The hard part of data work is almost never the analysis. It's that the export dropped canceled orders without telling you, two dashboards define an active user differently, and the join silently duplicated a chunk of your rows.

So what helps isn't a tool that runs a regression. It's one that checks the file before it answers, tells you what it found, and shows its work on every number so you can reproduce it.

What Strawberry can help you do

Not sure where to start? Tell your companion what decision the number is for, and it'll suggest the right place to begin. See the skill

Get an answer out of what you have

The most valuable thing here usually isn't the analysis. It's catching the reason the analysis would have been wrong. So the file gets checked before anything is calculated, and you hear what turned up before you see a chart.

Then every figure that matters arrives with the filters applied, the rows excluded and why, and the formula behind it. What the data shows stays separate from what it implies.

Skill

Analyze a dataset I already have

Analyze this dataset for the question I need answered.

Collect what doesn't exist yet

Sometimes the dataset you need isn't anywhere. It's spread across a few hundred pages that each hold one row of it. Strawberry works in the browser, so it can go and get those fields, validate them as it goes, and hand back a table with a source against each row.

Skill

Turn websites into a clean spreadsheet

Collect these fields across these sources into a clean table.

Settle the definitions

Most arguments about numbers are really arguments about definitions. What counts as active. Whether the month is calendar or rolling. Which timezone the day ends in. Two tools can both be right and still disagree.

Because your companion works across the logged-in tools you already have open, it can compare how each one defines the same metric rather than taking one dashboard's word for it. That's usually where the discrepancy turns out to live.

Show people what you found

An analysis nobody acts on may as well not have run. Once the findings are accepted, they can become a presentation that leads with the answer and keeps the evidence traceable underneath it.

Skill

Present findings to stakeholders

Turn this accepted analysis into a presentation.

Keep the room honest

Most analysis meets its real test in a meeting, where someone asks why this number disagrees with the one in their dashboard. Turning up with the definitions already reconciled is the difference between a review and an ambush.

Stop rebuilding the same report

Recurring reports have a habit of drifting. The definitions move slightly, someone applies a different filter, and two quarters stop being comparable without anyone deciding that.

Once a report is right, the definitions, filters, quality checks, and output can be saved so next period runs the same way. If several people build versions of the same thing, that accepted method can be shared instead.

Skill

Set Up a Shared Team Workflow

Turn this report into something the team runs the same way every period.

When it genuinely repeats on a schedule, it can become a Routine that pulls the data, runs the same checks, and brings you the result to review.

Research deep enough to bet onThe wider research system: sourcing, mapping a landscape, and structured evidence. Turn websites into a clean, reviewable datasetBuild the dataset itself when the numbers you need are spread across many pages. Run finance on autopilotTake the same validation discipline into recurring reporting and performance review.
Official Strawberry skill
Copy skill file Download skill file
name
getting-started-with-data-work-in-strawberry
description
Help someone doing data work collect the data they need, analyze data they already have, reconcile conflicting definitions across tools, and present findings people will act on. Use to choose a first data workflow and complete one real result.

Getting Started with Data Work in Strawberry

Help the user finish one real piece of data work, then show them the workflow behind it. Most of the time in this job goes on getting the data into a usable state and on arguing about definitions, not on the analysis itself.

Find out what the answer is for

Ask what decision this feeds before touching the data. A number with no decision behind it invites a tour of the dataset rather than an answer, and that is how an afternoon disappears.

Route to the work

  • Analyze data the user already has: strawberry/research-analysis/analyze-a-dataset — quality checks first, then definitions, then the question, with the limits stated.
  • Collect data that does not exist yet: strawberry/research-analysis/extract-web-data for the same fields across many pages, into a validated table.
  • Define or review metrics: strawberry/product-engineering/review-product-metrics when the question is whether the framework itself is right.
  • Reconcile conflicting definitions: strawberry/marketing/review-marketing-performance when two tools disagree about the same number and someone has to settle it.
  • Build a recurring report: strawberry/finance/prepare-a-recurring-finance-report for approved data that must be validated the same way every period.
  • Present the findings: strawberry/research-analysis/create-beautiful-slide-deck once the analysis is accepted.

Protect the answer

Say where each number came from and what was excluded to produce it. Keep observed values, derived calculations, and interpretation clearly apart. Never present a correlation as a cause because the timing is suggestive.

When the data cannot support the question, say so and explain what would be needed instead. That is more useful than a confident answer built on forty rows.

Where the browser matters here

Data work usually spans several logged-in tools. Strawberry works in the browser, so it can pull from the analytics tools, dashboards, and admin panels the user already has open, compare how the same metric is defined in each, and bring the numbers together without an export-and-paste round trip for every source.

Keep what works

Preserve the accepted definitions, filters, quality checks, and output format as a custom skill so the same report is not rebuilt slightly differently each period. Use strawberry/operations/set-up-shared-team-workflow when the team needs one shared version. Add a Routine once the source, cadence, audience, and review point are settled.