%202.png)
Multi-Source Research Synthesis
Combine primary research, internal data, operational records, expert input and previous studies into one assessment, with each source's status and provenance kept visible.
Free. Works on Claude, ChatGPT, Gemini or any assistant that accepts a skill file.
What this skill does
The method, encoded.
Real projects almost never rest on one kind of evidence. A typical assessment draws on a survey you ran, a data extract the client supplied, a set of usage records, two published reports, a study from three years ago, and the view of practitioners with a decade in the category. These cannot be pooled, and the failure is subtle: they get written into one narrative in one voice, and the reader can no longer tell which sentence rests on a measured finding and which on somebody's impression.
This skill provides the discipline. It builds a source inventory before any synthesis begins, assigns each source an epistemic status saying what it can and cannot establish, and maps which source is authoritative for which question rather than treating all sources as evidence for everything. It aligns definitions and time periods, which almost never match and where forcing them creates durable false precision. It tests whether apparent triangulation involves genuinely independent errors. And it supplies a five-check procedure for the delicate case where a client's internal data contradicts your primary research, one that neither defers automatically nor dismisses.
The output carries full provenance on every claim, and states what no source covers.
Best used for
- Integrating commissioned research with an organisation's own operational data
- Reconciling a client's internal reporting with primary fieldwork
- Programme and performance evaluation drawing on monitoring data and beneficiary research
- Building a business case from mixed market, internal and primary evidence
- Bringing expert or stakeholder input into an evidence base without inflating it
- Assessments where budget covered only part of the question
Typical inputs
What you give it.
The decision or question the integrated assessment must serve, Primary research outputs (survey, qualitative, behavioural), Client-supplied internal figures and data extracts, Operational, transactional or monitoring records, Appraised published and secondary sources, Expert, practitioner or stakeholder input, Previous studies with their method and date of data, Optional data dictionaries, field definitions and query logic, Optional definition-change history for operational systems
Typical outputs
What you get back.
Source inventory with type, provenance, date of data, method, definitions and limitations, Epistemic status statement per source, saying what it can and cannot establish, Question-by-source authority map, Definition alignment table with aligned, mappable or unalignable verdicts, Period and reference-frame alignment notes, Populated integration frame, question by source, Triangulation assessment with independence stated per convergence, Conflict diagnosis for internal data versus primary research, Gap list of questions no source covers, classified by kind, Integrated findings with full provenance and confidence per answer
Method coverage
What the skill works through.
- Why different kinds of evidence cannot be pooled
- Building a source inventory before any synthesis
- What each source type can and cannot establish
- Which source is authoritative for which question
- Aligning definitions across sources, and when alignment fails
- Aligning time periods and recall windows
- Triangulation: what convergence actually buys
- When internal data disagrees with primary research
- Handling expert and stakeholder input without inflating it
- Finding the questions no source covers
- Keeping provenance on every claim
Download
Free skill. One file.
Enter your email once. Every skill you download after that takes a single click.
How to install
Add the skill file and the five kernel protocols to a Claude Project, a ChatGPT Project, a Gemini Gem, or paste them at the top of any assistant conversation. Then give it your real research material, not a description of it.
Download skillQuestions
Common questions.
How do I combine survey data with the client's internal data?
Do not combine them into one figure. Establish which is authoritative for which question, align the definitions at the level of the operational rule, align the periods, then report each within its own scope. Behavioural and operational records are usually authoritative for what happened within their coverage; primary research is usually authoritative for why, and for the population the records cannot see.
The client's numbers contradict our research. What do we do?
Run five checks in order and stop at the first that explains the gap: definition, population covered, period and recall window, mode of measurement, and the provenance of the internal figure itself. Most conflicts dissolve at the first or second. If the conflict survives all five, do not pick a winner. State what each source is authoritative for, quantify the gap, and put the choice in front of the decision-maker.
What is triangulation, really?
Convergence between sources whose errors are genuinely independent. A survey and a set of system records agreeing is meaningful because self-report error and instrumentation error are unrelated. Two internal reports built on the same data feed agreeing is not triangulation at all. Always state what makes the sources independent before using the word.
Can I treat client-supplied figures as evidence?
Yes, within limits, and always labelled. A client-supplied figure is evidence about what the client's systems record, which is frequently exactly what you need and is not the same claim as evidence about the world. Label it as client-supplied at every appearance including the summary, and where you could not see how it was derived, say so at the point of use.
How should expert opinion be used in an evidence base?
As authoritative for mechanism, for interpreting anomalies and for the parts of a market nobody measures. Not as a source of prevalence, however experienced the expert. Attribute it, describe how it was gathered, and word it as judgement. The risk in a mixed-source frame is that a cell for expert input looks like the same kind of thing as a cell for a measured finding.
What is the difference between this and evidence synthesis?
Evidence synthesis combines comparable studies of the same kind, weighting them against each other claim by claim. Multi-source synthesis handles source types that cannot be pooled at all, and its work is inventory, authority mapping, alignment and provenance rather than weighting. The two are complementary: a set of comparable studies can be synthesised first, then enter a multi-source frame as one source.
Why do definitions matter so much?
Because they almost never match and the labels hide it. "Active customer" in a system, "regular user" in a survey and "account with a transaction in 90 days" in a finance report are three different constructs. Combining them produces a number that is more quotable than its caveat, and it will outlive the project.
What should the output look like?
A source inventory near the front rather than in an appendix, an authority map, a definition and period alignment note, findings question by question with the carrying source named and confidence stated, a conflict section with the diagnosis, and an explicit list of what no source covers. Every claim names its source, period and status without the reader having to look anything up.
The skill chain
Works well with.
Research where people already are.
Analyse it where you already work.
Yazi helps researchers conduct surveys, AI interviews and longitudinal research directly through WhatsApp.
