Walk me through how you'd apply conditional formatting to a chart, and how you decide the thresholds.

Select the widget, enable conditional formatting in the properties panel, choose the measure that drives the colour, then define bins with their threshold values and assign a colour and a readable label to each. The technical part is trivial; the thinking is in the thresholds. I set them to real business boundaries — quota, SLA, the margin floor — not round numbers, because a threshold that matches a decision boundary makes the colour actionable. I also rename the bins to Low/Medium/High or the team's own wording, since those labels appear in the legend and tooltips and make the widget self-documenting.

When would you use a gradient instead of discrete bins?

Bins when there's a genuine cut-off the reader needs to see — above or below quota, inside or outside SLA. The hard colour change communicates which side of the line a value sits on. A gradient when I'm showing relative magnitude with no meaningful boundary, like revenue across regions, where inventing a threshold would imply a rule that doesn't exist. My rule of thumb: if I can state the threshold in a sentence, use bins; if I'm making a number up to justify one, use a gradient.

Why does a toggle or list filter widget need a step, and what do you group it by?

Filter widgets don't generate values on their own — they display whatever a step's query returns. So the workflow is always step first, then drag the widget and bind it. For dates I group by a date part or a custom bucket field from the recipe, never the raw timestamp, because a raw date has thousands of distinct values and would render thousands of buttons. I also filter the step itself so only the periods anyone cares about appear — curating the buttons is a query-level decision.

A user selects a year on one toggle and the quarter toggle's options mostly vanish. Explain what's happening and how you fix it.

That's faceting. By default, every widget's step is filtered by selections made in other widgets, so the quarter step re-queries with the year selection applied and only returns quarters present in that year. Great behaviour for charts, wrong for filter widgets — a filter should always show the full set of choices so the user can change their mind. The fix is to uncheck 'apply filters from faceting' on that widget, which stops its step listening to other selections. I leave faceting on for charts and tables, and off for toggles and lists.

How do you decide when a dashboard needs a values table rather than a chart?

Charts answer 'how much' and 'which is biggest'. The moment the question becomes 'which records exactly' — which deals are at risk, which cases breached — you need row-level detail and that's a values table. I set sensible default fields on the dataset so every table starts from a reasonable set of columns, order them identifying first, categorical next, numbers on the right, and use fit-data-to-widget spacing so more rows fit in the space available.

Someone changes a field's label in a dashboard step and it doesn't stick elsewhere. What's going on?

They're in the wrong editor. The Fields panel inside a dashboard step only controls what that query returns and how that one widget displays it. Real field metadata — display labels, number formatting, default columns, hiding fields — lives on the dataset and has to be edited in lens mode. It's a very common source of confusion because both places are called Fields and look similar.

How would you display product images or owner photos in a dashboard?

If the dataset carries the image URL as a field, bring it into a values table as a column and switch on 'show data as image' in that column's properties. CRM Analytics then renders the picture instead of the URL text. Owner photos on a top-performers table are the classic use — faces make a leaderboard instantly scannable. The caveat is that the image must be reachable by the viewer's browser, so anything behind a separate login renders broken for everyone but the author.

What can you do with a custom tooltip that you can't do with the default one?

You control exactly which dimensions and measures appear, and you can add custom text that interpolates step field values, so instead of a stack of label/value pairs you get a sentence about that specific data point. I usually hide the defaults that just repeat the axis. I also use tooltips as inline help — putting the definition of a metric like pipeline coverage right where people get confused kills the recurring email thread and documents the dashboard for whoever inherits it.

What's a chart marker, and when do you use one?

A marker is an annotation pinned to a specific bar or data point, with a label, a longer message, and a colour. It's for the story the data can't tell you itself — the month revenue dipped because a major account churned, or the quarter the methodology changed so a step-change isn't a real trend. I use them sparingly, one or two per dashboard, and only on stable anchored charts, because a marker is tied to a data point and can end up next to the wrong thing if users re-filter or re-sort the chart.

How would you build the KPI row at the top of a dashboard?

A number widget with shortened formatting and conditional formatting for the value and its status, paired with a timeline chart stripped of axes, titles, legends, and gridlines to act as a micro chart showing direction. Timeline rather than bar because it uses a true time axis, spaces points correctly, and supports the projected line. A polar gauge for the one or two metrics with a real target, since it shows position within a range better than a number does. Then a consistent theme across the strip and reduced cell spacing so it reads as one band.

You enabled the projected line on a timeline and it doesn't appear. How do you debug it?

The projection extends from today forward, so the first check is whether the data spans today's date — if the dataset ends last quarter, or a dashboard filter or facet trims the range before today, there's nothing to anchor to and the line silently doesn't render. Then I check there are enough historical points for the projection to be meaningful. I also caveat it in front of leadership: a trend extension is directional, not a forecast model.

How do you connect a set of dashboards so users can move between them?

Link widgets, pointed at saved dashboards, with 'apply selections' enabled so the user's current filters travel with them — otherwise they re-apply the same filters on every hop and the set feels like separate reports rather than one application. I style the links with a background and rounded corners so they read as tabs, keep the strip identical and in the same position on every dashboard, and highlight the current one. For that to work the dashboards need compatible filters on shared datasets with consistent field naming, so I design them as a family from the start.

Is there a way to make an image clickable in CRM Analytics?

There's no image-button widget, but you can compose one. Drop a container widget, give it a background image, then overlay a link widget covering the same area and set both its background colour and its font colour to fully transparent. The link still works and still carries filters — it just renders nothing visible. Both colours matter: leave the font opaque and you get stray text floating over the picture. Since there's no hover feedback, I add an affordance like a border or a caption so it's obvious the tile is clickable.

sponsorCard.titlesponsorCard.desc

Designing Dashboards: Interview Questions

Rehearse common CRM Analytics dashboard design interview questions — conditional formatting, faceting, steps behind filter widgets, tooltips, chart markers, KPI design, and linking dashboards with filters.

Designing Dashboards: Interview Questions

You've reached the end of the Designing Dashboards section — nice work. Before the graded quiz, use this page as a study sheet and rehearse each answer out loud.

Dashboard design questions in interviews are rarely "which menu is it in". They're about judgement: why this chart type, why this colour rule, why this filter behaves the way it does. Try to answer each one in your own words before reading the model answer — that active recall is what makes it stick under pressure.

The two questions that separate a confident candidate from a shaky one are faceting and carrying filters across a link. If you can explain both cleanly — what the default behaviour is, and why you'd deliberately turn it off or on — you'll sound like someone who has actually shipped dashboards. Then head to the Graded Quiz; you need 85% to pass, and your score is tracked in your profile.

When you're ready, continue to the graded quiz.

Interview questions

13 common questions with model answers — rehearse each one aloud.

1
Walk me through how you'd apply conditional formatting to a chart, and how you decide the thresholds.
A
Select the widget, enable conditional formatting in the properties panel, choose the measure that drives the colour, then define bins with their threshold values and assign a colour and a readable label to each. The technical part is trivial; the thinking is in the thresholds. I set them to real business boundaries — quota, SLA, the margin floor — not round numbers, because a threshold that matches a decision boundary makes the colour actionable. I also rename the bins to Low/Medium/High or the team's own wording, since those labels appear in the legend and tooltips and make the widget self-documenting.
2
When would you use a gradient instead of discrete bins?
A
Bins when there's a genuine cut-off the reader needs to see — above or below quota, inside or outside SLA. The hard colour change communicates which side of the line a value sits on. A gradient when I'm showing relative magnitude with no meaningful boundary, like revenue across regions, where inventing a threshold would imply a rule that doesn't exist. My rule of thumb: if I can state the threshold in a sentence, use bins; if I'm making a number up to justify one, use a gradient.
3
Why does a toggle or list filter widget need a step, and what do you group it by?
A
Filter widgets don't generate values on their own — they display whatever a step's query returns. So the workflow is always step first, then drag the widget and bind it. For dates I group by a date part or a custom bucket field from the recipe, never the raw timestamp, because a raw date has thousands of distinct values and would render thousands of buttons. I also filter the step itself so only the periods anyone cares about appear — curating the buttons is a query-level decision.
4
A user selects a year on one toggle and the quarter toggle's options mostly vanish. Explain what's happening and how you fix it.
A
That's faceting. By default, every widget's step is filtered by selections made in other widgets, so the quarter step re-queries with the year selection applied and only returns quarters present in that year. Great behaviour for charts, wrong for filter widgets — a filter should always show the full set of choices so the user can change their mind. The fix is to uncheck 'apply filters from faceting' on that widget, which stops its step listening to other selections. I leave faceting on for charts and tables, and off for toggles and lists.
5
How do you decide when a dashboard needs a values table rather than a chart?
A
Charts answer 'how much' and 'which is biggest'. The moment the question becomes 'which records exactly' — which deals are at risk, which cases breached — you need row-level detail and that's a values table. I set sensible default fields on the dataset so every table starts from a reasonable set of columns, order them identifying first, categorical next, numbers on the right, and use fit-data-to-widget spacing so more rows fit in the space available.
6
Someone changes a field's label in a dashboard step and it doesn't stick elsewhere. What's going on?
A
They're in the wrong editor. The Fields panel inside a dashboard step only controls what that query returns and how that one widget displays it. Real field metadata — display labels, number formatting, default columns, hiding fields — lives on the dataset and has to be edited in lens mode. It's a very common source of confusion because both places are called Fields and look similar.
7
How would you display product images or owner photos in a dashboard?
A
If the dataset carries the image URL as a field, bring it into a values table as a column and switch on 'show data as image' in that column's properties. CRM Analytics then renders the picture instead of the URL text. Owner photos on a top-performers table are the classic use — faces make a leaderboard instantly scannable. The caveat is that the image must be reachable by the viewer's browser, so anything behind a separate login renders broken for everyone but the author.
8
What can you do with a custom tooltip that you can't do with the default one?
A
You control exactly which dimensions and measures appear, and you can add custom text that interpolates step field values, so instead of a stack of label/value pairs you get a sentence about that specific data point. I usually hide the defaults that just repeat the axis. I also use tooltips as inline help — putting the definition of a metric like pipeline coverage right where people get confused kills the recurring email thread and documents the dashboard for whoever inherits it.
9
What's a chart marker, and when do you use one?
A
A marker is an annotation pinned to a specific bar or data point, with a label, a longer message, and a colour. It's for the story the data can't tell you itself — the month revenue dipped because a major account churned, or the quarter the methodology changed so a step-change isn't a real trend. I use them sparingly, one or two per dashboard, and only on stable anchored charts, because a marker is tied to a data point and can end up next to the wrong thing if users re-filter or re-sort the chart.
10
How would you build the KPI row at the top of a dashboard?
A
A number widget with shortened formatting and conditional formatting for the value and its status, paired with a timeline chart stripped of axes, titles, legends, and gridlines to act as a micro chart showing direction. Timeline rather than bar because it uses a true time axis, spaces points correctly, and supports the projected line. A polar gauge for the one or two metrics with a real target, since it shows position within a range better than a number does. Then a consistent theme across the strip and reduced cell spacing so it reads as one band.
11
You enabled the projected line on a timeline and it doesn't appear. How do you debug it?
A
The projection extends from today forward, so the first check is whether the data spans today's date — if the dataset ends last quarter, or a dashboard filter or facet trims the range before today, there's nothing to anchor to and the line silently doesn't render. Then I check there are enough historical points for the projection to be meaningful. I also caveat it in front of leadership: a trend extension is directional, not a forecast model.
12
How do you connect a set of dashboards so users can move between them?
A
Link widgets, pointed at saved dashboards, with 'apply selections' enabled so the user's current filters travel with them — otherwise they re-apply the same filters on every hop and the set feels like separate reports rather than one application. I style the links with a background and rounded corners so they read as tabs, keep the strip identical and in the same position on every dashboard, and highlight the current one. For that to work the dashboards need compatible filters on shared datasets with consistent field naming, so I design them as a family from the start.
13
Is there a way to make an image clickable in CRM Analytics?
A
There's no image-button widget, but you can compose one. Drop a container widget, give it a background image, then overlay a link widget covering the same area and set both its background colour and its font colour to fully transparent. The link still works and still carries filters — it just renders nothing visible. Both colours matter: leave the font opaque and you get stray text floating over the picture. Since there's no hover feedback, I add an affordance like a border or a caption so it's obvious the tile is clickable.

Discussion

No comments yet — be the first to start the discussion.