At most one email a week. Unfollow anytime with one click.
Transcript
The full narration of this video. The current sentence is highlighted while the video plays; click any sentence to jump there.
Show transcript
Introduction
Hi, and welcome to the closing episode of the research session. Good research can go to waste if the report is confusing. Today we learn to present it neatly. We'll use a five-part structure, gather material in a Project, draft it as an Artifact, adapt the report to its readers, and check it before sending.
Step 1: Research report structure
Dewi has complete material on the opportunity to sell stationery to schools: the deep research report, the fact-check sheet, and interviews with five school treasurers. Now the director is waiting for the report. The main principle is simple: conclusion up front, details at the back. A busy director should understand the report from the first page, without reading all of it. Use five parts. First, the executive summary: the answer to the research question in five points. Second, background: what questions were researched and how, for example web research plus interviews with five schools. Third, findings: facts and numbers, each with its source. Fourth, recommendations: options for next steps with their pros and cons. Fifth, the complete source list with dates. This part is rarely read, but it's very important when someone asks where a number came from. This order also works for small research reports, such as Laras's competitor price research. The report may be just one page, but all five parts are still there. Why does this structure matter? Because of four common mistakes. First, the conclusion is saved for the last page. Busy readers often give up before they get there. Second, every finding is included. The important findings get buried among dozens of trivial ones. Choose the ones related to the decision. Third, numbers with no origin. One doubtful number can make the reader doubt the whole report. Fourth, vague recommendations, such as the phrase needs further consideration. That phrase helps no one make a decision.
Step 2: Gathering material in a Project
It's best to gather the report material in one Project, the workspace we learned about in the Projects session. The first reason: all the material is in one place. Second, the project instructions stay in effect. Dewi doesn't need to re-explain who the reader is every time she asks for a revision. Third, revisions stay consistent, because Claude always reads the same sources. The numbers in the summary and in the findings won't contradict each other. So that all the material is read together, Dewi creates a Project called School Market Report. The first file: the deep research report with dozens of web sources. The second file: the fact-check sheet, with the status of every claim. Now Claude knows what's proven and what's still an estimate. The third file: interview notes from five schools. Notice the description: the interviewees' names have been removed. Just write School A, School B, and so on. Interviewees' personal data doesn't need to go into AI. The fourth file: the office report template, so the result follows the format Maju Bersama Co. usually uses. On the upper right are the project instructions: the reader is the director, a five-minute read, conclusion up front, every number needs a source, and estimates must be marked. These instructions apply to every chat in this project.
Step 3: Drafting it as an Artifact
Inside the Project, Dewi asks for a report draft using the office template, starting with the executive summary. The draft appears as an Artifact in the right panel. Five summary points answer the director's question, complete with source numbers. The spending per school is marked as an estimate because it comes only from interviews. All the numbers here are examples. Before revising, Dewi reads the summary as if she were the director. Her question: if I only read this part, could I decide? The answer is almost, except for the recommendation. Dewi feels the recommendation point is too general. She doesn't ask for a full rewrite, just the recommendation section with concrete steps for the first three months. Claude updates the Artifact to version two. The old version is still saved, so Dewi can compare or go back if needed. Revising section by section like this keeps things much more under control.
Step 4: Numbers, tables, and sources
Now the findings section. Dewi asks for number cards from the interviews and a competitor table, complete with sources. In version three, three number cards catch the eye right away: five schools interviewed, three of them willing to try a new supplier, and they ask for thirty-day payment terms. The competitor table below shows each one's strengths and weaknesses. Notice the labels W2 and W4: those are interview codes, not people's names. Combining web sources with interview codes makes every row traceable, without revealing who the interviewees are. There are four rules for writing numbers in a report. First, every number has a source or an estimate label. No orphan numbers that come from who knows where. Second, consistent units. If you use per semester, use per semester throughout the report. Third, round sensibly. Eight to fifteen million is much easier to read than a number down to the last hundred rupiah. Fourth, keep facts and opinions apart. Findings in one section, recommendations in another. You can write these rules in the project instructions so Claude always follows them.
Step 5: Adapt for the reader, then final check
One piece of research can produce several versions of a report. For the director, one page is enough. For the field team, two to three pages. The content is different too. The director needs the summary, the risks, and the decision being requested. The field team needs a task list, a schedule, and the schools to contact. The style also differs: concise and formal for the director, practical and step by step for the team. Just ask Claude: make a field team version of this report. Claude will rework it from the same material in the Project. Dewi gives it a try. Still in the same Project, she asks for a field team version: a task list per person and a three-month schedule. Claude turns the recommendations into monthly tasks, complete with who does what. Same material, different shape. Then Dewi asks for a cover email to the director, just three sentences. The short email mentions the attachment, the conclusion, and the decision requested with its deadline. Dewi just rereads it, adjusts the greeting, and sends it herself from her office email. Before sending, run a final check. First, the summary answers the research question in five points. Second, every number has a source or an estimate label. Third, you've opened and matched the sources for the key numbers yourself. Fourth, there's no personal data about interviewees, such as school treasurers' names or phone numbers. Fifth, the recommendations are concrete: who does what, and when. Sixth, the format and file name follow office practice. You can copy or download the document from the Artifact panel. The screens in this video are simulations, so menu names may differ slightly because Claude is updated often. This checklist takes just five minutes. Five minutes that make your report trusted and help your boss decide faster.
Done!
The research session is complete. We've used web search, deep research, fact-checking, and competitor research, and now we present it all as a tidy, sourced report. From the next episode we enter the data analysis session, starting with the step people skip most often: cleaning and tidying data. See you there!
Practice exercises
Try these on your own after watching to lock in what you learned.
Write your research report in five parts: summary, background, findings, recommendations, sources
Ask Claude for a one-page version for your boss and a task-list version for the team
Run the final checklist and fix at least one finding that has no source yet
Ask the class material
Type a question or keyword; we find the part of this episode that explains it and jump the video there.
My notes
Notes are saved with the current video time and are only visible to you.
Questions & answers
Stuck on a step? Ask here; the instructor and other learners can answer.
No questions yet.
Your instructor
Cerdas Inovasi Digital Team
Software developers, BIM & 3D engineers, and AI practitioners
53 video classes1,281 episodes
We build software, BIM models, games and AI automations for clients every day, and turn that hands-on work into step-by-step classes: every step is shown on screen and comes with a PDF guide you can keep.
