CBC Shop: Enhanced Choice-Based Conjoint

1. What Is CBC Shop?

CBC Shop is a Choice-Based Conjoint study presented like a real store. The respondent answers the familiar question, which of these would you choose? Each alternative is still a bundle of attribute levels, the choice is recorded exactly as in traditional CBC, and the utilities are estimated with Hierarchical Bayes on the same path as CBC. What changes is the presentation: the alternatives appear as an e-commerce list, as gallery cards, or as products standing on a shelf with a price label.

The reason is realism. A task that looks like the place where the product is really bought carries context that a bare grid leaves out. For packaged goods, CBC with realistic store-shelf displays is "a robust approach" (Orme, 2020, p. 48), although virtual shopping environments sometimes pay off in better data and sometimes do not (Orme, 2020, pp. 35-36).

2. When to Use It (and When Not)

Use CBC Shop when the look of the buying context plausibly changes how people choose: products browsed as online listings, goods picked from a shelf, or categories where the product photo and the price tag drive the decision. If a plain card grid already captures the decision, traditional CBC is simpler and statistically just as good.

3. Opening the Editor

The browser tab reads "CBC Shop Configuration", and the left rail carries the badge "CBC Shop". On a Free account a project can hold at most two conjoint tasks.

The left rail

What saves automatically and what needs a click

The two locks

• A design exists. The Attributes step shows "Attributes are locked. Delete the generated design before changing attributes or levels. Your existing configuration will be kept." with the button Delete generated design to edit attributes. Prohibitions are locked too. The Shop Display, the question, images and fixed-task definitions stay editable, because they do not change the design.
• Responses exist. A banner at the top reads "Design locked - responses already collected" with the number of active and trashed responses. To change the design, open Data > Bin and permanently delete all responses, or duplicate the project to keep them. Preview, downloads and design diagnostics stay available.

The editor can also show "Design Warning: ..." when the stored design no longer matches the configuration, and "Regeneration required. This generated design conflicts with active prohibitions and cannot be published until it is regenerated."

4. General

No.	Control	What it does
1	Task Name *	Internal name and column prefix in your data (cbcs_pizza_oven). Up to 50 characters, start with a letter, then letters, numbers, underscores and hyphens.
2	Question / Instruction *	Rich text shown above every task. The default reads "If you were buying a new product and these were the only alternatives available, which one would you choose?". Up to 10,000 characters and 10 inline images.
3	Footer	Optional rich text below the task (up to 10,000 characters).
4	Attribute Display	Capitalize attribute names (UPPERCASE), off by default.
5	Image Options	Enable image zoom on click and Show labels under images (only applies to level images). Both off by default.

Unlike traditional CBC there is no Layout Style and no Option Header here: the whole look is set on the Shop Display step (section 6).

5. Attributes, Price Attribute and No-Choice Option

Attributes

No.	Control	What it does
1	Attributes are locked.	Shown while a generated design exists. The orange button deletes the design so you can edit attributes again.
2	Add Attribute	Opens the Add Attribute dialog (Figure 3).
3	Drag handle	Changes the order of the attributes.
4	Hide / Show	Hidden attributes stay in the design and are estimated, but are not displayed. Use it when the product photo already shows the attribute.
5	Edit	Edit Attribute dialog: name and levels as small rich-text editors (without images), levels reorderable, Shift+Enter for a line break, Save changes.
6	Delete	Deletes the attribute after confirmation.

No.	Control	What it does
1	Attribute Name	Up to 50 characters.
2	Levels	One row per level (up to 50 characters each), with a counter badge and a bin to delete a row. In this dialog levels cannot be dragged; reorder them later in the Edit dialog. Enter jumps to the next level.
3	Add Level	Adds a level row. At least 2 levels are required.

The same server rules as in CBC apply: 2 to 20 levels per attribute, at most 20 attributes, names unique ignoring capitalization. Free plan: at most 4 attributes with at most 4 levels each (the five-attribute Pizza task needs a paid plan).

Price Attribute and No-Choice Option

No.	Control	What it does
1	Price Attribute	"-- No price attribute --" or one of your attributes. The chosen attribute is shown as a price (large, styled with Price Size and Price Color) instead of an ordinary attribute line, and gets the "Price" badge. The Shelf layout needs it for its price labels. Saved as soon as you change it.
2	Enable "None of these" Option	Adds a None choice to every task. Off by default for a new CBC Shop task (on in the Pizza task).
3	Button Text	The label of the None choice, "None of these" by default (up to 200 characters).
4	Expected None Share (Optional)	Optional value from 0 to 1 (for example 0.20). Metadata for the design diagnostics only; it does not change the generated tasks.

Below these fields, Save No-Choice Options stores the None settings. CBC Shop has no Position control for the None option; in the shop layouts it appears as a wide button below the products (Figure 19).

6. Shop Display

The Shop Display step holds the whole presentation. It changes what respondents see, never the design or the data the model receives. Save it with Save changes at the top right.

No.	Card	What it does
1	Shop Theme Preset	Five one-click looks (section 6.1).
2	Main Layout	Shop Horizontal, Shop Grid or Shelf (section 6.2).
3	Attribute Display	How the levels read inside a product card (section 6.4). Hidden on Shelf.
4	Online Shop Settings	Products per row, card and button style, button text, image behaviour (section 6.5). Shop layouts only.
5	Selection & Interaction	Accent colour and how a product is selected (section 6.6).
6	Typography & Price Display	Collapsed by default: text and price styling (section 6.7).
7	Card Styling	Collapsed by default: corners, spacing and background of the cards (section 6.8). Hidden on Shelf.

On Shelf, a Shelf Settings card (section 6.3) appears directly below Main Layout, and the cards that only make sense for online shops disappear. Table 3 shows what is visible in which layout.

6.1 Shop Theme Preset

"Quick-apply a complete shop look including layout. Customize individual settings afterwards." A preset sets the layout and a group of styling options at once. It never changes the typography (font size, weight, text colour), Image Interaction, the checkmark or the click-anywhere setting.

6.2 Main Layout

There is no separate sub-layout control. Pick the layout that matches the channel in which your product is really bought.

6.3 Shelf Settings

Two notes appear in the card. The blue note recommends product images, because a shelf shows products, not cards: products without a matching rule under Product Images get a neutral placeholder. The yellow warning names every ticked attribute whose first level is longer than 14 characters ("Stone + pizza peel" in the example); the label shortens such text with "...", as the label preview shows. Shorter level names or fewer ticked attributes avoid it.

Messages on this card:

• "Product images are strongly recommended for Shelf. Products without a matching rule under Product Images use a neutral placeholder." Shelf does not use level images.
• "No price attribute defined!" when no Price Attribute is set. The shelf needs it for its labels; set it in the Price Attribute section of the Attributes step.
• "X is not shown on the price label. If the product image shows it, hide the attribute." for unticked attributes.
• "Levels for "X" may be truncated." when the first level of a ticked attribute (other than the title) is longer than 14 characters.

The Preview panel next to the list redraws one price label as you tick attributes. For the full shelf with products, generate a design and open the Preview step.

6.4 Attribute Display

Decides how the levels read inside a product card (Shop Horizontal and Shop Grid; hidden on Shelf).

Levels-only formats look most like a real listing, but only work when the levels explain themselves. "3.9 ★" does; "420 °C" is clearer with its attribute name.

6.5 Online Shop Settings

Shown for Shop Horizontal and Shop Grid, under the heading Product Grid.

6.6 Selection & Interaction

Every product must stay selectable. If you switch off click-anywhere and hide the button at the same time, saving fails with "Each product must remain selectable. Enable click-anywhere selection or show a visible Select button." Keeping both on is the safest combination on phones.

6.7 Typography & Price Display

Price is usually the strongest driver of choice, so present it as it is presented in the real channel: a discounter shows large, coloured prices, a premium store smaller ones.

6.8 Card Styling

7. Fixed Tasks

A fixed task shows the same products to every respondent. Choicery leaves fixed tasks out of estimation and uses them as holdouts for hit rates.

No.	Control	What it does
1	Fixed Tasks + badge	"Applied (N rows)" once the fixed tasks are placed in the design (2800 rows here), "Draft (not applied)" before.
2	Add	Opens Create Fixed Task (Figure 8). At most 20 fixed tasks.
3	Show N alternatives	Expands the card. Pencil edits, bin deletes.

No.	Control	What it does
1	Number of Alternatives	2 to 9, default 3. A fixed task may show more products than the random tasks, for example a full shelf of 9.
2	Randomize All	Fills every selector with a random level; the shuffle icon of an alternative fills only that one.
3	Save Fixed Task	Enabled once every attribute of every alternative has a level.

Without fixed tasks the step warns that hit rates cannot be calculated. Positions are set on the Generate step (section 12). Holdouts with many alternatives are "more realistic than using smaller choice sets" (Rao, 2026, p. 261), and one holdout is not enough (Orme, 2020, pp. 191-192; see section 15).

8. Product Images

Product images give each product its photo, which the shop and shelf layouts are built around. A rule shows an image when a concept contains a given combination of levels.

No.	Control	What it does
1	Conditional Product Images	The list of rules: name, conditions, image file, and icons to edit, switch on or off, and delete.
2	Add Rule	Opens the Add Product Image dialog (Figure 10). At most 30 rules.

No.	Control	What it does
1	Rule Name	Up to 80 characters.
2	Trigger Conditions	Attribute and level, then Add; at least one condition. Conditions form groups with an AND/OR switch; New condition group adds a group, groups are combined with AND; Preview shows the expression.
3	Display Content	Content Type Image (Choose Image... from the library) or Text (up to 1,000 characters). For CBC Shop use Image: the product photo is what the shop and shelf layouts show.

9. Level Images

No.	Control	What it does
1	Library	Opens the project image library.
2	Add	Assigns an image to this level; it can later be changed or removed.

"Replaces the level text with a small inline image. For main product images use Product Images. Not used in Shelf." Image files may be PNG, JPG, JPEG, GIF or WebP, up to 5 MB and 4100 x 4100 px each (GIF up to 2 MB and 1024 px); up to 100 images and 200 MB per project.

10. Prohibitions

No.	Control	What it does
1	Add Prohibition	Opens the dialog in Figure 13. Disabled without attributes and while a generated design exists.

No.	Control	What it does
1	Name	Required, for example "No touch screen at $69".
2	Conditions	Attribute and level, then +. At least 2 conditions, each attribute only once per prohibition. Confirm with Add.

Prohibitions can be switched off, edited or deleted; at most 50 per task. After a change the editor reports fixed tasks removed from their positions or a design that must be regenerated. Use prohibitions "sparingly or not at all" (Orme, 2020, pp. 53-54): they almost always lower efficiency (Orme & Chrzan, 2017, ch. 3). A realistic shelf makes implausible products stand out, but a combined attribute is often the better fix than a prohibition.

11. Templates

No.	Control	What it does
1	Save as Template	Template name (required), Description (optional), Save.
2	Available Templates	Your templates with Apply, rename and delete, a count and a storage summary.

A template stores attributes and levels, hidden attributes, level images, prohibitions, product image rules, the General settings, the Shop Display settings, the No-Choice settings and the fixed tasks. Templates work across CBC, DR-CBC, BW-CBC and CBC Shop. The confirmation names only attributes, images, prohibitions and product images, but applying replaces all stored parts. Applying is refused while a generated design exists.

12. Generate

Before generation

No.	Control	What it does
1	Versions	Different questionnaires; one per respondent, reused when there are more respondents. 1 to 1,000, default 200.
2	Tasks	Number of random tasks per respondent; fixed tasks come on top (10 + 2 = 12 screens in the Pizza task). 1 to 30, default 10. Free plan: at most 6 including active fixed tasks.
3	Concepts	Products per task, without None. 2 to 9, default 3. Free plan: at most 3.
4	Method	Balanced Overlap (default, recommended): level and pair balancing with moderate overlap. D-Optimal: coordinate-exchange search for a low D-error under zero priors; slower. Random: no optimization.
5	Generate Design	Starts generation with a progress panel and Cancel; at most 300 seconds.

There is no seed field: every generation draws a new random seed. The editor warns when Concepts exceed the levels of an attribute (repeated levels are then unavoidable). The shop layout plays no part in the design.

After generation

No.	Control	What it does
1	Show Design Diagnostics	Opens the diagnostics (section 13).
2	Generate Design	Generates again after the confirmation "Replace existing design?". Blocked once responses exist.
3	Delete design	Deletes the design and unlocks attributes and prohibitions. Disabled once responses exist.
4	Fixed Task Positions	A Pos: selector per fixed task, from 1 to the total number of tasks; each position once.
5	Apply Positions	Inserts the fixed tasks at those positions.
6	Remove from Design	Takes the fixed tasks out of the design; their definitions stay.

The status bar shows versions, random and fixed tasks, concepts, method and date, with buttons for the design report and the design CSV.

13. Design Diagnostics

Below the tiles: the number of prohibitions applied, the note that fixed-task rows were excluded from the efficiency calculation, and the sections Efficiency Details (D-efficiency, A-Efficiency Ratio, parameters, No-Choice ASC, Expected None Share), Per-Level SE Proxies (target about 0.05 or less for main effects, Orme & Chrzan, 2017, ch. 3), Level Balance, Attribute Association, Two-Way Level Frequencies, Overlap by Attribute and Concept Slot Balance (whether levels are spread evenly over the product positions).

14. Preview and the Respondent View

No.	Control	What it does
1	Design version	Picks the version; the button next to it reloads. The preview below is the real participant page ("Preview Mode. Responses will not be saved."), so you click through it with Next.

Before a design exists the step shows an example preview. Check the preview on a phone as well: Shop Grid and Shelf get crowded on small screens.

The recorded data is identical to traditional CBC: the shown levels, the chosen product and the products it beat. The order in which the products were displayed is stored as well, which allows the optional display-position control in estimation (section 16).

15. How Large Should Your Study Be?

The editor accepts up to 20 attributes with 20 levels each, 30 tasks and 9 concepts. The literature recommends far less. CBC Shop is standard CBC, so the CBC recommendations apply.

Where the sources disagree

• Maximum number of attributes. 6 to 8 (Orme, 2020, p. 50); four or fewer for CBC in the chapter by Orme in Rao (2026, p. 527); concern only at 15 or more (Orme & Chrzan, 2017, ch. 7).
• Number of tasks. 8 to 12 in practitioner sources (Orme, 2020, p. 109); in standard CBC predictive validity can peak well before the last task (Sablotny-Wackershauser et al., 2024, Web Appendix H, pp. 139-140).
• Sample-size threshold. 500 as the minimum with 1,000 recommended (Orme, 2020, p. 64) versus 1,000 (Orme & Chrzan, 2017, ch. 8).
• Holdouts. "Usually 2 or so" in older ratings-based practice (Rao, 2026, p. 122) versus many more in the simulation of Orme and Chrzan (2017, ch. 6).

Shop-specific evidence

• Show products as buyers see them in the market; mixing graphic and text attributes can bias choices toward the graphic ones (Orme, 2020, p. 50). In the Pizza task the photos carry Controls and Included accessories, the text carries the rest.
• Screen position matters: without randomized order the first of three concepts was chosen 39.6 % of the time and the third 24.5 % (Mao, Kessels & Mee, 2025, p. 2). This was tested with three concepts, not on shelves. The Estimate Display Position Effects option (section 16) and Concept Slot Balance address it.
• When respondents do not know market prices, they learn them during the tasks and the None share grows; giving typical prices before the tasks removed this effect (Hung, Kurz, Bailey, Huber & Allenby, 2025, pp. 11-12).

16. Estimating & Interpreting Results

Estimation is standard CBC. Open Conjoint Analysis, select the CBC Shop task and use Preference Estimation (HB).

One control exists only for CBC Shop: Estimate Display Position Effects adds control terms for the displayed product order (Display Position 1, 2, ...), which separates position effects from attribute utilities. Everything else (Run Name, Burn-in, Draws/Chain, Variable Coding with linear price, covariates, In-App and Create Package) works as described in the Preference Analysis (HB) guide. Reading the results (convergence, importances, utilities) is described in the CBC Traditional guide, section 16.

With utilities in hand, the Market Simulator answers "what if?" questions: Market Simulation, Sensitivity Analysis, Willingness to Pay, Price Elasticity and Reservation Price.

17. Best Practices & Pitfalls

Do

• Match the layout to the real channel: Shelf for in-store, Shop Horizontal or Shop Grid for online.
• Set the Price Attribute first; Shelf cannot label its products without it.
• Add Product Images rules for every combination that defines the photo before you tune the look.
• Keep 6 to 8 attributes at most, 8 to 12 random tasks, and plan at least 300 respondents.
• Add more than one fixed task as holdout.
• Preview on desktop and on a phone, and pilot the survey before fielding.

Do not

• Do not let styling hide the attributes. If respondents cannot read the levels, the choice is noise.
• Do not hide both the Select button and click-anywhere selection (the editor refuses to save it).
• Do not add prohibitions liberally.
• Do not copy a recognisable retailer or brand.
• Do not expect CBC Shop numbers to differ in kind from CBC; they are the same utilities from a more realistic task.

18. Key Definitions

19. Ecological Validity & Realistic Presentation

The idea that a study's stimuli should resemble the natural environment of the behaviour under study goes back to Egon Brunswik, who argued for representative design and coined the notion of ecological validity (Brunswik, 1956). Applied to choice experiments: preferences are measured through a task, and the more that task resembles the real decision, the more the measured preferences resemble real ones.

Stated-choice methodologists make the same point. Louviere, Hensher, and Swait (2000) stress that the realism of the choice context shapes how well stated choices transfer to the market, and Lancsar and Swait (2014) frame the external validity of discrete choice experiments around whether survey choices predict real ones. A central threat to that transfer is hypothetical bias: choices made with nothing at stake can drift from choices made in a real store. Presentation realism is one lever against it.

A second, stronger lever is incentive alignment: giving respondents a real chance to receive what they choose, which improves out-of-sample prediction (Ding, Grewal, & Liechty, 2005; Ding, 2007). In meta-analyses of the authors' own studies, incentive alignment and adaptive designs both improved prediction (Sablotny-Wackershauser et al., 2024, Web Appendix H, pp. 118-120). CBC Shop works on presentation only; it does not put anything at stake by itself. Where an incentive-aligned study is impractical, a realistic presentation closes part of the same gap.

20. The Shared CBC Engine

CBC Shop rests on the same random utility theory as every CBC-family method (McFadden, 1974). Respondent h assigns to profile j a utility equal to the sum of its levels' part-worths plus a random error, and chooses the profile with the highest total utility. If the errors are independent extreme-value, the choice probability takes the multinomial logit form:

Individual part-worths are estimated with Hierarchical Bayes, drawing each respondent's β_h from a population normal and borrowing strength across respondents (Lenk, DeSarbo, Green, & Young, 1996; Train, 2009). Effects coding, the zero-centered-diff rescaling and attribute importances are computed exactly as in CBC. Choicery has no separate estimator for the "Shop" label; the shop layout sits on top of the standard CBC/HB path.

For the full derivation (random utility model, Gibbs sampler, effects coding, zero-centered diffs, importances, experimental design), see Part II of the CBC Traditional guide.

21. References