Data Analysis with Balansor

This guide explains how to process and analyze survey data using Balansor, Askalot's statistical analysis service. Balansor extracts survey responses into a Bundle, codes open-ended answers, applies post-stratification weighting, refines columns, evaluates data quality, and exports results in multiple formats — all driven from one pipeline board.

The Balansor pipeline board — one card per Bundle, each showing its Bronze → Silver → Gold stages.

Core Concepts

Bundles

A Bundle binds one Project, one Questionnaire, and a chosen — non-empty — subset of that Questionnaire's Campaigns to exactly one Bronze → Silver → Gold chain. It is the unit you work with in Balansor: the home page is a pipeline board of Bundle cards, and every action — extract, code, weight, refine, export, check quality — happens on a specific Bundle's stage.

A Bundle owns at most one dataset per stage. If you want to try a different weighting approach or a different set of computed columns on the same raw data, you clone the Bundle — the clone starts with an identical copy of the ready stages and can be re-derived independently, so you can compare variants side by side on the board.

No more picking datasets from a list

Earlier versions of Balansor asked you to pick a dataset from a flat list on every page. That's gone — everything now flows from the Bundle card and its three stage nodes.

The Medallion Architecture

Balansor organizes each Bundle's data using a medallion architecture with three progressive stages. Askalot's stage names carry specific meanings that differ slightly from generic "medallion" usage elsewhere — the plain-language subtitles on each board node ("raw extraction", "coded + weighted", "analysis-ready") are there so you never need to think in jargon:

🥉 Bronze          🥈 Silver              🥇 Gold
(raw extraction) → (coded + weighted)  →  (analysis-ready)

🥉 Bronze — raw extraction

• One row per completed survey from the Bundle's chosen campaigns — everyone the fielding returned, drawn respondents and volunteers alike
• No statistical corrections applied — original response values preserved
• Foundation for everything downstream; re-extracting picks up newly completed surveys

🥈 Silver — coded and weighted

• Open-ended text is coded first. If the questionnaire has open-text questions, Balansor turns their answers into labelled categorical variables before anything else happens on this stage — see Open-End Coding.
• Weighting runs once, on the coded data. Post-stratification weights are then applied so the sample's demographics line up with your target population — and the newly coded dimensions are available as weighting variables alongside demographic factors.
• Includes a _weight column for statistical analysis and one column per coded dimension you kept.

🥇 Gold — analysis-ready

• Final dataset ready for publication or delivery
• The Silver file with the columns you chose, in your order and under your names, plus computed columns (arithmetic, range-bucket recode, concatenation)
• Highest quality tier for client delivery

Why coding comes before weighting

This ordering matches how experienced survey researchers work: you code open-ended answers into categories first, then weight the whole case base once — including the new categories as inputs, not as an afterthought. Weighting a dataset a second time, or weighting anything other than the Bundle's own raw extraction, isn't offered — it would produce numbers that don't mean anything.

Why This Staged Approach?

• Reproducibility: Raw data preserved; every processing step is visible on the board
• Flexibility: Export from any ready stage based on what a client needs
• Quality assurance: Progressive refinement with checks at each stage
• Auditability: Full lineage from raw responses to final deliverables

Getting Started

Accessing Balansor

Navigate to Balansor from the main navigation menu or directly via your tenant URL:

https://balansor.{tenant}.askalot.io

The Pipeline Board

The board is the Balansor home page and the bundle picker. Each card represents one Bundle and shows:

• Bundle name and how many campaigns feed it
• Three stage nodes — Bronze, Silver, Gold — each showing its current status: missing, working, ready, failed (with a reason), or stale (needs re-deriving)
• Per-node actions: extract/re-extract Bronze, derive/re-derive Silver, refine/re-refine Gold directly from the card, plus a Quality button

Click a Bundle's title (or the Quality button) to open its Overview page — the full Representativeness story, the analysis-base threshold, current-Strategy controls, and the Research Paper section. From there, the page row under the navbar opens the Bronze, Silver and Gold phase pages for that Bundle — each hosts the full editing experience for its stage (the Silver page's Calibration Targets editor, the Gold page's Column Layout, and so on) that the board's inline actions don't expose, and each ends with an Export section for that stage.

When the controls are hidden

If you can read a Bundle but not change it — you hold the analyst role, or you are visiting a sealed organization such as the demo — the board still shows every Bundle you have access to with full status and quality, but the create/run/delete/clone controls are hidden.

Creating a Bundle

1. On the board, click "New Bundle"
2. Give it a name
3. Choose the Project
4. Choose the Questionnaire — this defines the dataset's columns

5. Select the Campaigns to include (a "select all" option is available). At least one is required — a Bundle is defined by the fielding it analyses, so there is nothing to extract without one.

   A volunteer who walked in on one of your campaigns, or arrived through its open link, needs no setting: their survey belongs to that campaign and arrives with it, flagged so you can tell them from the people the sample design selected. Previews — a designer's test walk, a reviewer opening a share link — never arrive at all, and there is no option that admits them 6. Click "Create"

The new Bundle appears on the board with an empty Bronze node, ready to extract.

Extracting Bronze

Click "Extract" on a Bundle's Bronze node. This pulls one row per completed survey across the chosen campaigns, using the questionnaire's structure to define the columns — so the schema stays consistent regardless of which campaigns are included. Each row records how that person arrived, drawn or volunteer, and how the survey was taken, its mode (direct, ai_assisted or field_interview), so every later stage can report either split.

Multi-part items expand into exactly the columns the questionnaire declares: one column per sub-question in a question group, and one per row × column cell in a matrix question (a 4-row, 1-column matrix yields exactly four columns). The extracted schema always mirrors the questionnaire definition — the same declared structure the formal verification covers — with no padding or placeholder columns.

Multi-select questions get one extra column per option. A multi-select answer is stored as the combination a respondent chose, which is faithful to what they submitted but cannot be weighted: every observed combination would become its own weighting cell, and the options do not divide the sample into groups anyway, since one person can pick several. So alongside the combination column, extraction adds a yes/no column for each declared option — the same shape open-end coding produces — and those are what you can weight and cross-tabulate on. A respondent who skipped the question leaves the option columns empty rather than "no", so a skip is never counted as a rejection of every option.

An older Bundle needs a re-extract to get them

Bundles extracted before this existed keep their original columns until you re-extract; nothing changes underneath a dataset you have already worked with. So if a questionnaire has multi-select questions but the Calibration Targets editor offers no option columns for them, the extraction predates the feature — re-extract and they appear. Re-extracting marks Silver and Gold stale as usual. (Reading a Bundle's Calibration Targets over the API also returns a bronze_schema_stale flag that answers this directly.)

Respondent attributes. Each row also carries the respondent's age, gender and location, in the columns _demographic_age, _demographic_gender and _demographic_location. A respondent's custom attributes are different: one becomes a column, named _custom_<key>, only when a Sampling Strategy factor or a Calibration Target of this Bundle names it. An attribute nobody names is never copied into the dataset, so contact details and other fields an import carried stay on the respondent record. A dataset that was already extracted is not rewritten when a respondent is edited or deleted afterwards, or when your design stops naming an attribute: the values it copied stay in it, and in files exported from it, until you re-extract. A re-extract is what brings a dataset back in line with the current design.

Re-extracting later (as more surveys complete) picks up the new responses and marks the Bundle's Silver and Gold stages stale — they still hold their prior data and settings, but need a one-click re-derive before they can be trusted or exported again.

Data-suspect surveys are left out

A completed survey whose run hit a critical problem — logic in the questionnaire that failed part-way through, leaving the branching that followed to run on incomplete values — is excluded from extraction. Its answers were accepted at the time, but they cannot be relied on, and a dataset is the wrong place to discover that.

Nothing is deleted: the survey stays fully readable in Targetor, where the campaign page lists its degraded surveys and the survey detail page shows exactly where the problem occurred. Fix the questionnaire logic, re-field, and re-extract. If a Bronze row count is lower than the campaign's completed count, this is the first thing to check.

Deriving Silver: Coding and Weighting

Open a Bundle's Silver phase page. Two things happen in order when you click Derive:

1. Coding. Balansor turns the questions you selected for coding into one labelled categorical column per dimension you kept — see Open-End Coding for the mechanics. The Open-End Coding accordion on the same page is where both choices are made: its Questions to code picker lists every open-text question the Bronze offers, and the dimension review below it is where you propose dimensions, read their categories, and tick the ones to keep. Both choices feed the next derive, not the one just run. With nothing selected — because you chose nothing, or the questionnaire asks no open-text question — this step is skipped and weighting runs directly. A question that is selected but whose dimensions nobody has reviewed produces no coded column, and the Silver stage and the export both say so rather than shipping a file that is quietly missing them. Coding that is attempted and yields no column at all is a different case and fails the derive: one question that cannot be coded is left uncoded while the rest proceed, but zero columns out of every question tried stops the stage with a message naming each one, instead of publishing a Silver that looks finished and contains none of the coding you asked for.
2. Weighting. Post-stratification weighting (raking) then runs once, on the coded case base, against the Calibration Targets accordion — see below.

Because coding and weighting can take a little time, Balansor runs this derivation in the background — the Silver node shows "working" while it processes and updates automatically once it's ready (or reports a reason if it failed).

Calibration Targets

Calibration Targets are a per-Bundle spec you edit on the Silver phase page, separate from the Bundle's current Strategy (which grades the Bronze measure instead). They hold what Silver weights toward, and the outside figures you want the sample checked against. The first time you open the Silver page, Balansor seeds Calibration Targets once from the current Strategy's factors; after that, you edit them freely:

• Untick Include on a Strategy factor you don't want in the targets. The current Strategy's factors always stay listed, with the Strategy's targets shown, so you can still see that the design had that margin. Ticking it again weights against the Strategy's own targets, which is also how you reset a Strategy factor whose targets you edited.
• Add any dataset column as a factor — a Bronze demographic or answer column, a multi-select option column, or a coded open-end indicator — with its own target proportions (must sum to 100%). A factor you added can be dropped again.
• Choose each target's role. Every included target is either Raked (the default: Silver weights the sample toward it) or Benchmark only (the weights leave it alone, and Bundle quality grades the sample against it). Use Benchmark only for an outside figure on a survey answer, such as home ownership from a national survey or vote share from an election, or for any column you want checked without weighting on it, respondent attributes included. At least one target must be raked: a list that is all benchmarks is refused, because the derive would have nothing to weight on.
• Copy from another Bundle. Pick any Bundle you can read, and Balansor adds the targets that fit this Bundle, each with its role, to the unsaved list. Targets that do not fit are listed with the reason: this Bundle has no Bronze yet, the column is missing here, a category is not an option of this Bundle's questionnaire, or the target clashes with one already in the list. Nothing is saved until you click Save.
• Add a respondent custom attribute. The picker has a separate Respondent custom attributes group. It lists the attributes this Bundle's own surveyed respondents carry, together with those already in the dataset, and marks each one in the dataset or not in the dataset: needs a Bronze re-extract. Attributes of respondents outside this Bundle are never listed. Naming an attribute here copies its values into the dataset files at the next Bronze extraction.

Reassigning the Strategy elsewhere never re-seeds Calibration Targets — the two are independent once seeded. The new Strategy's factors appear unticked until you tick them.

Two kinds of column are never offered, because neither can be weighted: the bookkeeping columns extraction records around each row (identifiers, status, completion time, response timing, the weight itself), and the raw combination column of a multi-select question — pick its per-option columns instead, one target per option.

Every save checks the categories. A target on a question's answer must name that question's options, by option code or by option label, in the Bundle's questionnaire. A category the question does not offer is refused, and the message names the column and the categories it accepts; this matters most for a benchmark, where a mistyped category would be graded as a miss that no respondent could avoid. If a column's categories cannot be checked, for example a slider or a free-text answer, the targets are saved with a warning saying so.

When the dataset is behind your targets. You can save a target on a custom attribute the dataset does not hold yet; that is how a new attribute enters the analysis. The save succeeds and the page tells you that Bronze needs a re-extract before Silver can be derived. Until you re-extract, a derive that would weight on that attribute is refused with a message naming it. The same card also shows:

• Custom attributes not in the dataset yet: attributes your Strategy or targets name that the current Bronze was not extracted with. One that only the Strategy names (and you do not weight on) does not block the derive; its factor reads as not measurable on the quality measures until you re-extract.
• A quieter note when Bronze still carries an attribute nothing names any more. A re-extract drops it.

Named a new attribute? Re-extract, then derive

Re-extracting Bronze writes every attribute the Strategy and the Calibration Targets name at that moment, then marks Silver and Gold stale for a one-click re-derive. If the refusal names an attribute none of this Bundle's respondents carries, the name is misspelt: a re-extract would only add an empty column, so correct the name in the Calibration Targets or the Sampling Strategy instead.

A target category is the value the column stores, not its caption

Weighting groups the data by the value actually in the column — for a coded answer that is the option's code, not its text; for an option or coded-theme column it is 0 and 1. In the editor you pick a category by its option text and Balansor stores the matching code for you, so the picker cannot get this wrong. A target written straight through the API, or one inherited from an older recipe, can.

If none of a factor's categories occurs in its column, the derive fails and the Silver node reports why, showing your declared categories next to a sample of the column's actual values. It does not quietly drop the factor, and it does not produce a dataset whose weights are all 1.0 alongside a quality score. A factor with some categories present is weighted normally; a declared category that nobody fell into is an ordinary, expected result.

Weighting Results

After a successful derive, the Silver phase page's Quality accordion shows:

Metric	Description
Converged	Whether the algorithm reached stable weights
Iterations	Number of raking iterations performed
Mean Weight	Average weight (should be ~1.0)
Design Effect	Variance inflation due to weighting
Weight Range	Min and max weights assigned
Effective Sample Size	Equivalent unweighted sample size

Interpreting Design Effect

A design effect of 1.5 means your effective sample size is 67% of your actual sample size. Higher design effects indicate more extreme weighting adjustments.

Quality Checks

Quality belongs to the Bundle, not to an individual dataset — click a Bundle's title on the pipeline board to open its Overview page (the Bronze and Silver phase pages also surface their own measure inline). The page follows one sample from the design you intended to the file you deliver, in the order the distortion happens:

1. Strategy → Pool — did sampling achieve the design? One reading per Pool your Campaigns drew on, against the Strategy that generated it. A Pool built by hand has no design to measure against, and says so. The reading uses each member's values as they were when the Pool was drawn, so later edits to a respondent do not move it; a Pool without a complete record of those values (an older, merged or split Pool) reads as not measurable.
2. Strategy → Actual — how far off was the realized sample, before any weighting? Against the Bundle's current Strategy, with a per-campaign breakdown. Reports no strategy until one is assigned.
3. Strategy → Weighted — did weighting recover the design? Against the same Strategy as (2), which is what makes the pair readable: the two differ by exactly what weighting recovered. Its headline is labelled Weighted fit and says how many Strategy factors it covers: only the factors the weighting targeted are graded, and raking matches those by construction, so a fit near 100% confirms the weighting rather than showing that the sample is representative. A Strategy factor the weighting did not target is left out ("not weighted on"). Read Recovery (the error over every factor, before and after weighting) and the Benchmarks for how representative the weighted sample is; the card's How to read the weighted grade box explains why weighting cannot reach 100% on every factor.

Those three share one referent, so they sit on one scale. A fourth measure deliberately does not:

1. Pool → Actual — what did fielding itself contribute? Measured against the Pools the Surveys came from, over pool-borne Surveys only. It isolates the fielding step, which the three above cannot separate from a Pool that was already off-target — so read it as a diagnostic, not as a fourth point on their scale.

And one measure has no referent at all:

1. Response Quality — are questions producing informative, diverse responses? Normalized Entropy, Straightlining Score, Cronbach's Alpha, Acquiescence Bias. See the Quality Metrics Reference.

The four representativeness measures compare demographics against their targets using RMSE, MAE, Chi-Square, Max Deviation, and a Composite Quality Score — see the Quality Metrics Reference. None of them ever reports a zero in place of a gap: a factor whose data cannot support measurement is named as unmeasurable and left out of the score, because a zero would read as a bad grade rather than as a missing one.

Recovery, and what it cost. Attached to the weighted measure, the Overview shows how much distance from the design weighting closed — measured in error units over the factors graded on both sides, so the figure means the same thing on every Bundle. Immediately beneath it are the design effect and effective sample size: because no respondent is ever discarded to balance a distribution, the price of recovery is variance, and this is where you see it. Below 1.5 is a low cost; 1.5–2.5 warrants caution with small subgroups; above 2.5 the weighted estimate rests on a small effective sample.

Benchmarks. Every Calibration Target set to Benchmark only is graded per category on Strategy → Actual (the unweighted sample) and on Strategy → Weighted, shown beside the Strategy's respondent factors and never folded into their score. Benchmarks are graded whether or not the Bundle has a current Strategy. This is where outside outcome figures belong: a Sampling Strategy describes only the respondents a Pool is drawn from. If a benchmark's targets name no category its column holds, it is shown as unmeasurable ("Targets match nothing") rather than graded as a full deviation.

Raking reached. A raked target is not graded: the weighting aimed at it, so a grade would only say that raking converged. Strategy → Weighted instead shows, for each raked target, the weighted share against the target, labelled raking reached, so you can see how close raking got. A target you switch to Benchmark only after the Silver was weighted on it keeps reading as raking reached on Strategy → Weighted until you derive Silver again, because the current weights were still fitted to it; Strategy → Actual grades it as a benchmark straight away.

The current Strategy is re-assignable from the page header, and reassigning it recomputes all three Strategy-referenced measures. The Overview page can also launch the AI Analyst report — see that guide for how the analyst interprets these metrics.

The measurement base and the analysis base

Two different populations sit behind those numbers, and the page names which one every count refers to.

The measurement base is everything fielding returned — the Bronze data. It is what the fielding measures count, and it can include partly-finished and engine-flagged Surveys if you admit them: the campaign-set dialog's Measurement base toggles decide whether partial and suspect Surveys are extracted at all. Admitting them re-extracts Bronze and re-derives everything downstream.

The analysis base is the subset that passes the Bundle's completeness threshold — its own card on the Overview page, defaulting to 50%. A Survey's completeness is the share of the items it actually reached that it answered. Reached means every distinct question the respondent was shown, including the one on screen when they stopped. Skip logic therefore never counts against a respondent, while someone who answered one question and left the next one open reads one out of two, so the threshold can see an abandoned Survey. A Survey that reached no applicable item has no completeness at all rather than a zero. Changing the threshold re-derives Silver and Gold; Bronze is untouched, so nothing needs re-extracting.

The completeness threshold is one of two cuts you control. The other is whether volunteers stay in the analysis base — the people who took part without being drawn into the campaign's pool. It is off by default, so volunteers are analysed like everyone else unless you decide otherwise, and it cuts at Silver rather than at extraction: Bronze keeps everyone the fielding returned, so the drawn/volunteer split stays reportable whichever way you decide, and changing your mind re-derives Silver and Gold without touching Bronze.

There is no equivalent choice about previews. A designer's test walk or a reviewer's run through a share link is not a case, and nothing admits one into a dataset.

The threshold is a cutoff you introduce, so the page shows what it removed: the excluded set is profiled against the analysis base on your Strategy's own factors. Two row counts cannot tell you whether a cut introduced bias — a threshold that removes 12% is harmless if that 12% looks like the people it kept, and is a bias you created if it does not. When nothing was excluded, the block says so rather than showing a divergence of zero.

Calibration Targets

Weighting doesn't target the current Strategy directly — it targets the Bundle's Calibration Targets, a separate, editable spec that lives on the Silver phase page. Raked targets are the weighting instruction, not the yardstick: the result is still graded against the current Strategy, so raked targets that drift away from your design show up as distance from that design rather than as a perfect score against their own aim. Benchmark-only targets are the opposite: a yardstick the weighting never aimed at, graded on the Overview. The first time you open that page, Balansor seeds it once from the current Strategy's factors; after that, editing the Strategy elsewhere never touches it. From then on you edit it freely: drop a seeded factor, or add any weightable dataset column — a Bronze demographic or answer column, a multi-select option column, or one of the categories discovered by open-end coding — with its own target proportions, or a respondent custom attribute, which reaches the dataset at the next Bronze extraction. See Calibration Targets above for what makes a target valid.

Improving Representativeness

If quality scores are low:

• Check respondent demographics: Missing demographic data reduces matching capability
• Review strategy targets: Ensure targets are achievable with your respondent pool
• Consider broader factor definitions: Combine small categories with few respondents
• Increase sample size: More respondents enable better matching

Refining to Gold

Open a Bundle's Gold phase page to decide what the delivery file contains. The page has three parts: the Column Layout, where you choose, order and name the columns; Recipe, a read-only list of what the layout changes; and Derive, which builds the Gold file.

A layout you have not touched makes the Gold file a copy of the Silver file. Once you change it, the Gold file differs from the Silver only by your changes: every column you did not remove is still there, row-metadata and respondent-attribute columns included, and each kept column keeps its question text and value labels in SPSS, R and JSON exports.

The Column Layout

The layout has two lists side by side.

• Silver columns (left) lists every column of the Silver file, in the file's order. It never changes. A mark shows how many times a column is served (×1, or ×2 and more when it has aliases) and how many computed columns read it (ƒ with a count); a column nothing uses is drawn dashed. Filter the list by name, or tick Unused only. Serve all shown and Un-serve all shown act on the columns the filter currently shows. Filtering changes what you see, never the layout.
• Served columns (right) is the Gold file: these columns, in this order, under these names. A served column is a Silver column the Gold file includes. Each one shows its position, its name, the Silver column it comes from when you renamed it, and a mark for row metadata, respondent enrichment or the weight column. Aliases and computed columns sit in the same list, wherever you put them.

What you can do, each on the column itself:

To	With the pointer	With the keyboard
Serve a column	Drag it from Silver columns into Served columns; it lands where you drop it	Focus it in Silver columns and press Enter or Space; it takes its natural place in the Silver's order
Un-serve a column	Drag it back onto Silver columns	Delete on the column in Served columns; Delete in Silver columns removes every copy of it
Serve a column again under another name (an alias)	Drag an already-served column from Silver columns into Served columns; the alias lands where you drop it	Focus the column in either list and press A; the alias goes right after the column's last copy
Reorder	Drag by the grip	Space picks the column up; arrow keys move it by one, Home and End to either end; Space drops it, Escape puts it back
Move far	Click the position number and type the new position	P, then type the position
Rename	Double-click the name	Enter or F2; Escape cancels
Find a served column	Type in Find a served column; matches are highlighted and Enter steps through them	—

Tab reaches each list; arrow keys move between its columns. Ctrl+Z (⌘Z on a Mac) undoes the last change. The keys are also listed at the bottom of the page.

Several columns at once

Every column in both lists has a tick box. Tick the columns you want, in any order and wherever they are in the list, then act on all of them in one step:

• In Silver columns: Serve selected serves every ticked column at its place in the Silver's order; Un-serve selected removes every copy of each.
• In Served columns: Remove selected removes the ticked columns; Move to position puts them, one after another and in the order they had, at the position you type.
• Dragging a ticked column carries every ticked column of its list: into Served columns they land together where you drop, in the Silver's order; inside Served columns they move together and keep their order; onto Silver columns they are removed.

With the keyboard, X on a focused column ticks or unticks it, and Tab reaches the buttons. A set change is one change: Ctrl+Z puts every column of it back at once. A ticked computed column moves with the others, but Remove selected keeps it, because other computed columns may read it; delete it on its own chip.

Un-serving never deletes anything and never asks for confirmation: the column stays in the Silver list and can be served again at any time.

Aliases

An alias serves a Silver column a second time under another name, for example age and age_at_interview. The Gold file then holds two columns with the same values, question text and value labels. A new alias gets a free name (age_2) that you rename like any other column, and it is marked alias with the Silver column it copies.

• An alias can sit anywhere in the file and is removed on its own with Delete.
• An alias needs its column. If you un-serve the column itself and an alias remains, the first alias takes its place: it stops being an alias and becomes the column, under the alias's name.
• The weight column's alias is a plain copy of the weights. The Gold file's weight column is always the column itself, so analyses and exports keep using it.

Names you give must be unique. A name you type for a served or computed column may not match another column of the Gold file, and names that differ only by letter case (Score and score) count as the same name. An empty name is not allowed, and spaces at the start or end of a name are removed when you confirm it. The columns involved are marked in red and the layout cannot be saved until they are fixed. If the Silver file itself holds two columns whose names differ only by letter case, they can both be served as they are; that is the Silver's own naming and nothing you need to fix. A layout must serve at least one column.

Computed Columns

A computed column is a new column calculated from others. A new one is added at the end of Served columns; move it anywhere in the file as you move any other column.

To make one, drag the columns it reads onto the New computed column area, or press the New computed column button and pick them from the list. Give it a name and choose what to compute. The number of columns you picked decides which kinds can be selected:

Kind	Needs	Effect	Example
Arithmetic	Exactly two columns	+, -, * or / between them	total = price * quantity
Recode into buckets	Exactly one column	Each bucket turns the values from its minimum to its maximum (inclusive) into its label. A bucket's minimum may not be above its maximum	Age → 18-34, 35-54, 55+
Concatenate	One or more columns	Joins them in order with a separator	full_name = first_name + " " + last_name

• Arithmetic and recode read each value as a number. A value that is not a number gives an empty result for that row, as do a division by zero and a value outside every bucket.
• A computed column reads the Silver columns under their Silver names, whether or not you serve or rename them. You can rename age to Age, or leave income out of the file, and still compute from either.
• It can also read a computed column that was made before it. Where the two sit in the file does not matter: moving a computed column never changes what it can read.
• It cannot read an alias; read the Silver column the alias copies.
• On a computed column, Enter opens its form and Delete deletes it.
• Its name must differ from every Silver column name, including columns you do not serve.
• Edit on a computed column (or Enter) reopens the form; Delete (or the Delete key) removes it. If a computed column reads something that no longer exists, it is marked as broken with the missing input named, and the layout cannot be saved until you fix or delete it.

Saving the Layout

Nothing is stored until you press Save layout. The bar at the bottom of the page shows whether you have unsaved changes and offers:

Button	Effect
Undo	Takes back the last change, one step at a time
Discard changes	Returns to the saved layout
Reset layout	Serves every Silver column in file order under its own name and removes every computed column. It asks first, and Undo brings your layout back
Save layout	Stores the layout. Available only when something changed and nothing blocks it

Saving a layout marks an existing Gold as stale, because that Gold was built from a different layout, and marks the Bundle's Study as stale if it is part of the Research Paper. That is why Save is unavailable when nothing changed.

When someone else saved first

If the saved layout changed after you opened the page (a colleague or an AI assistant saved one), or the Silver was derived again in the meantime, your save is declined with a message. Your arrangement stays on screen so you can note what you did; Load the saved layout replaces it with the current one.

When the Silver lost a column

A saved layout can name a column that a later Silver no longer has. The page then opens with a notice naming those columns, sets them aside, and shows unsaved changes: press Save layout to store the layout without them.

Recipe lists the operations your layout stands for (remove, rename, order, compute), exactly as Save would store them. It is the same recipe an AI assistant or the API reads and writes (see create_gold_dataset). It describes the layout, not how the current Gold file was built.

Deriving Gold

Derive Gold builds the file from the Bundle's ready Silver and the saved layout. While you have unsaved changes the button is unavailable: save or discard them first.

• If the saved layout removes or orders a column the Silver no longer has, the build goes ahead without that entry and tells you which entries it skipped.
• If a rename or a computed column refers to a missing column, the build is declined with a message naming it. Open the layout, fix it, save and derive again.
• If the layout is saved again while a Gold is being built, that Gold is discarded and you are asked to derive again, so a ready Gold always matches the saved layout.

Reading Without Editing

Changing the layout and deriving are open to a manager who owns the Bundle's Project, and to administrators. Anyone else who can read the Bundle (an analyst, a manager who does not own the Project, or anyone in a sealed organization such as the demo) sees the saved layout as a plain numbered list of the columns the next Gold would hold, with the same marks and the stored recipe, and no editing controls.

Exporting Data

Export from Any Ready Stage

Every stage page (Bronze, Silver, Gold) ends with an Export section for that stage, closed until you open it. Its header shows the stage's live status, row count and column count, driven entirely by the database (never a file or directory scan). Opened, a ready stage shows the format buttons, what the stage changed against the one before it, and a preview of its first rows (lock any column to keep it in view while scrolling); a stale (invalidated), still-processing, or errored stage shows its status instead, with no export control — re-derive it first.

Stage	Use Case
Bronze	Raw data for custom analysis or archival
Silver	Coded and weighted data for statistical software
Gold	Final deliverable for clients

Export Formats

Format	Extension	Best For
CSV	.csv	Universal compatibility, Excel, Python, R
Excel	.xlsx	Microsoft Excel with formatting
SPSS	.sav	SPSS/PSPP with variable labels
Parquet	.parquet	Big data pipelines, efficient storage
R	.rds	R — categoricals arrive as factors, value labels as factor levels
JSON	.json	Scripts and AI assistants — the rows plus each column's question text and value labels, in one file

Every export format is available on every plan

CSV, Excel, SPSS (.sav), Parquet (.parquet), R (.rds) and JSON (.json) export are open to every authenticated user. An export is declined when the dataset is not ready (re-derive the stage and try again), or when it is a Silver whose coded questions still await dimension review (review them, re-derive, then export). The same rules apply whether you export here, through an AI assistant, or through the API.

Export Process

Click a format button in a ready stage's Export section to export and download it — each button exports that stage in that format immediately, with no separate options dialog. If a ready stage's underlying file has gone missing on disk (a rare data-integrity issue), the export fails loudly with an explicit error rather than downloading a broken file — re-derive the stage to fix it.

Workflow Examples

Standard Survey Analysis

1. Create a Bundle: choose the project, questionnaire, and campaigns
2. Extract: create the Bronze stage from completed surveys
3. Derive Silver: code any open-ended questions, then weight against your Calibration Targets
4. Refine: arrange the Gold Column Layout — order and rename the key columns, add computed columns for derived metrics — then save and derive
5. Check quality: open the Bundle Overview or the Silver phase page's quality accordion — compare Bronze vs Silver representativeness
6. Export: download Gold as SPSS for client delivery

Comparing Variants

1. Build out a Bundle through Silver
2. Clone the Bundle — the clone starts with an identical Bronze and the same recipes
3. In the clone, change the weighting targets (or the refine recipe) and re-derive
4. Compare the two Bundles side by side on the board

Integration

MCP Integration

For AI-assisted data analysis, the Bundle pipeline is available through MCP (Model Context Protocol) tools covering Bundle lifecycle, Bronze extraction, coding + weighting, Gold refinement, quality metrics, and export.

See the MCP Interface Reference for the complete list of available tools and their parameters.

Best Practices

Bundle Management

• Name Bundles descriptively: Include the wave or purpose, not just the questionnaire name
• Archive, don't delete: A Bundle's Bronze is your foundation — clone before experimenting rather than mutating a shared Bundle
• Re-extract deliberately: Re-extracting invalidates downstream stages; do it when you actually have new completions to fold in

Weighting Strategy

• Define Calibration Targets before collection: Better to weight to pre-defined targets than to retrofit them
• Avoid extreme weights: Weights > 5.0 or < 0.2 indicate sampling problems
• Check convergence: Non-convergent raking suggests incompatible targets
• Review effective sample size: Significant reduction indicates over-correction

Quality Assurance

• Always check quality metrics before delivering weighted data
• Compare Bronze vs Silver before and after weighting
• Re-weight if targets change by editing Calibration Targets and re-deriving Silver, rather than editing exported files

Export Strategy

• Match format to recipient: SPSS for statisticians, Excel for clients
• Use consistent naming conventions across projects
• Only export ready stages — a stale or working node has nothing trustworthy to hand out yet

Troubleshooting

Common Issues

"No strategy assigned"

The Bundle has no current Strategy, so the Bronze quality measure reports no_strategy until one is assigned:

• Create a strategy in Targetor
• Assign it from the Bundle Overview page's header (the "Assign Strategy" control)

This does not block weighting — Calibration Targets can be edited manually on the Silver phase page even with no current Strategy assigned.

"Weighting factor … matches no value in column …"

The derive stopped because one factor's target categories do not occur in the column it names. The message puts both sides in front of you — the categories you declared, and a sample of what the column actually holds — so the mismatch is usually readable at a glance ("your categories are 'Sales' and 'Engineering'; the column holds 2, 3, 5, 6" means the targets were written as captions where the column stores codes). Almost always one of:

• The targets were written as captions where the column stores codes (or the other way round) — open the Calibration Targets editor, which shows both side by side, and re-enter the categories.
• The factor names a column from a different questionnaire version or a renamed item.
• The Bronze was re-extracted and the column's coding changed underneath the target.

Fix the factor (or drop it) and re-derive. A factor whose categories only partly occur is fine and is not what this reports.

"Silver would be weighted on the custom attribute …, which this Bronze was not extracted with"

The Calibration Targets (or, while the targets still follow the Strategy, the Strategy's weighted factors) name a respondent custom attribute that was not part of the design when Bronze was extracted, so the dataset has no column for it:

• Re-extract Bronze, then derive Silver. The re-extract writes every attribute the design names now.
• If no respondent of this Bundle carries an attribute by that name, the name is wrong. Correct it in the Calibration Targets or the Sampling Strategy; a re-extract would only add an empty column.

Raking did not converge

The algorithm couldn't find weights that match all targets simultaneously:

• Check if targets are mathematically achievable
• Review factor correlations (e.g., age-gender combinations)
• Consider relaxing target precision
• Verify sufficient respondents in each category

A failed derive leaves no weights behind

When a derive fails, the Silver stage keeps its previous file but drops every claim about it — the weight flag, the weighting timestamp and method, and the quality score all clear, so the node reports the failure and nothing else. It will never show a score earned by an earlier run next to an error saying this run produced nothing. Fix the cause and re-derive.

Low quality scores

Weighted data doesn't closely match targets:

• Check for missing demographic data in respondents
• Review target distributions against available population
• Consider if factors are too granular for sample size

Export is unavailable on a node

The stage is stale or still processing — re-derive it first. Export only appears once a stage is ready.

Missing demographics in Bronze

Extracted dataset lacks demographic columns:

• Ensure demographic enrichment was enabled during extraction
• Verify respondents have demographic data in Targetor
• Re-extract with the option enabled

Next Steps

• Campaign Management

  ---

  Design campaigns with sampling strategies and respondent pools

  Guide

• Create Surveys

  ---

  Design questionnaires with QML and SMT validation

  Guide

• Execute Surveys

  ---

  Run surveys with dynamic flow and lazy evaluation

  Guide

• API Reference

  ---

  Integrate programmatically via REST or MCP

  REST API