Verified against Perplexity Labs · 2026-07-31
Brief Perplexity Labs to produce a report, a data table, and a dashboard in one linked pass
A deliverables-first brief for Perplexity Labs that names each output asset explicitly, so a multi-step research task produces a linked report, spreadsheet, and simple dashboard that all agree with each other instead of just a written summary.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Run this as a Labs task. I want more than a chat answer — build each of these as a distinct, linked asset in the same task, not just a written summary with numbers mentioned in passing. DELIVERABLES REQUIRED a two-page written report, a spreadsheet of the underlying comparison data, and a simple one-page dashboard summarizing the top 3 findings TOPIC / BRIEF Compare the total cost of ownership of three CRM platforms for a 40-person sales team over 3 years. WHO THIS IS FOR Our VP of Sales, who needs to justify a platform switch to finance DATA REQUIREMENTS Break costs into licensing, implementation, and estimated migration cost separately, not as one bundled number. CROSS-ASSET CONSISTENCY RULES - The written report and the data table must agree with each other — if the report states a number, that exact number must appear in the table alongside its source, not only in prose where it can't be checked against the underlying data. - Any chart or dashboard view must be built from the data table produced in this same task, not from a separate, unlinked source pulled in just for the visual. - Use the same units, currency, and time period consistently across every asset — if the report states an annual figure and the table happens to show a monthly one, convert one to match the other and say which conversion was applied. SOURCING Cite sources on the data table itself, row by row, not only in the written report's prose or footnotes — a reader should be able to check any single data point without cross-referencing a different document. GAPS If a requested asset genuinely cannot be built from available data — for example, no public data exists for one specific comparison point you were asked to include — say so explicitly on that specific asset and explain what's missing, instead of quietly omitting it from the final deliverable set or filling the gap with an unlabeled estimate. FINAL CHECK Before finishing, re-read the report and the table side by side and flag any number that doesn't match between them, rather than assuming they're consistent because they came from the same task.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
Perplexity Labs is the slower, agentic mode built to output more than one artifact type per run — a written document, a spreadsheet of the underlying data, and a simple web page or dashboard — instead of a single chat-style answer, and it decides for itself which of those output types to bother building based on how open-ended the request reads. Naming the exact deliverable list up front removes that judgment call: an open request like 'research this and give me a report' can come back as text only, while explicitly listing a report, a spreadsheet, and a dashboard as required assets makes Labs treat each as a distinct commitment it has to satisfy, not an optional extra it can skip if the topic seems to call for less. The rule that the report and the table must agree, and that any chart must be built from that same table, exists because a multi-asset agentic task can otherwise generate each artifact from a slightly different pass over the sources, producing a report and a spreadsheet that quietly disagree with each other on the same number — linking them forces internal consistency instead of three separately-sourced outputs bundled together under one task label. The final side-by-side check step matters because self-consistency checking works dramatically better as a distinct, named step performed after the fact than as an implicit expectation baked into generation — a model asked to 'make sure it's consistent' while writing has no actual mechanism to compare a number it wrote three sections ago against one it's writing now, whereas asked to re-read and compare afterward, it's running an actual comparison pass over completed text rather than trying to hold every number in working memory while still generating.
Verified against
Perplexity Labs 2026 · 2026-07-31
Changelog
- 2026-07-31 — Initial publish, verified against Perplexity Labs multi-asset output mode.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
