword is a string a database can match. One concept usually needs several keywords, because different authors name the same idea differently and a database matches characters rather than intentions.
Select the most relevant database using those concepts
The source's instruction is to look in the library databases and choose on the basis of your concepts and keywords, noting that research engines work in different ways. It names five kinds of resource: a catalogue, a discovery service, an institutional research archive, a national aggregated database, and a database of dissertations and theses. Which is most relevant is a judgement about coverage, and it is one you may have to revisit at step 7.
Perform the search
Run the query built from your concepts and keywords. This is the only step in the list with no elaboration at all — the source supplies its query-construction advice separately, in the six search-strategy rules, not here.
Evaluate the results
Four instructions in one: look at the results, look at the citation abstracts and subject headings, ask whether they are relevant, and mark the useful citations. Note what you are evaluating. You have not retrieved any full text yet, and the source does not ask you to before marking an item as useful.
The concept-to-keyword move at step 1 is where most searches are won or lost, and the source gives it one sentence. The table below is an illustration built for this page, using a topic area drawn from the source's own list of project management research themes.
FROM CONCEPT TO KEYWORDS — AN ILLUSTRATION CONSTRUCTED FOR THIS PAGE
| Concept | Keywords and variants worth trying | Why the variants matter |
|---|---|---|
| Handover from construction to operations | handover · hand-over · transition to operations · commissioning · operational readiness | Engineering, operations and contract functions name the same event differently, and each community publishes under its own term |
| Readiness of the receiving organisation | operational readiness · asset acceptance · ready for operations · owner's team | The concept exists in practice long before it acquires a settled label in the literature, so the search has to cover several candidate labels |
| Defects identified at takeover | defect* · snagging · punch list · latent defect | Truncation catches singular, plural and derived forms in one term — the source's own worked examples of this are globali*ation and cook* |
The theme is taken from the supplied source's list of project management research themes. The keyword expansions are constructed here as an illustration and are not from the source.
Subject headings deserve a second look at step 4. The source names them as something to inspect without explaining what they are. As a general principle — this is not stated in the supplied material — they are the terms the database itself uses to index an item, so a clearly relevant result hands you the vocabulary the database expects, which is frequently not the vocabulary you started with. Feeding those terms back into step 1 is the cheapest improvement available to you.
Steps 5, 6 and 7: the three diagnostic branches
Enter one of these only when step 4 has told you the set is wrong. Each branch carries a different diagnosis behind it, and applying the wrong remedy makes the search worse rather than better.
Which branch you are in
WHAT EACH BRANCH SUPPLIES, AND WHAT IT LEAVES TO YOU
| Branch | Condition | The source's remedy | Threshold given |
|---|---|---|---|
| Step 5 | No results | Searches work best with keywords only; check for errors | None |
| Step 6 | Too many results | Check use of OR and AND; use the refinement options; restrict by publication type; search the abstracts field only | None |
| Step 7 | Too few results | Add terms and combinations; find synonyms; use truncation or wildcards; try other databases in the subject area | None |
"Too many" and "too few" are left entirely to your judgement. The source gives no counts, and no counts are supplied here.
Notice too that the step 6 remedies are of two different kinds. Checking your operators changes the logic of the query; restricting by publication type or searching only the abstracts field changes what the database will look at. The first fixes a query that was asking the wrong question; the second narrows a query that was asking the right one too broadly. The source lists them together without distinguishing them.
OR and AND: named once, never explained
Step 6 tells you to check your use of OR and AND. That instruction is the only appearance of either operator anywhere in the supplied material.
One practical consequence while the gap stands: if a search behaves in a way you cannot account for, do not guess at the operator logic. Rebuild the query in the database's advanced search form, which presents the same logic as labelled fields and drop-downs — the source's own search rules already tell you to use advanced searches and categories of searches — and compare the result counts against your typed query.
Steps 8 and 9: saving the set and reaching the full text
Step 8 is a single line — save the search results, using the email, print, save or export options. Step 9 handles the common case where a citation you marked is not available as full text: click the link the source calls a "Find it" link, and if that does not deliver the item, check the discovery service or the catalogue.
Two observations. Step 8 saves the results, not the search — there is no instruction anywhere to record the query, the database or the date. And step 9 stops at the point where the item is not held anywhere; the source says nothing about what to do next.
A search you can still defend in twelve months
- The concepts are written down separately from the keywords that express them
- Each concept has more than one keyword, including at least one term you took from a subject heading
- Every database searched is named, with the date it was searched
- The exact query string is recorded, operators included, not a description of it
- The result count and the number of items marked are both recorded
- The marked set has been exported or saved rather than left in a session that will expire
- Every marked citation that is not available as full text has a recorded status: obtained, requested, or not held
How the nine steps fail in practice
Three predictable failures
One search, run once
The numbering conceals a loop, so the list reads as something you complete. A single query against a single database, evaluated once, is a first pass — not a literature search, and not something you can describe as one.
Relevance decided on titles
Step 4 names three things to read: the citation, the abstract and the subject headings. Titles are the weakest signal available at that moment, and they are the one most often used alone because they are the only thing visible without a click.
Your vocabulary, not the field's
A search built entirely from the words you already use will return the literature that happens to share your vocabulary. The synonym instruction at step 7 and the subject headings at step 4 both exist to break that loop, and both are routinely skipped.
There is also a boundary worth marking. Step 4 asks whether results are relevant, and that is as far as the nine steps go on quality. Whether a marked item is any good is a separate assessment against separate criteria — see assessing the academic value of sources. Whether it earns a place in your review is a third question again, answered only by reading it properly: see evaluating an article before you cite it.
Everything you mark also has to be captured as a reference at the moment you mark it, not later. That discipline is covered on note-taking and bibliographic records, and the wider daily routine it belongs to is on checking the literature for your topic.
What to carry forward
- Six of the nine numbered tips are the procedure; steps 5, 6 and 7 are branches you enter only when the result set is wrong.
- Concepts come before keywords, and keywords before the database. One concept needs several keywords because the field will not have settled on yours.
- Evaluate on citations, abstracts and subject headings. Subject headings give you the database's own vocabulary — feed it back into the search.
- No results and too few results are different diagnoses with opposite remedies. Strip back the first; broaden the second.
- OR and AND are named and never defined in the supplied material. Get them from the database's own help documentation, not from a guess.
- The source saves your results but never your search. Record the database, the query, the date and the counts yourself.
Frequently asked questions
Are the nine steps meant to be followed in order?
Only six of them. Steps 1 to 4 build and run the search, and steps 8 and 9 save the results and retrieve the full text. Steps 5, 6 and 7 are diagnostic branches for a result set that came back empty, too large or too small, and you enter at most one of them. The supplied source numbers all nine in a single list without saying this.
How many results count as "too many" or "too few"?
The source gives no numbers at either end and no rule for deciding. Practically, the ceiling is what you can actually triage on citation, abstract and subject headings, and the floor is the point at which you are no longer seeing the same authors and works cross-referenced from different directions. Neither threshold is stated in the supplied material.
What do OR and AND actually do in a database search?
The supplied teaching source names both operators in the too-many-results branch and never defines them, gives no example and mentions no other operator. This library does not invent a definition. Take it from the advanced search help of the database you are using, since that is the only account guaranteed to match that database's behaviour, then confirm against your library's search guidance.
Should I read the full articles before marking them as useful?
Not at step 4. The source has you evaluate on the citation, the abstract and the subject headings, and mark useful citations on that basis; retrieving full text is step 9, after the set is saved. Reading properly is a separate later activity with its own set of questions.
What do I do if the full text is not available anywhere?
The source takes you as far as the "Find it" link and then the discovery service or catalogue, and stops there. It gives no further instruction. Record the item as not held rather than dropping it silently — an item you know exists but could not obtain is a real limitation of your search, and it is one you can state honestly.
Do I need to search more than one database?
The source only raises it as a remedy in the too-few-results branch: try other databases in your subject area. It does not require multiple databases as standard practice, and it does not say how many is enough. Since it also states that research engines work in different ways, a single-database search is a coverage decision you should be able to justify.
References and source attribution
- The supplied teaching source: master's-level teaching material on reviewing the literature, supplied as study notes and accompanying slides. The nine numbered tips and the three diagnostic branches are quoted from that material.
- O'Leary, Z. 2017, The Essential Guide to Doing Your Research Project, 3rd ed., Sage Publications, London (listed among the source's references).
- Bryman, A. 2016, Social Research Methods, 5th ed., Oxford University Press, Oxford (listed among the source's references).
- Naoum, S. G. 2013, Dissertation Research & Writing for Construction Students, 3rd ed., Routledge (set as reading in the supplied material).
Suggested questions for Ask KEVOS
- Turn my research question into concepts and a keyword set for a database search.
- My search returned nothing. Which of the source's branches applies and what should I check first?
- My search returned several thousand results. How do I narrow it using the source's step 6 remedies?
- What should I record about each search so I can describe it in a methodology chapter?
- Where should I go to learn how OR and AND work, given the source never explains them?