Market Research Dashboard: Built From the Decision

A market research dashboard is usually built from the variable list. What the BI adoption numbers show, and how writing the decisions first changes the scree...

Fieldwork on a brand tracker closes on a Tuesday. By Thursday the market research dashboard is live. There is a grid of charts, a filter bar, brand colors, and a link sent to nine stakeholders. The following Monday the commercial director asks a question in the meeting, and somebody answers it from a spreadsheet.

Key Takeaways

• A market research dashboard earns its place when a named person can point at a tile and say which decision it settles.
• Analytics and business intelligence tools are used by an average of 29 percent of employees, per Gartner figures published by IBM.
• A BARC and Eckerson study of 214 companies put average adoption at 25 percent, with no real movement across seven years.
• The default panel that ships with a collection platform is built from the variable list, which is why it answers questions nobody asked.
• Writing the decision list before the tile list is the whole method. It always makes the dashboard shorter.

What a market research dashboard is supposed to do

A market research dashboard is a live screen that shows survey results to the people who commissioned them. It updates as new waves land, so a stakeholder can read the answer without asking the research team for a fresh cut.

A tile is one chart, table or number on that screen, with its own question, its own filter and its own cut of the data.

Greenbook's introduction to market research dashboards lists what the market treats as standard: real time data, filters, drill down, sharing, alerts. Every item on that list is a capability. None of them is a question.

How a team gets its dashboard today

The sequence is short, and it is the same almost everywhere.

1. Fieldwork closes and the data file is cleaned and weighted.
2. The collection platform opens its own dashboard on the study, already populated.
3. Someone picks, from the variable list, the questions that make a decent chart.
4. Someone fixes the colors, the order and the titles.
5. The link goes out by email to the stakeholder list.

Nobody in that sequence writes down a decision. The starting point is the variable list. That list exists because of what the questionnaire asked, and the questionnaire was written to cover a brief, not to feed a screen. Greenbook's guide describes the resulting artifact accurately, feature by feature.

The tool decides the method here, the same inversion that shows up in [where a research community is asked to live](/blog/research-community-on-whatsapp). And most teams still export the panel to PowerPoint for the readout, because the question the director asks is not on any of the tiles. The questionnaire behind it gets [tested for path coverage before field](/blog/survey-testing-before-fieldwork). The screen it feeds gets tested by nobody.

The default panel that ships with the collection platform

PERSONAL EXPERIENCE This is the meeting, and we have sat in it more than once.

The panel opens. There are 18 charts on it. Awareness by wave, consideration by wave, the funnel, a demographic breakdown nobody asked for. The commercial director looks at it for forty seconds. Then they ask whether the September promotion moved anything among shoppers who buy the large pack.

That answer needs two variables that are not on the same screen and a filter that was never built. The researcher writes it down, goes back to the data file, and sends the cut on Thursday. The dashboard was open the whole time.

What the adoption numbers actually say

Research should borrow this evidence from business intelligence, which has been measuring the same failure for a decade.

Gartner figures published by IBM put the average share of employees who use analytics and BI tools at 29 percent. That is after 87 percent of organizations reported growing their analytics user base. The tools spread. The using did not.

The second study is more uncomfortable, because of its sample size and its span. BARC and Eckerson surveyed 214 companies and found average adoption at 25 percent, flat across seven years of tracking. In the same work, 50 percent of data leaders said usage had increased greatly. The belief moved. The behavior did not.

A research dashboard is a BI dashboard with a smaller data set and a narrower audience. There is no reason to expect it to behave differently.

Why the default panel answers the wrong question

The default panel assumes the reader arrives already knowing what to ask.

IBM's write up names three causes for the adoption gap: the tools are complex, their scope is too limited to act on, and they cannot surface the unknown unknowns. The third is fatal for research. An unknown unknown is a finding the reader would have wanted but never thought to ask for, so no filter on the screen leads to it.

A stakeholder does not open a dashboard with a query. They open it with a decision they are stuck on, and they need the screen to meet them there. A tile built from a variable cannot do that, because the variable does not know what the decision is.

Start from the decision, not from the variable list

UNIQUE INSIGHT Reverse the order of the two lists and the screen changes shape.

1. Write the decisions the dashboard has to support, in business language, as sentences somebody would say in a meeting.
2. Name the owner of each decision by role, then check that role against a real person.
3. Write the literal question that owner will ask when the decision comes due.
4. Only then choose the cut, the base and the chart that answer that question on one screen.
5. Delete every tile that cannot be tied back to a decision on the list.

Step five is the one that gets argued about. It always removes tiles. Removing tiles feels like losing value to a team that bought a platform advertised on how many chart types it offers.

Every tile answers to a decision

A map read left to right in three columns: the decision, the role that owns it, and the tile that settles it. First row: keep or drop the September promotion, owned by the trade marketing lead, settled by promotion weeks against baseline broken by store format. Second row: which region gets the next media flight, owned by the media planner, settled by awareness by region across the last three waves. Third row: whether the new pack goes national, owned by the category director, settled by trial and repeat by pack size in the test cities. Fourth row: which claim leads the next campaign, owned by the brand manager, settled by claim preference by segment against the current claim. Below a dividing line sits a fifth row drawn in grey and dashed, with nothing in the decision column and nobody in the owner column. Its tile is deleted. The rule the whole map states is that no decision means no tile.

The grey row is the method. A tile with no decision behind it and no owner in front of it comes off the screen.

The tile that survives the deletion test

UNIQUE INSIGHT A tile that cannot name the decision it supports is a tile nobody will open twice.

The test runs in one pass over the screen. Point at each tile and ask which decision it settles and who owns it. A decision owner is the single named role that will act on what the tile says, so a tile without one answers to nobody.

If the answer is that the tile gives useful context, the tile is decoration. Decoration is not neutral. It costs screen space and load time. It teaches the reader that most of the screen can be skipped. That habit is how a dashboard ends up unopened.

Where Cassi.ai comes in

Cassi.ai is a software engineering company specialized in the pains of market research, innovation and insights, working with enterprise research teams and agencies. Cassi.ai builds the dashboard as an application, not as a template: the decision list comes first, the tiles are written against it, and the statistical work the question actually needs sits inside the panel instead of in a separate export.

Not every study earns that, and the last section says which ones do not. The rest of what we automate for research operations sits in the Cassi.ai portfolio, including [the platform side of the same problem](/blog/why-chatgpt-is-not-a-research-platform).

Tracking dashboard compared with a project dashboard

The feature set both start from is in Greenbook's dashboards guide. The split between the two, and what each has to freeze, is our own reading from building them.

What to measure once the dashboard is live

UNIQUE INSIGHT Licenses issued is the wrong number, and it is the number most teams report.

Three measures are worth keeping, all off the decision list.

1. How many of the named decisions were taken with the dashboard open.
2. How many tiles nobody opened across a full wave.
3. How many stakeholder questions still arrive by email because the panel could not answer them.

The third is the honest one, because it counts failures directly. Set the benchmark against the 29 percent average usage IBM reports from Gartner. A research dashboard with a named audience of nine people and a decision list behind it should beat a company wide BI rollout. If it does not, the decision list was wrong.

When the platform's own dashboard is enough

Sometimes the built in panel is the right answer, and building anything else is waste.

Consider a single study with one question, no next wave, and no audience beyond the team that ran it. That study does not need an application. Neither does an internal check only one analyst will act on, such as a quick read on sample composition.

The BARC and Eckerson study of 214 companies found average adoption stuck at 25 percent. A screen that four people will open twice is not where that problem gets solved, and a custom build for it returns nothing.

FAQ

What is a market research dashboard?

A live screen that shows survey results to the people who commissioned them, updated as new waves land, so a stakeholder can read the answer without asking for a new cut. Greenbook lists the standard feature set: real time data, filters, drill down, sharing and alerts.

Why do most dashboards go unused?

Because they are built from the variable list rather than from a decision. Average usage of analytics tools is 29 percent of employees, per Gartner figures published by IBM, and BARC and Eckerson measured 25 percent adoption across 214 companies, flat for seven years.

What is the difference between a tracking dashboard and a project dashboard?

UNIQUE INSIGHT A tracking dashboard is opened every wave by the same owners, so question wording, base definitions and weighting have to stay frozen for wave on wave comparison to hold. A project dashboard serves one study, one audience and a few weeks, and it follows the analysis plan agreed before fieldwork.

How many charts should a research dashboard have?

UNIQUE INSIGHT As many as there are decisions on the list, and no more. The count is an output of the method, not a target. The deletion test holds it there: a tile that cannot name the decision it settles and the role that owns it comes off the screen.

Is the dashboard that comes with my survey platform enough?

Sometimes. For a one off study, with one question, no next wave and no audience beyond the team, the built in panel is the right call. If stakeholders keep emailing the research team for cuts the panel cannot produce, that is the signal it was built from the variable list.

A dashboard is measured by the decisions that came out of it, and by nothing else. Tile count, chart variety and seats issued are easy to report. None of them says whether anyone acted.

If you have a tracker running, the exercise worth an hour this week is to open the dashboard, point at each tile, and name the decision it settles and who owns it. Whatever you cannot name is what the next version deletes.

Published by Cassi.ai. Read the full article at https://www.cassiai.com/blog/market-research-dashboard-decision-first.