Running both strands at once, comparing them honestly, and treating disagreement as a finding rather than a problem.
In a convergent design both strands are collected in the same period, analysed separately, and then brought together to see where they agree, extend each other or conflict. In an embedded design a smaller strand sits inside a larger one, answering a secondary question. The decision that determines whether a convergent design works is made before collection: what you will do when the strands disagree. Divergence is usually the most informative result these designs produce, and only if you planned to treat it that way.
Both strands run in the same period and are analysed independently, each on its own terms and with its own quality criteria. Only then are they brought together.
The independence matters. If the qualitative analysis is conducted with the survey results already in mind, it is no longer an independent view of the question — it becomes an illustration of the numbers. Analysing separately before merging is what makes the comparison meaningful, and it should be stated in the methodology chapter that this is what you did.
Use it when you want two independent perspectives on the same question, when both are equally important to the answer, and when you have the capacity to run both properly. The last condition is the one people underestimate: a convergent design is two full studies conducted simultaneously, held to the standards of both traditions.
The sampling decision. The two strands may use the same participants or different ones, and both are legitimate. Same participants allow case-level comparison — you can examine whether an individual’s survey responses match what they said in interview, which is a genuinely powerful move. Different participants give a broader base but limit you to comparison at group level. Decide deliberately and say why, since the choice determines what kind of integration is available to you later.
Sample sizes do not need to match, and generally should not. Each strand is sized by its own logic: the quantitative strand by what its analysis requires, the qualitative strand by the depth its question needs. Recruiting four hundred for one and sixteen for the other is entirely normal and should not be apologised for.
This is the situation the design exists to handle, and the point at which most doctoral candidates lose their nerve.
The instinct on finding a contradiction is to treat it as an error — to look for what went wrong, or to quietly favour whichever strand supported the expected conclusion. Both responses discard the most interesting result the study produced.
Divergence usually means something specific: the two strands are capturing different aspects of the phenomenon. Common patterns worth recognising:
The methodology chapter should say in advance how divergence will be handled: examined for these explanations, reported in full, and interpreted rather than resolved. A study that reports a well-explored contradiction and what it means is considerably more convincing than one where everything agreed — which sometimes prompts the question of whether disagreement was looked for.
Where the divergence turns out to be a methodological artefact rather than a substantive finding — an item that was misread, a sampling difference between strands — that is also a legitimate conclusion, provided you show the reasoning that led to it.
A smaller strand sits inside a larger design, answering a secondary question the primary method cannot address. The strands are explicitly unequal, and saying so is part of the design rather than a weakness to conceal.
The classic uses:
What the design asks of you is precision about the secondary strand’s role: what question it answers, why the primary method cannot answer it, and how its findings will be used. An embedded strand without a stated job tends to appear in the thesis as a short unconnected chapter that adds nothing.
Where a qualitative strand is embedded in an intervention study, decide when it runs. Interviewing during the intervention risks influencing it; interviewing afterwards risks reconstructed accounts. Either is defensible with a reason.
The structural problem in both designs is the same: results reported strand by strand, with integration deferred to a paragraph in the discussion. That is two studies in one document.
For a convergent design, report each strand’s results in enough detail to stand on its own, then a distinct integration section that does the comparison explicitly. A joint display — a table setting the quantitative finding, the corresponding qualitative finding, and what the combination means side by side — is the standard tool and gives an examiner something concrete. Divergent cells should be as visible as agreeing ones.
For an embedded design, report the primary strand as the main study, with the embedded strand positioned where it does its work — a process evaluation belongs alongside the outcome findings it explains, not in a separate chapter at the end.
In both cases, state your quality criteria separately for each strand: validity and reliability for the quantitative work, trustworthiness for the qualitative. Applying one framework to both is a small error that signals a larger confusion, and it is easy to avoid.
A design where quantitative and qualitative strands are collected during the same period, analysed independently of one another, and then brought together to compare. The independence of the two analyses is what makes the comparison meaningful — analysing the qualitative data with the survey results already in mind turns it into an illustration of the numbers rather than a second view of the question.
Either is legitimate, and the choice determines what integration is possible later. The same participants allow comparison at case level — whether an individual’s survey responses match what they said in interview — which is a powerful move. Different participants give a broader base but restrict you to group-level comparison. Decide deliberately and state the reason.
Report it and interpret it — it is usually the most informative result the design produces. Divergence commonly means the strands captured different aspects of the phenomenon: what people report versus what they do, an average concealing subgroups, or participants understanding a construct differently from how the instrument defined it. Examiners find a well-explored contradiction more convincing than tidy agreement, provided you engage with it rather than favouring the strand you preferred.
A design where a smaller strand sits inside a larger one to answer a secondary question the primary method cannot address — most commonly a qualitative process evaluation inside an intervention study, addressing how something was implemented and why it worked in some settings and not others. The strands are explicitly unequal, and the secondary one needs a clearly stated job or it ends up as an unconnected chapter.
No, and matching them usually damages one or both. Each strand is sized by its own logic: the quantitative strand by what its analysis requires, the qualitative strand by the depth its question needs. A survey of several hundred alongside a dozen or so interviews is entirely normal and does not need apologising for in the limitations section.
What you will do if the strands disagree is a methodology decision, and making it afterwards looks like choosing the answer you preferred. Describe your design and a PhD in your field will work through it with you.
Leave an email or a WhatsApp number — whichever you prefer — and tick how we should reach you. We reach out within 30 minutes.
