River
Y CombinatorBacked by Y Combinator
FREE TEMPLATE

Framework Matrix Template for Research

Two documents and three sheets that chart every case against fixed dimensions and give each blank cell a reason before it counts as a finding.

Free download  ·  No account needed

Framework Matrix

One dimension, six cases, and a pattern in the blanks

CaseReferralD5: Job-search behaviour since completing
P01Jobcentre-referredApplied to two warehouse roles since finishing; the barcode module is named in one cover letter
P03Jobcentre-referredOne completed online application, submitted with a family member present
P06Jobcentre-referredSet up two job-alert emails; has not yet applied to anything from them
P08Employer-referredNot applicable
P09Employer-referredNot applicable
P12Employer-referredNot applicable

All three employer-referred cases here hold the same blank on this dimension. The full matrix has five of five. Gap Cells names the reason so the blank reads as one finding about the dimension, not three cases with nothing to say.

A framework matrix crosses cases against a fixed set of dimensions, one cell per intersection, and every cell starts blank. A blank can mean one of four things: the dimension does not fit the case, it was asked and got no answer, the interview never reached it, or nobody has charted it yet. NVivo's own documentation for the feature is explicit that a cell is blank whenever no summary has been entered for that case and theme, with no distinction between the four. A spreadsheet renders all of them identically: empty.

The pack is two documents and three sheets. Framework Definitions fixes each dimension's boundary before a single case gets charted, the way the framework method's originators describe identifying the thematic framework as its own stage, ahead of charting. Framework Matrix charts every case against that boundary. Evidence Register holds the quote and transcript reference behind any cell that would read as a guess without one. Gap Cells gives every remaining blank one of four named reasons, and a pattern in those reasons is what the standing rule will not let the analysis skip past.

In the worked example, one dimension returns a structural not-applicable for five of twelve cases, and all five share one attribute: an employer referral for in-role upskilling rather than a Jobcentre referral tied to unemployment. Reading those five blanks as thin data would understate the matrix. Reading them as one pattern is the finding: the dimension needs a second form for employed cases, not five cases with nothing to say. Open the pack in River and the agent charts alongside you, or download the sheets and classify your own case set's blanks the same way.

Every sheet in the pack

The Framework Matrix (six of twelve cases), the Evidence Register and Gap Cells.

Framework Matrix

Six of the pack's twelve cases, charted against all five dimensions. Full dimension names are in Framework Definitions.

CaseReferralD1 EnrolmentD2 RelevanceD3 BarriersD4 ConfidenceD5 Job-search
P01JobcentreCaseworker booked her on after a second sanction warning; she asked to go once she heard the barcode-scanning module was includedHigh. Wants a warehouse role and names the inventory-scanning module as the reason she kept attending [Q-01]None reported. Attended all six sessionsWent from avoiding the shop's self-service checkout to using a laptop unsupervised by session fiveApplied to two warehouse roles since finishing; the module is named in one cover letter [Q-02]
P03JobcentreReferred after missing an online-only application deadline the month beforeThe online-forms module directly, naming it as the reason the earlier deadline was missedThe one bus route that reached the venue on time was cut mid-course; missed two sessions [Q-04]Limited change reported. Still asks a family member to check a form before submitting itOne completed online application, submitted with a family member present
P06JobcentreReferred as a condition of his claim; open about resenting the requirementRated the jobs-board walkthrough as the only module worth the six weeks [Q-07]No sessions missed, though he said twice he “only came because he had to”Not askedSet up two job-alert emails; has not yet applied to anything from them
P08EmployerStore manager enrolled her and three colleagues after head office mandated a new scheduling systemEntirely about the scheduling-software module; the rest “were not really why I was there”None. Sessions were scheduled inside her paid shiftMoved from asking a colleague to build the weekly rota to building it herself within three sessionsNot applicable
P09EmployerEmployer paid for the course after a paper stock system was replaced at her warehouseHigh for the spreadsheet module, used daily; called the CV-builder “not for me”One session missed for a delivery backlog at workNow builds her own stock-count formulas instead of asking the systems team [Q-09]Not applicable
P12EmployerPut forward by his line manager after a training-needs review flagged a gap in basic web toolsVery high for the website-building module; asked to redo that session's exercise a second time [Q-12]None. Manager blocked his calendar for all six sessionsWent from no experience building a web page to a working one-page site by the final sessionNot applicable

P06's blank is a skipped question, not a finding. P08, P09 and P12's are the same reason repeating. Gap Cells tells the two apart.

Evidence Register

The quotes and transcript references behind the bracketed tags in the Framework Matrix, for the six cases shown there.

IDCaseDimensionVerbatim excerptRef
Q-01P01D2 Relevance“Once I heard scanning was in there I thought, right, that's the one I need for the warehouse job.”L096
Q-02P01D5 Job-search“I put the scanning bit right in the cover letter, told them I'd already used the system.”L211
Q-04P03D3 Barriers“The 46 got cut in week three. That was the only bus that got me there on time, so that was that, twice.”L134
Q-07P06D2 Relevance“Only the jobs-board bit, walking through the actual site. The rest I could take or leave.”L102
Q-09P09D4 Confidence“I build the count formula myself now. Used to just ask whoever was free on the systems desk.”L149
Q-12P12D2 Relevance“Could I do that one again? The website bit. I want to get it right before I put it live.”L141

Not every charted cell needs a tag. These six are the sentences that would read as a guess without the quote behind them.

Gap Cells

Every blank cell left in the twelve-case matrix, classified by reason before the Analysis Narrative is allowed to cite any of them.

CaseReferralDimensionReasonNote
P04JobcentreD3 BarriersAsked, no answerAsked directly twice; both times she said “no, it was fine” with no elaboration
P05JobcentreD2 RelevanceNot yet chartedInterview fully coded; charting queued behind the D5 pattern below
P06JobcentreD4 ConfidenceNot askedInterview overran on D2 and D3; a follow-up call is scheduled
P08EmployerD5 Job-searchNot applicableAlready employed; referred for in-role upskilling, not job-seeking
P09EmployerD5 Job-searchNot applicableSame reason as P08
P10EmployerD5 Job-searchNot applicableSame reason as P08
P11EmployerD5 Job-searchNot applicableSame reason as P08
P12EmployerD5 Job-searchNot applicableSame reason as P08. Five of five employer-referred cases; zero of seven Jobcentre-referred

One dimension accounts for five of the eight blanks, and all five share a reason and a referral route. That repetition is the finding the standing rule exists to surface.

What's in the pack

01

Framework Definitions

Fixes what counts as a case and what each dimension covers before a single transcript is charted, including the boundary line a later not-applicable pattern gets checked against. Works as well from transcripts already prepared for coding as from an appraised document corpus directly.

02

Framework Matrix sheet

Every case charted against every dimension in the case's own terms, with a bracketed tag into Evidence Register wherever a cell would read as a guess without the quote behind it. Where more than one person charts, reliability scoring between coders is the complementary check, not a substitute for agreeing the dimensions first.

03

Evidence Register sheet

The verbatim quote and transcript reference behind any tagged cell, so a charted summary can be checked against what the case actually said rather than taken on trust.

04

Gap Cells sheet

Every remaining blank cell, given one of four reasons: not applicable, asked and no answer, not asked, or not yet charted. A not-applicable pattern sharing a case attribute is a finding about the dimension, not a run of thin data.

05

Analysis Narrative

Reads across each dimension before down any single case, citing the register rather than asserting a pattern, and states plainly what the matrix could not answer. If this analysis sits alongside survey results, the joint display is where the two come together.

06

A blank cell is not read until it is classified

The standing rule every prompt reads first. No conclusion gets drawn from an empty cell until Gap Cells says which of four reasons put it there.

How to use it

  1. 1

    Open in River, or download it

    Open the pack in River and the agent charts cases with you, or download the two Word documents and three CSV sheets instantly, filled in with the worked example.

  2. 2

    Fix the dimensions first

    Framework Definitions turns the evaluation's own questions into five or six dimensions with a stated boundary, before a single transcript gets read for charting.

  3. 3

    Chart each case, tag the evidence

    Framework Matrix reduces each case's transcript to one sentence per dimension, with a bracketed tag into Evidence Register wherever a quote is doing work a paraphrase alone could not.

  4. 4

    Classify every blank, then write up

    Gap Cells gives each remaining blank one of four reasons before Analysis Narrative is allowed to cite it. Reporting checklists ask whether an analytic frame was applied consistently, which this discipline is what actually answers.

Frequently asked questions

Is this template free?

Yes. Download the whole pack as Word documents and CSV sheets with no signup and no credit card. Edit with AI is a separate, optional path that has the agent chart the cases and classify the gaps with you. The template library holds the rest of the packs.

I already build framework matrices in NVivo. Why use this?

NVivo builds the grid and lets you type a summary into each cell, and its own help pages are explicit that an empty cell just means no summary exists yet, with no distinction between the reasons. This pack adds that distinction: a required sheet that gives every remaining blank one of four named reasons before an analysis narrative can cite it.

What format are the downloaded files?

CSV for the three sheets and Word documents for Framework Definitions and the Analysis Narrative, in one zip. They open in Excel, Numbers, Sheets, Word, Pages and Google Docs with nothing to convert. Inside River the same content opens as live Docs and Sheets.

How is this different from a blank framework-matrix spreadsheet?

A blank grid gives you rows, columns and nothing else; every empty cell you end up with still just reads as empty. This pack ships the boundary for each dimension before you chart a single case, plus a register for the quotes behind non-obvious cells. A required sheet classifies every blank one of four ways before it gets read as a finding.

What does 'Edit with AI' actually do?

It creates a free River account and installs this exact pack as a private workspace. Send the research questions and the case transcripts, and the agent charts each case against the fixed dimensions, holds the supporting quotes in the register, and classifies every remaining blank cell before drafting the narrative.

How many cases and dimensions does this need?

The worked example uses twelve cases against five dimensions, which is a workable size for a single matrix read across and down. Five or six dimensions is typical; past that, cells stop getting meaningfully compared against each other, which is the reason a matrix exists over separate summaries.

What if a dimension genuinely does not fit some of my cases?

That is a not-applicable, one of the four standing reasons a blank cell can carry, and it is worth reporting rather than hiding. The worked example shows the shape: one dimension does not fit an entire subgroup, and the fix is a second form of that dimension for them, not a forced answer.

Find out which of your dimensions a whole subgroup can't answer

Send the research questions and the case transcripts. The matrix comes back with every blank cell classified, not just charted.

Edit with AI