Example · Market Research Analyst

50% of your week as a market research analyst is still yours.

AI now programs surveys and builds the deck. Deciding what to study and why stays with you.

Make my planA typical week at the middle pace. Your plan uses your own tasks.

Is market research a dying career?

In a typical week at the middle pace, 35% of this work needs your judgment and 15% is work you do using AI. AI could do about 50%. Production work is about half of your week, and a model does it today. It is also the half that justified hiring you.

Read more

The industry's own vendors now sell survey programming, tabulation, the tracker refresh, the deck and the competitive scan. Your profession also faces a rare threat, because the respondent can be synthesized too. Qualtrics started selling prebuilt synthetic panels this spring, and a consumer brand was among the first to test them.

What neither the model nor the synthetic panel can do is the part people already come to you for: knowing whether a sample can answer the question, turning a brand manager's email into a study that will change a decision, and sitting in a backroom while a brand team hears a consumer call the concept confusing. That work is often not yet your title, and it is exactly what an Insights Manager posting describes. The plan is about making it your title before the brand teams learn to run the synthetic panel themselves.

The week, task by task

Who does each hour of a typical 40-hour week.

See every task
You35% · 14 hours

Needs your judgment.

  • Study design: business question to objectives, method, sample, questionnaire architecture9%
  • Synthesis: the "so what" and a recommendation against the business question8%
  • Moderating and the backroom: live groups and IDIs, brand team behind the glass, same-day debrief6%
  • Agency and panel vendor management: briefs, bids, SOWs, budget5%
  • + 2 smaller tasks8%
You using AI15% · 6 hours

AI does the slow parts. You still own it.

  • Fielding management: panel quotas, incidence, speeders, fraud and bot cleaning8%
  • Focus group coordination: screener, recruiter brief, facility, first-draft discussion guide6%
AI50% · 20 hours

AI could do most of it.

  • Building the insight deck: charts, formatting, headline writing12%
  • Survey programming and logic testing in the survey platform9%
  • Data cleaning, quality checks, tabulation, weighting, crosstabs8%
  • Secondary and competitive synthesis (scan data, syndicated reports, trade press)8%
  • + 2 smaller tasks13%
Why some tasks still need a person
Telling a brand manager why the survey in their email will not answer their question

This rests on how hard mistakes are to catch. A bad study returns a confident number and nobody finds out until the launch underperforms. Catching it before fielding needs someone who has watched a leading question produce a false winner.

Turning "should we launch the spicy line?" into objectives, method, sample and a questionnaire architecture

This depends on context. The model can write a concept test, but it does not know that sales already promised the retailer a Q2 slot, that R&D has two viable formulas, or that the last test failed because the sample skewed to current buyers.

The backroom and the same-day debrief

This needs presence and context. AI-moderated one-to-one interviews are live products (Outset, Conveo). A live group with the brand team behind the glass, and you reading the room and re-ordering the guide on the fly, is a different job.

Saying "that move is inside the margin of error" to a VP

The accountability here is borrowed. Someone must stand behind the number when a brand team wants to act on noise. Today you likely do it as the defendant, after the fact. The plan changes when and who initiates.

Knowing which past study already answered the question

What protects this is context that won't last long. It is worth a lot because nobody wrote it down. In Week 7 you write it down on purpose and trade the task for the credit.

Good to know

What is hard about this field, and what you can do about it.

The hard truth

Online opt-in samples can mislead you.

Their answers can swing results, especially for young adults. Checking the data before you share it protects your work and your name.

What you can doBecome the person who vouches for the data. Add quality checks, compare against trusted benchmarks, and flag weak samples before results reach a deck.
Sources (2)

What's next for a market research analyst?

Plan A is the best move inside the work you do now.

APlan A

Grow in the job you have

The main focus. Most weekly sessions build toward it.

Your next stepResearch lead for one brand
Each week
4 sessions
Ready in
4 to 6 months
Why it fits

Why you. If you already tell brand managers why their surveys fail, want to be deciding what gets studied rather than building what someone else decided, and the seat above you is open or thin, this is the move.

The gap. You have no charter and no ledger of studies you designed and decisions they changed. Your method breadth (conjoint and MaxDiff, two ways to measure trade-offs, and pricing) is thinner than an Insights Manager posting asks. You also need to handle the politics of telling a brand director "that isn't a research question."

The trade-off. You depend on one brand director, and you will be arguing, in writing, for tools to do the work you were hired for. That is the right argument, and it still costs you. You get fewer studies, bigger bets and more visible mistakes.

2 other moves that fit
  • Own the admissibility gate for synthetic and AI-moderated methodsReady in 6 to 9 months
  • Qualitative moderator (the hedge)Ready in about 12 months to be good at it

What's a plan B for market researchers?

Plan B is meant to be different from this job. It was picked because skills from this role carry over.

BPlan B

Get ready for a new kind of job

A second option. A newer job AI is creating that these skills already fit.

A job to aim forAI evaluation specialist

Catching a leading question before fielding is the same instinct that decides whether an AI output, or a synthetic panel, can be trusted.

Each week
1 session
Outlook
Hiring now
Pay
$230K to $270K a year (one posting)One Anthropic safety evaluations posting, 2026; AI lab pay runs well above most employers.
2 other options
  • AI value and ROI analystYou already tell a VP when a move sits inside the interval, and AI investments need someone who will say that about their claimed returns.
  • AI product managerTurning a vague launch question into objectives, method and sample is product discovery, and AI features need people who can frame the right question.

In your own plan you can switch to any of the 47 roles.

The 13 weeks

Small steps that add up. One plan for both: most sessions build Plan A, one a week builds Plan B.

By week 13 you will have

AFor Plan A

  • A tracker memo with thresholds the brand teams agreed to
  • Written rules for when synthetic results can be used
  • A published AI test write-up

BAnd for Plan B

  • A draft write-up of your AI test
  • A published write-up of your AI test for your portfolio
Month 1Weeks 1 to 4
  1. 1Measure your week
  2. 2Test AI on survey building
  3. 3Replace the tracker deck
  4. 4Plan a synthetic panel test

You finish withA side-by-side test plan and your Director's answer on a trial

Month 2Weeks 5 to 8
  1. 5Start using a decision brief
  2. 6Moderate a focus group
  3. 7Index the study archive
  4. 8Run the test and decide

You finish withA written decision on which move to pursue

Month 3Weeks 9 to 13
  1. 9Write rules for new methods
  2. 10Check the market and pay
  3. 11Measure your results
  4. 12Send the memo and ask
  5. 13Decide your next step

You finish withA one-paragraph decision with a date

Three courses from the plan
  • Bisbee, Clinton, Dorff, Kenkel & Larson, "Synthetic Replacements for Human Survey Data? The Perils of Large Language Models"Political Analysis, 2024 (doi 10.1017/pan.2024.5); free preprint at osf.io/5ecfa · Free (open access)

    Yes, Week 4. It is the clearest evidence that synthetic panels get the mean and miss the spread.

  • Park et al., "LLM Agents Grounded in Self-Reports Enable General-Purpose Simulation of Individuals"arXiv 2411.10109, v3 June 2026 · Free

    Yes, Week 4. It is the strongest case for the other side, so know it before someone quotes it at you.

  • FTC Policy Statement Regarding Advertising SubstantiationFTC · Free

    Yes. Week 9.

Checked 2026-09-22.

The fuller picture

Risks beyond AI: moderateBudget risk

Research budgets are discretionary and get cut with marketing budgets in a volume downturn. Agency retainers are often the first to go. Consumer-goods office headcount has been contracting on its own, before synthetic panels shipped. Brand-side research budgets are rising overall, but the money is flowing to platforms rather than analyst seats.

Where this could be wrong
  • The shares are guesses until Week 1. If your log shows far more deck production than we assumed, more of your week is automatable, junior openings shrink faster than we show, and the research lead move becomes urgent.
  • Treating survey programming as automatable rests on vendor claims, not a benchmark. Your Week 2 count is the number that matters. If you find four faults per twenty questions, AI only helps with the task rather than doing it, and your seat is safer.
  • The synthetic evidence is moving. Bisbee tested ChatGPT in 2023; Park's v3 is June 2026. If a vendor shows segment-level spread that matches your human waves on your categories, the method gate narrows fast.
  • We assumed a mid-size company with no MBA brand program. At a P&G or General Mills the brand-management door is real and our decline of it is too strong.

This is an example. Your plan uses your own week.

Make my plan

Also matched: market research analyst, consumer insights analyst, insights analyst. Research checked 2026-09-22. Page updated .