Technology · Partner practice

Synthetic research

A survey where synthetic respondents answer instead of people. Test ads, names, packaging or price in days, not in weeks of fieldwork.

1–3 daysfrom brief to report
10 typescreative, name, packaging, price, UX
2,160profiles in the platform library
1.5 hoursrun time for one study

What it is

Synthetic research is a survey in which synthetic respondents answer instead of real people: digital profiles built on data about the Russian consumer. The method has already entered global research practice — Kantar, NielsenIQ and Ipsos use it.

We use the Synthetic Research platform — a partner technology we bring into our work. The panel is built from research data: from the national statistics service to industry and syndicated market studies. The audience is split into 72 socio-demographic and 18 behavioural groups, 24 respondents per cell, with the sample assembled for the task.

From there it works like any study: brief, a panel for your audience, questions built on the method, a report with findings and respondent quotes. The difference is that the panel is digital and the run takes an hour and a half.

What you can test

Ten types of research — from a quick check of a hunch to expert research by your own methodology.

01
Quick testThe program picks the method for the task. A first pass: check a hunch before a full test.
up to 3 options
02
Strategic researchPositioning, messages, barriers and motivations. Brand launch, repositioning, entering a new category.
up to 6 options
03
Creative testEvaluate ad layouts before a campaign: compare options and pick the material for the channel.
up to 3 options
04
Product and concept testDemand, strengths, barriers. Test a concept before investing in production.
up to 3 options
05
Name testSound, associations, risks of misreading. Shortlist a final name from several options.
up to 3 options
06
Packaging testShelf visibility. Pick a design or check a redesign before printing a run.
up to 3 options
07
Price testPerception thresholds and ranking by Van Westendorp and Gabor–Granger.
up to 6 price points
08
Message testRanking by persuasiveness: shortlist the strongest wording for a campaign.
up to 6 options
09
UX testWhere users get stuck in a flow: sign-up, form, funnel — before rebuilding anything.
up to 3 options
10
Expert researchResearch by your own methodology, when the design is non-standard.
1 client scenario
If you have more options than the standard scope, the study is still possible — we agree it separately.

How a study runs

01BriefTask, audience, geography. We check the brief before launch and tell you what is missing.
02PanelA panel of 8–50 personas is assembled for the task: each with a profile, consumer jobs and media.
03AnswersQuestions follow the method of the research type, from warm-up to key. Each persona answers from its own profile.
04AnalysisFindings get a stability label; the study as a whole gets a stability index from 0 to 100.
05ReportDOCX, PPTX, XLSX. Respondent quotes, recommendations and a field-check plan.

Why this is not “just asking a language model”

The difference comes down to three things, and each one can be checked.

A panel, not improvisation. Profiles rest on data about the Russian market — the model does not invent an audience on the fly. Market structure, five geographic layers, income levels and media preferences are measured in Russia.

Dissenters in the panel. The sample includes critical respondents: sceptics, those loyal to a competitor, price-sensitive buyers. Today that is 30% of the panel. Otherwise a study turns into a chorus of agreement.

Stability labels. Key findings receive one of three labels, and the wording is honest: stable pattern if 70% of personas or more answered alike; notable tendency between 40% and 70%; isolated signals below 40% — a hypothesis to be checked with real people. The study as a whole receives a stability index: a score below 50 is called preliminary outright.

The methods are not invented either: the theory of planned behaviour by Ajzen, Van Westendorp price thresholds, the Gabor–Granger test, the Kano model, Kahneman's work. Each research type rests on its own method, and the method is described in the report.

Can it be trusted

KantarNamed synthetic data among ten marketing trends of 2026
NielsenIQLaunched idea screening on synthetic respondents in 2025
IpsosPublished guidance on synthetic data and announced a Stanford partnership on AI and synthetic data
ICC/ESOMARThe 2025 code defined synthetic data and personas and made AI disclosure mandatory
Kantar, NielsenIQ, Ipsos, ICC and ESOMAR are not affiliated with the platform's developer, took no part in building it and have not evaluated it. Information is cited from public company and industry materials.

Where the method is strong, and where it is not

Works well for

  • Filtering weak creative, names or packaging before an expensive test
  • Checking hypotheses before fieldwork and narrowing the discussion guide
  • Understanding the audience's language and barriers when there is no time
  • Reviewing a product flow before rebuilds
  • Testing an idea when there is no budget for classic research

Does not replace

  • Sensory evaluation: taste, smell, touch
  • Exact shares and projection onto the whole audience — that is quantitative work
  • Real market behaviour: whether people buy or click
  • Trademark clearance of a name — that is a patent attorney's job
  • The group dynamics of a live focus group
We say this plainly, in the report as well: synthetic research is a qualitative method. Its findings are hypotheses to be checked with a live audience, and every report includes a plan for that check. The ICC/ESOMAR code requires disclosure of AI use — we do disclose it.

What you get

A report in three formats. DOCX, PPTX, XLSX — findings, reasoning, respondent quotes and segment differences.
Stability labels. In the Excel table every key finding is marked stable pattern, notable tendency or isolated signal.
Stability index. A score from 0 to 100 across four axes, as a separate document you can pass on.
A field-check plan. Method, sample, indicative budget and timeline — if you decide to verify with real people.
Respondent quotes. Under every finding you see what it was built on, not just the conclusion.
Ownership of the report. Exclusive rights belong to the client: you can pass it on and include it in your materials.

Timeline

Guaranteed turnaround — no more than three working days after payment and receipt of the full set of materials.
Usually faster — on the day materials and payment arrive, or the next working day.
Program run — up to 1.5 hours per study. That is not the order turnaround.

Why a media agency needs it

We work in a logic of “calculate first, launch second”. Synthetic research fills a step that used to be skipped: a quick check of materials and hypotheses before the media budget is already spent.

Three scenarios where it works most often.

Before a pitch. Test positioning and messages so the strategy rests on audience reactions, not on guesses.

Before a campaign launch. Compare three to six creative options and keep the ones that read.

Before production. Check the name, packaging or price line — before the print run and the shoot.

“Synthetics does not answer ‘how many will buy’. It answers ‘what will people see in our message’ — and it answers before the budget is spent.”

Method Media · practice note
Assured Ad Solutions
115114, Moscow, Derbenevskaya emb.,
bld. 7, str. 9, BC "Novospassky"
Privacy policy
© 2020–2026 Method Media