The message arrives late, from an account with no photograph and a handle that looks like a typo. Three sentences, one of them a name, one a date, one a claim that would matter if it were true. Most tips like this go nowhere. A few become the best work you will ever do. The difference is rarely luck. It is a set of small, unglamorous habits applied in the same order every time, and it starts with resisting the urge to write anything at all.
Write Down the Tip Before You Interpret It
Save the message exactly as it arrived — screenshot, timestamp, the channel it came through, the precise wording. Do not clean up the grammar and do not paraphrase. Later, when the person on the other end denies ever contacting you, that file is the whole argument.
Then take the tip apart. Most messages contain three things bundled together: something the source saw, something the source heard secondhand, and something the source has concluded. Separate them into three columns and watch how quickly the solid part shrinks. "The invoices are fake" could mean forged signatures, duplicated payments, or a contractor billing for work that never happened. Those are three different stories with three different proofs, and only one of them may be reachable.
Keep a second list beside it, headed What we don't know. It will be longer. That list is your assignment sheet, and it is also the honest record of how much work stands between you and a publishable sentence.
Build the Timeline Before You Build a Theory
A chronology is the cheapest investigative tool you own. Build it as a table with four columns: date, event, where the date comes from, and confidence. Mark confidence plainly — documented, corroborated, single source, unconfirmed. The moment those labels sit next to each other in a spreadsheet, you can see the shape of your evidence without flattering it.
Ordering does the work. A company incorporated in February that signs a municipal contract in March is a different story from one incorporated in April that signs the same contract. Neither sequence is automatically damning; both raise a question you now know to ask. Gaps matter too. A two-year silence in a payment record, a meeting that appears in one diary but not another, a hiring freeze that coincides with an unusual purchase.
Resist the urge to smooth vague dates. If a source says "sometime in the spring," write "spring, exact date unconfirmed." A timeline that quietly tidies itself will eventually tidy itself in the wrong direction, and you will publish something you cannot defend.
Build a second timeline for your own reporting: who you contacted, when, on what channel, what they said, what you promised. That log becomes your protection and your memory.
The Verification Ladder
Facts do not get stronger through repetition. Move each one up a ladder and stop where it stops.
- Get the primary document. The contract itself, all pages including annexes, not the summary a friendly official read to you over the phone.
- Find a second, independent confirmation. Independent means a different route to the same fact, not a second person who heard it from the first.
- Check the public record. Registries, filings, tender notices, meeting minutes. These confirm existence, dates and roles, which is often enough to make a vague tip concrete.
- Put the claim to the institution. A denial, a non-answer, a silence — each is a recorded datapoint, and each must be reflected fairly in the story.
- Look for the physical or digital trace. A photograph with metadata, a receipt, a travel booking, an email header.
For every source, ask one question and then ask it again differently a week later: how do you know that? People who witnessed something answer consistently. People who absorbed a rumour tend to drift, adding detail where there was none. Note the drift. It is not proof of lying, but it tells you where the firm ground ends.
Care for the Source Like a Fact
You will be tempted to promise more than you can deliver. Don't. Anonymity is a newsroom decision, not a favour one reporter can grant on the phone. Explain plainly what can be protected — a name, a workplace, an identifying string of details — and what cannot. If your story only works by describing a role so specific that only one person could hold it, the source needs to know that before they speak, not after.
Agree one channel of contact and stick to it. Give them a realistic sense of timing, and tell them the story might not run at all, or might not say what they hope. Then do something most beginners skip: go back to them before publication with the exact claims you intend to attribute to them and confirm the wording.
Assume the source has a motive. Everyone does. Note it in the file, weigh it, and carry on — a motivated source is not automatically a wrong one.
Editorial Checkpoints Along the Way
Set three reviews and keep them, even when you are impatient.
At intake. Is there a story shape — a person or institution, an action, a consequence, someone affected? If not, you have a lead, which is a fine thing to have.
At the halfway mark. Could a colleague who knows nothing about the tip reconstruct your main claim from the documents alone? If they cannot, your structure is thin, not your evidence.
Before publication. Right of reply to everyone named or clearly identifiable, a careful read by an editor, and a written answer to one question: what single fact would sink this story, and have we genuinely tried to find it? If you cannot say what would prove you wrong, you have a belief rather than an investigation.
Where a story carries real risk of defamation, breach of confidence or harm to a source, take professional legal advice rather than reasoning it out in the newsroom. That is not caution; it is part of the craft.
When the Story Falls Apart
Sometimes the final check kills it. Shrink the story rather than force it. "The contract was awarded without a competitive tender" is a smaller claim than the one you started with, and it is publishable. Keep the file, note what failed and why, and revisit it when a new document surfaces. The story you cannot stand behind is worth less than nothing, because it will be used to discredit the next one you can.
A One-Page Template You Can Start Tonight
Keep a single document with five blocks: the claim in one sentence; the timeline table; the source list with what each person knows and how they know it; the verification to-do list with dates attached; the open questions. Update it before you stop work each day.
Then test it. Read only the first block aloud. If you cannot state the claim in a sentence that is both specific and true to your evidence, you are not ready to report it out — you are ready to keep asking questions. That is where every good investigation actually begins.
Photo: KenamiRyoko / Pixabay
