Make a weekly report reusable with Orkas: inputs, checks, and a sample output

A weekly report needs more than a summary of the latest figures. The next person, or the next AI task, also needs to know which inputs were used, how the numbers were calculated, and what remains unresolved. Orkas is an open-source, local-first desktop application for macOS and Windows that can help organize this work. Its Commander coordinates specialist agents for research, writing, documents, presentations, and other project tasks.

This guide proposes a small reporting workflow. The input rows and report below are illustrative, with arithmetic checked by hand. They provide a concrete target to inspect when running a similar task with your own data.

1. Keep the inputs small and explicit

Imagine a seller has the following weekly export. Revenue here means attributed revenue in the supplied export. Spend means advertising spend. Neither column is a profit figure.

csv
week,channel,revenue,ad_spend
A,search,900,300
A,social,600,200
B,search,1000,400
B,social,500,250

Keep a short note beside the file: the reporting currency, the date range covered by each week, and the export source. For this illustration, the figures use the same unspecified currency and comparable week lengths. A real report should name those details instead of assuming them.

Orkas can work with supplied files and project context. Put the relevant export and notes in the project workspace, and give Commander a task that names the inputs and the expected result. Connecting a live advertising account is unnecessary for this file-based example.

2. Give the task an acceptance target

Here is a proposed instruction you can adapt:

Use the supplied weekly export and reporting notes. Compare week B with week A. Calculate total attributed revenue, total advertising spend, and revenue divided by advertising spend for each week. Show the arithmetic and identify the channel changes. Produce a short report with a summary, a data-quality section, and three follow-up questions. Use only the supplied information. Flag missing facts rather than guessing the cause of a change. Save the report for review; leave campaign settings and outbound communications unchanged.

The useful part is the output contract: a named comparison, defined metrics, visible calculations, and questions the source data cannot answer. Commander can coordinate the work, while the project files keep the inputs available for subsequent tasks. The user still checks the result before using it.

3. Inspect a complete sample report

For the four illustrative rows, a hand-checked report would read as follows:

Weekly comparison: B versus A

Attributed revenue remained at 1,500. Advertising spend increased from 500 to 650, a rise of 150 or 30%. Revenue divided by advertising spend fell from 3.00 to approximately 2.31.

Calculations:

  • Week A revenue: 900 + 600 = 1,500.
  • Week B revenue: 1,000 + 500 = 1,500.
  • Week A advertising spend: 300 + 200 = 500.
  • Week B advertising spend: 400 + 250 = 650.
  • Spend change: (650 - 500) / 500 = 30%.
  • Revenue-to-spend ratio: 1,500 / 500 = 3.00 in A; 1,500 / 650 = 2.3077, rounded to 2.31, in B.

Search contributed 100 more attributed revenue while using 100 more advertising spend. Social contributed 100 less attributed revenue while using 50 more advertising spend. Together, these changes explain the unchanged revenue total and the additional 150 of spend.

Data quality and limits: The file contains no order counts, gross margin, returns, attribution window, or customer-level information. It cannot establish profit, customer acquisition cost, or why a channel changed. The comparison also depends on the weeks covering equivalent periods and the export using consistent attribution rules.

Three follow-up questions: Did the attribution window or reporting definition change? Did the social campaign mix or creative change between the two periods? Are returns or delayed conversions likely to alter either week's revenue total?

This output gives a reviewer something specific to accept or correct. It separates calculations supported by the rows from explanations that require more evidence.

4. Preserve the reasoning for the next task

Save a short handoff note alongside the report:

Input: weekly export with four illustrative rows, weeks A and B.
Metric definitions: attributed revenue; advertising spend; revenue/spend ratio.
Confirmed arithmetic: revenue 1500 -> 1500; spend 500 -> 650.
Open questions: attribution settings, social campaign mix, delayed conversions.
Next step: add the relevant source notes and update the comparison.

For a real project, replace the illustrative input description with the actual source and date range. Orkas's project context, Library sources, memory, and reusable skills can support later work, such as turning the reviewed report into a presentation. Keep the underlying figures and unresolved questions available so that the next deliverable carries the same assumptions.

A recurring task can prepare a report on an agreed schedule, provided the assigned device, required access, and tools are available. Start by checking one report manually before making the workflow recurring.

5. Match the setup to the data

The Orkas core desktop application is free; model usage and optional services can add costs. Files and history are managed locally by default. Using external model providers or connected services can send the data needed for the task to those providers, so choose inputs and access accordingly.

The practical result of this workflow is a reviewable report plus enough context to continue it. Reusing a checked source file, explicit definitions, and a short handoff note helps keep subsequent writing and presentation tasks grounded in the same evidence.

Shared by ColunTina for the Orkas project. Product information checked against the official Orkas site on 5 November 2026.