A client emails the week before fieldwork and asks you to add a fourth concept to a three-concept test. In most research agencies that request goes to the survey programmer, then to link testing, then to data processing, then to whoever builds the charts, and each of them has a queue.
In the MX8 Labs Research Platform, the researcher can make the change, review the revised survey, and carry it through to reporting.
Judgment and production
By production we mean everything that turns an agreed design into a delivered study: programming the survey, testing it, cleaning and weighting the data, running the tables, and building the charts and slides. Researchers set the rules for that work, including how to clean and weight the data, and review the output.
Judgment is deciding what to ask, whom to ask, and what the answers mean for the client. It depends on the brief, the category, and earlier research, and two good researchers can reasonably disagree about it.
Researchers have long spent more of the week on production than on judgment. In Greenbook's GRIT Insights Practice survey of 531 insights professionals, fielded in the fourth quarter of 2020, respondents reported spending 17% of their time managing research execution and another 17% analyzing, charting, and reporting results. Consulting on implications took 13%. (MarketingCharts summary of the GRIT data)
One late change, start to finish
1. Change the survey. Describe the change to Chat in plain English: add the fourth concept, include it in the rotation, and ask the same follow-up questions. Chat proposes the edit, and you review and apply it. A new study programs from a Word document in about five minutes, and an edit to an existing one is quicker.
2. Check the build. Link testing meant someone clicking through every route of the survey to confirm the skips worked, and on a complex study it could take days. The platform now simulates respondents through the survey and flags any question that can never be reached. See Testing and reviewing your survey. You still read the survey through as a respondent would, because simulation checks the logic and cannot tell you whether a question is clear.
3. Check the sample. Four concepts need more respondents than three if every comparison is to keep its base size. Decide whether to add sample or accept smaller cells before fieldwork starts.
4. Read the results. Weighting, cross-tabs, and significance tests run on the platform as responses arrive, so the new concept appears in the same reports as the other three. See Reporting with MX8 Labs.
5. Build the deck. Ask survey Chat for the presentation. It proposes an outline and the reports it will draw on, waits for your approval, and returns an editable PowerPoint in your template with native charts. See Editable PowerPoint Charts in MX8 Labs.
For comparison, the legacy research cycle runs about 18 days from brief to report. A late change sends part of that cycle around again.
What the change still asks of you
The platform did not decide whether a fourth concept belongs in this study. Someone has to judge whether the new concept is comparable with the three already written, whether the sample supports four cells, and whether the client's question has changed along with the stimulus. Someone also has to explain to the client what the result means once the data is in.
Those decisions were always the researcher's. On the old timetable they competed for attention with chasing the build.
What this means for the agency
Fewer queues between stages. Today a researcher writes programming notes and waits overnight for a survey programmer in another time zone. The grid comes back wrong, so there is another note and another night. Then the survey goes to link testing, the data goes into the data processing queue for weighting and tables, and someone else builds the charts. Each of those is a separate team or an outside vendor, and each one needs the study explained again. When the platform does that work, the researcher makes the change and sees it in the survey within minutes.
One researcher can take a study from brief to deck. The person who understands the client's question is the person who changes the survey, checks the sample, reads the results, and asks for the presentation. Fewer people need briefing, and less is lost between them.
The agency is paid for its judgment. Production has been a large share of what a study costs. As that cost falls, clients can afford more studies on the same budget, and what they are buying in each one is the design and the recommendation. Agencies with strong researchers gain from that.
None of this needs a reorganization on day one. Most of the work the platform takes on is work agencies already send out, so the first effect is on vendor spend and turnaround.
Questions researchers ask
Will AI replace researchers?
It automates programming, testing, and report assembly. Researchers still decide on the design, cleaning and weighting rules, and recommendations, and review what the platform produces.
Do I still need to test the survey?
Yes. Simulation exercises the logic. Read the survey as a respondent, share a test link with the client, and start fieldwork with a soft launch.
Can I keep using my own analysis tools?
Yes. Data exports to SPSS and Excel, so any stage can be done in the tools your team already uses. See SPSS export format.
How do I compare this with my current process?
Run one study both ways. Record the programming turnaround, the number of corrections, the time spent preparing the report, and the time spent reviewing it.

