Image-Based Question
An image-based question lets respondents choose from visual answer options instead of plain text, making a quiz faster and more engaging to complete.
Key takeaways
- Each tile carries the same stored value and point weight as a text answer.
- Images help only when the options look genuinely different from one another.
- Captions under every tile keep the question usable without loaded images.
- Uncompressed assets erase the speed advantage the format was chosen for.
- Abstract choices such as budget bands have no useful picture to show.
In depth
An image-based question renders each answer option as a tile: a picture or icon with a short caption beneath it, wired to the same answer record a text option would use. When the respondent taps a tile, the stored value and its point weight are written to the response exactly as a radio button would write them. The image is a label, not decoration. The layout is usually a grid that reflows to one or two columns on narrow screens, and selection is shown with a border or overlay rather than a checkbox.
What makes the format work is visual distinctiveness between the options. When four tiles show clearly different things, such as a spreadsheet, a shared inbox, a dedicated tool and nothing at all, the choice is faster than reading four sentences. When they show four generic stock photos of people at desks, the pictures cost loading time and carry no information, so respondents fall back to reading the captions anyway. File weight is the other lever: uncompressed photos delay first paint and remove the speed the format promised.
The format fits questions whose answer is genuinely a thing you can show: a product tier, a layout style, a room type, a report format, a current tool. In a scorecard quiz, plan-tier mockups let a prospect self-select their size, and the chosen tile carries the point value that decides which tier they land in. Keep a caption on every tile so screen readers and slow connections still convey the option, and keep the tile count low enough that the grid does not scroll.
Images stop helping as soon as the distinction becomes abstract. Budget bands, agreement levels and time horizons have no natural picture, and inventing one forces the respondent to decode a metaphor before answering. The format also raises production cost: every option needs an asset, and changing an answer means commissioning a new one, which quietly discourages iteration. On very slow connections a tile grid renders as a row of empty boxes. Text answers with one supporting illustration are often the better compromise.
Example in practice
How to measure it
Compare drop-off at the image question with the neighbouring text questions in the same quiz. Divide the number of people who answer the step by the number who reach it, and see whether that step-level completion sits above or below its neighbours. Time on step is the second signal. If the image version takes longer than the text question it replaced, the tiles are not reading at a glance.
Watch the distribution of selections as well. If one tile takes an outsized share, check whether it is genuinely the most common answer or simply the most attractive image, or the first position in the grid. Rotating tile order across sessions separates the two. Finally, track whether leads from each tile behave differently later; a tile that predicts nothing is a question worth cutting.
Common mistakes
The frequent failure is treating the picture as decoration and the caption as the real answer. Teams pick four attractive stock photos that all show office scenes, respondents read the captions anyway, and the page is now slower for nothing. Before building, check whether someone could pick the right tile with the captions hidden. If they could not, the images are carrying no information and a plain text list will perform better.
The second is shipping full-resolution photography straight out of the design file. A tile displayed at two hundred pixels wide does not need a two thousand pixel JPEG, and four of them on a mobile connection push the question past the point where people wait. Export at the displayed size, serve compressed formats, and test the question on a throttled connection rather than on the office network.
Frequently asked questions
When should I use an image-based question instead of text?
Use images when the choice is visual or product-related, such as picking a plan tier, a style, or a use case. They also help on mobile-heavy traffic where reading long text labels lowers completion.
Do image-based questions still affect lead scoring?
Yes. Each image option carries its own score value just like a text answer. The respondent's selection feeds the scorecard and routes them into the correct result bucket.
What is the biggest mistake with image questions?
Using purely decorative images that all look similar forces respondents to guess. Make each image visually distinct and clearly tied to the answer it represents.
How many image options should one question have?
Two to six tiles is the practical range. Below two there is no real choice, and above six the grid starts to scroll on a phone, which removes the scanning advantage because the respondent can no longer see every option at once. If you need more options, split the question into two steps or group the images into categories first.
Do image-based questions work well on mobile?
They usually work better on mobile than text options, because a tile is a much larger tap target than a radio button and a two-column grid fits a phone screen naturally. The risk is file weight rather than layout. Export images at the size they are displayed, serve compressed formats, and test on a throttled connection before launch.
Should image answers still have text labels?
Yes. A caption under each tile keeps the question answerable when an image fails to load, gives screen readers something to announce, and removes ambiguity when two pictures look similar. Treat the image as an accelerator for the label rather than a replacement for it, and keep the caption short enough to read in a glance.
Where should the images for quiz answers come from?
Product screenshots, simple icons and your own photography usually beat stock libraries, because they show the actual thing the option refers to. Stock photos tend to be generic and interchangeable, which is exactly what breaks the format. Illustrated icons in one consistent style work well when the option is a category rather than a physical object.
Can images bias which answer people pick?
They can. A brighter, larger or more appealing image attracts selections regardless of what it means, and the first tile in a grid gets a positional boost. Keep the images consistent in style, crop and brightness, and randomise tile order if you suspect the effect. Comparing selection shares against a text version of the same question exposes it.
How do image questions affect quiz load time?
Each tile adds a request and bytes, so a four-option image question can weigh more than the rest of the quiz page combined. Preload the next step's images while the respondent answers the current one, serve modern compressed formats, and cap the displayed dimensions. A question that appears as empty boxes costs more completions than plain text ever would.