HOW I STRUCTURE AND WRITE A RESEARCH REPORT FROM SCRATCH

A research report has one job.

That job is to move a reader from zero context to a defensible conclusion through a chain of evidence they can verify. Every word in the report is meant to serve this function, else it doesn’t belong in the report.

This guide breaks the process into five steps: the research question, the thesis, the evidence chain, the structure, and the conclusion. Each step is illustrated with examples for both crypto and non-crypto readers.

The first step is the research question. Before any writing begins, the question must be stated in one sentence that specifies the asset, mechanism, or market being examined and what about it is unknown or unresolved. A vague question produces a vague report. The question also determines the evidentiary standard, which is the level of certainty and quality of evidence required to make a defensible conclusion.

E.g.

A precise research question: “Does adding a second checkout lane reduce average customer wait time below three minutes during peak hours?”

That question names the mechanism (second lane), the metric (wait time), and the threshold (three minutes). The evidentiary standard requires timestamped queue data from peak hours, not customer satisfaction surveys.

A vague research question: “How do checkout lanes affect the customer experience?”

That question does not specify which metric, which conditions, or what outcome is being examined. A report answering it can say almost anything and be technically correct.

A precise crypto research question: “Does Uniswap v3’s concentrated liquidity reduce slippage for stablecoin swaps above $100,000?”

That question names the protocol (Uniswap v3), the mechanism (concentrated liquidity), the metric (slippage), and the condition ($100,000+ stablecoin swaps). The evidentiary standard requires trade execution data at that size threshold across liquidity ranges.

A vague crypto research question: “How does Uniswap affect DeFi trading?”

That question does not specify which version, which mechanism, which metric, or which type of trading. Any answer is defensible because no answer is falsifiable.

The second step is the source hierarchy. Primary sources are protocol documentation, onchain data, audited financial statements, and original research papers. Secondary sources are analyst reports, news coverage, and commentary. You should use the Primary sources to anchor your claims, while the secondary sources can be used to provide context. Any claim that cannot be traced to a primary source is marked as unverified or cut. This step needs to happen before writing begins because it determines what the report can and cannot assert.

The third step is thesis formation. A thesis is a falsifiable claim about the research question. It takes a position, while the topic statement announces a subject. The thesis can be argued against, but the topic statement cannot.

E.g.

Thesis: “Requiring helmets reduces cycling deaths but discourages people from cycling.”

Topic statement: “This report examines helmet laws for cyclists.”

The thesis makes two falsifiable claims: deaths go down, ridership goes down. A reader can disagree with either. The topic statement just names the subject. No one can disagree with it.

or…..

Thesis: “LUCID’s scheme-agnostic design solves the enshrinement barrier but introduces subjective enforcement surfaces that weaken transaction validity guarantees.”

Topic statement: “This article examines how LUCID structures encrypted transactions on Ethereum.”

The fourth step is the structure. The report opens with the problem the thesis addresses and the evidence gap that makes the problem unresolved. The body builds the argument in layers, i.e., definitions and mechanism first, then empirical evidence, then implications, then counterarguments. Each section must be necessary for the next section to hold. If a section can be removed without breaking the argument, it is either redundant or misplaced.

E.g.

A report on whether a second checkout lane reduces peak-hour wait times would open by establishing that customer complaints spike during peak hours and that no internal study has isolated lane count as the variable. The body would define peak hours and the measurement method, present timestamped queue data before and after the second lane was added, draw implications for staffing costs, and address the counterargument that a second lane increases labor costs beyond the revenue retained by shorter waits. If you remove the measurement definition, the data would be uninterpretable. If you remove the cost counterargument, the recommendation would be incomplete.

A report on whether concentrated liquidity reduces slippage for large stablecoin swaps would open by establishing that large swaps on constant-product AMMs lose value to slippage, and that no existing study isolates the effect of concentrated liquidity at that trade size. The body would then define concentrated liquidity and how it differs from full-range provision. It would then present execution data comparing slippage across trade sizes. This data can then be used to draw implications for high-value users or institutional traders. It will also be used to address the counterargument that concentrated liquidity introduces just-in-time liquidity risks that offset slippage gains. If you remove the definitions section and the reader cannot interpret the execution data. If you remove the counterargument, the conclusion would be one-sided.

The fifth step is prose discipline. Every sentence must carry one unit of information. If a sentence can be shortened without losing meaning, it should be shortened. If two sentences say the same thing in different words, the one lacking the most information should be cut. Things like transitional filler, hollow intensifiers, and restatements of prior paragraphs should be removed on the first edit pass, and then finetuned with each edit iteration.

E.g.

Initial statement: “It is worth noting that the company’s revenue grew by 12% year over year, which is a meaningful and significant increase that demonstrates robust growth”

Rewrite: “Revenue grew 12% year over year.”

Initial statement: “Ethereum’s gas fee mechanism is a really important and crucial part of the network that plays a significant role in how transactions are processed and validated on the blockchain”

Rewrite: “Ethereum uses gas fees to price transaction execution.”

The final check is adversarial reading. The completed draft should be read as a skeptic looking for three things: claims without primary source support, logical gaps between sections, and conclusions that exceed what the evidence permits.

Any finding from that read is either resolved with evidence or removed from the report. If your report survives adversarial reading, it is publishable.

A conclusion in a research report restates the thesis in light of the evidence presented, states whether the evidence supported or complicated it, and names what remains unresolved. It does not introduce new information. It does not summarize every section. It answers the research question with the weight of the body behind it.

A conclusion in an article performs a narrower function. It states what the reader should take away and, where appropriate, what to watch next. It does not restate the introduction. It earns its final sentence by landing on the single claim the article exists to make.

E.g

For report conclusions:

“Does mandatory calorie labeling on restaurant menus reduce average calories ordered per meal?”

Conclusion: “Across 23,000 tracked orders at 140 locations over six months, average calories per order fell 8% in the first month after labels were introduced. By month six, the reduction stabilized at 4%. Calorie labeling produces a measurable initial drop in calories ordered. The effect decays by half within six months, which suggests labeling alone is insufficient as a long-term intervention without complementary menu redesign.”

“Does Uniswap v3’s concentrated liquidity reduce slippage for stablecoin swaps above $100,000?”

Conclusion: “Across 14,000 USDC-USDT swaps above $100,000 between January and June 2025, concentrated liquidity positions within 1 basis point reduced median slippage by 62% compared to Uniswap v2’s full-range positions. The reduction held during low-volatility periods but compressed to 34% during the March 2025 depeg event, when liquidity providers withdrew from active ranges. Concentrated liquidity reduces slippage for large stablecoin swaps under normal conditions. Its performance under stress depends on whether liquidity providers maintain their positions, which remains an open variable.”

For article conclusions:

Article topic: Should Lagos add a second checkout lane at the public library?

Conclusion: “Wait times at the downtown branch dropped from seven minutes to under three after the second lane opened in September. Foot traffic rose 18% in the same period. The second lane solved the wait problem. Whether the branch can staff it past the pilot phase depends on the January budget vote.”

Article topic: Ethereum’s transition from proof-of-work to proof-of-stake.

Conclusion: “The Merge eliminated mining from Ethereum and cut the network’s energy consumption by over 99%. Validator participation stabilized above 95% within three months. The execution risk that dominated pre-Merge discussion did not materialize. The open question is whether the concentration of staked ETH among three liquid staking providers creates a censorship vector that proof-of-work did not have.”