Research

Research is where Skeleton Key takes a Brand Profile and Campaign brief and produces five angles of market intelligence. Five analyst agents run in parallel, each covering a distinct dimension. The Research run takes about 5-7 minutes, and each report is saved as soon as its analyst finishes. On a new run the tabs fill in order (Category, Competitor, Consumer, Culture, Social), and the selected tab stays on the earliest report that has arrived until you pick one yourself.

1. Category Analyst: market landscape, drivers of growth, structural trends.
2. Competitor Analyst: positioning and messaging from direct and indirect competitors.
3. Consumer Analyst: who buys, what drives their choices, barriers, and journeys.
4. Culture Analyst: societal values, cultural shifts, and moments relevant to your brand.
5. Social Analyst: consumer signals from Reddit, Instagram, TikTok, Amazon, and similar platforms.

A run has two stages. A search step first gathers evidence for the Category, Competitor, Consumer, and Culture analyses: web findings, each recorded with its claim, a supporting quote, the markets and period it describes, and its source. The analysts then write from that evidence and run their own searches to fill gaps, while the Social Analyst gathers posts directly from the platforms it covers. The search findings stay with the run, and Strategy can look them up for those four analyses when a report leaves out a figure, period, or source. Every analyst applies its findings to the markets in your campaign's target geography (for brand-level Research, the markets your brand names) and does not carry a finding from one market over to another.

When your campaign targets a non-English market, the analysts search in both English and the local language. For example, in a campaign targeting Japan, that means querying モンベル alongside "montbell" and surfacing local-language sources, so the findings and citations reflect what people in-market are actually saying rather than only English-language coverage.

What it produces

Each analyst's findings appear in their own tab on the Research page. Every analyst report opens with an Executive Read: a short note on what the analysis covers (the markets, competitors, audiences, and evidence it drew on), then the four or five findings that matter most. The sections that follow are named for the analyst's findings rather than a standard template, and each one carries a table, chart, list, or quote as its evidence. Every report closes with Risks & Evidence Gaps, which flags the figures the analyst derived, the inferences it relied on, conflicts between sources, and weak evidence.

After all five complete, you can generate two synthesis outputs by clicking Start on the Synthesized Reports strip: a One-Page Summary that distills the key findings across all five analysts into a single scannable summary, and a Full Report that weaves the evidence, insights, and story from every analyst into a cohesive document.

Both reports appear in the Synthesized Reports strip above the analyst tabs and open in a full-screen modal when clicked. Export options live in the modal toolbar. See the Reports page for more.

How it fits in the workflow

Research requires a completed Brand Profile. That is the only prerequisite for brand-level Research. Campaign-scoped Research also requires a Campaign brief, which gives the analysts the context they need to focus their findings on your specific objectives, audience, and market.

The distinction matters for what comes next. Strategy and Media Plan are always campaign-scoped. When you start Strategy inside a campaign, you choose its Research source from that campaign's completed runs or from any completed brand-level run. Brand-level Research can therefore ground a campaign's Strategy, as well as serve as exploratory market intelligence outside any specific brief.

The full dependency order is: Brand Profile → Campaign → Research → Strategy → Media Plan. Each phase becomes available once the previous one is complete. The product enforces this automatically.

Refining and version history

Each time you run or refine Research, a new run is created. Prior runs are preserved as version history and accessible via the History icon on the Research page. You can restore any previous version with one click, which promotes it to the current run.

When you Refine, you submit feedback and the platform determines which of the five analysts need to re-run based on your input. Analysts whose outputs are unaffected are carried forward unchanged, and an analyst that does re-run changes only the sections your feedback touches. This keeps iterations targeted rather than rerunning everything from scratch. During a refine or rerun, reports appear as soon as they are saved rather than in tab order. If a refine fails or is cancelled, the page shows the reports from the latest saved version, which may not include the analyst whose refine failed, under the failed or cancelled banner. Earlier versions stay available in History. The full model, including how analyst routing works and how version history grows, is in the Refine workflow guide.

You can optionally attach documents to any Research run. PDFs, Word files, Markdown, CSVs, and Excel sheets uploaded on the Documents page can be selected at run time, and the analysts treat them as high-priority context alongside web research. One run reads about 600,000 characters of document text in total. If a file is trimmed, the run page says how much of it was read. Standing rules on the Guidelines page take precedence over attached documents when they conflict. See Documents for the details.

If this is your first Research run, the Quickstart walks through it end to end. Otherwise, the per-analyst pages and the Refine workflow are linked below.