What Verification-Grade Means

~4 min

The standard you are judged against

Every research delivery is judged against one standard. Every requested field is populated or explicitly marked unavailable after a real search, and figures are traceable to a source where the brief asked for sourcing. Isolated inaccuracies in publicly-sourced data are normal variance; fabricated findings are not. Read it twice. It does not ask for perfection. It asks for three things: no empty cells without a reason behind them, a trail back to a source whenever sourcing was requested, and nothing invented. This course teaches you to meet that standard on every task you claim.

Variance is forgiven, fabrication is not

Public data is messy. A phone number changes the week after you record it. A directory spells a name wrong and you carry the error forward. If you ran a real search and recorded what a source actually said, that is variance, and we treat it as normal. Fabrication is different: writing down something no source ever told you. A guessed email, an estimated price, a role assumed from a job posting. The operator cannot tell your 295 real findings from your 5 invented ones, so one discovered fabrication makes the whole file suspect. Variance costs a correction. Fabrication costs trust in every row you have ever delivered.

Why the bar is this high

Research deliveries get used, not just read. A client who orders owner emails for 300 dental clinics is about to send 300 messages. A wrong email bounces harmlessly. An invented one can reach a stranger, embarrass the client, or poison a mailing list. A pricing sheet feeds a pricing decision. A verification file decides which records stay in a database. You rarely see what happens after delivery, so treat every cell as something a real person will act on tomorrow morning. That is what verification-grade means: not academically perfect, but safe to act on.

Remember

The standard forgives sourced inaccuracies but never invented findings.
Every cell is either populated or marked unavailable after a real search.
One discovered fabrication makes every row in your file suspect.
Clients act on research; treat every cell as something used tomorrow.

Primary and Secondary Sources

~4 min

What primary means here

A primary source is the entity itself speaking about itself. The clinic's own website is primary for its address. A company's own pricing page is primary for its prices. A person's own LinkedIn profile is primary for the role they claim. An official registry is primary for registration facts. Primary does not mean guaranteed true, since a company page can be outdated, but it means no one retyped or summarized the fact on its way to you. When the fact belongs to the entity, the primary source is where the fact lives.

What secondary means, and its uses

A secondary source repeats what someone else found: directories, aggregator sites, news articles, comparison blogs, cached lists. Secondary sources are fast and wide, which makes them good for discovery. When you need to locate 300 clinics, a directory hands you the list in minutes. But every step away from the primary source adds a chance of error: old scrape dates, typos, merged records. Use secondary sources to locate candidates and to corroborate, not as the final word on facts the client will act on.

When each is enough

Match the source to the stakes. A low-stakes field with no sourcing requirement, such as a clinic's city or general specialty, can rest on one reasonable secondary source. A field the client will act on directly, an email they will write to or a price they will position against, deserves the primary source or corroboration. Pricing is a special case: it changes without announcement, so a secondary source that was right last quarter can be wrong today. When in doubt, ask what the client does with this cell. The more they will do, the closer to the source you go.

Remember

Primary means the entity speaking about itself; nothing retyped on the way to you.
Secondary sources locate candidates and corroborate; they are not the final word.
Match source strength to stakes: the more the client will act, the more primary you go.
Pricing changes silently; only the company's own page is current.

The Two-Source Rule

~5 min

Two independent sources for facts that matter

For any fact the client will act on, find it in two places that did not copy each other. If a directory and the clinic's own site agree on an email, record it with confidence. If they disagree, the primary source usually wins, and you note the conflict if the brief asked for sourcing. One strong primary source about itself, such as a company's own pricing page for its own prices, can stand alone; corroboration adds little there. The rule exists for everything else: names, emails, roles, and figures that passed through other hands before reaching you.

Independence is the whole rule

Two sources only count if they are independent. Dozens of directory sites resell the same data feed, so finding a phone number on three of them is finding it once. Signs of a shared feed: identical wording, identical mistakes, the same odd formatting. Real independence looks like a directory plus the clinic's own site, or a news article plus a registry. Before you count a second source, ask where it likely got the fact. If the honest answer is from the first source, keep looking.

Capture evidence as you go

When the brief asks for sourcing, a fact without a trail does not meet the standard. The trail is simple: the source URL and the date you saw it. Capture both at the moment you record the fact, not at the end of the task. Pages change and disappear, and the date tells the operator the fact was true as of that day, which protects you when data shifts after delivery. Follow the format the brief specifies; if it specifies none, a source column and a date column in the sheet is enough. Traceable is the word in the standard, and this is what it costs: one paste and one date per fact.

Remember

Facts the client acts on need two sources that did not copy each other.
Three directories reselling one feed count as one source.
A primary source speaking about itself can stand alone.
Record the source URL and the date seen the moment you find the fact.

Current, or Published Once

~4 min

The web does not age visibly

A page written six years ago looks exactly like a page written yesterday. Footers showing this year's date auto-update and prove nothing. The question behind every fact you pull is not whether it was published but whether it is maintained. A staff page nobody has touched in four years still lists the doctor who left. A pricing table can survive three price changes without anyone correcting it. Verification-grade research treats every undated fact as a claim of unknown age until something tells you otherwise.

Reading freshness signals

Look for evidence the page is alive: recent posts, dated updates, prices that match a current promotion, staff listings consistent with recent activity elsewhere. Then look for signs of neglect: a news section that stops years ago, broken links, past events described as upcoming, copyright ranges ending in the past. No single signal is proof; each one shifts your confidence. A fact from a clearly maintained page carries more weight than the same fact from a ghost site, and a stale signal is a cue to corroborate somewhere fresher before you record.

Your delivery is a snapshot

Everything you deliver is true as of the date you saw it, no more. That is not a weakness; it is what the date-seen column is for. Recording dates turns this is the owner's email into this source said so on this date, which is a claim you can always stand behind. The standard calls sourced inaccuracies normal variance partly for this reason: data drifts. Your job is not to promise the future. It is to report what maintained sources say today, and to mark the today.

Remember

Published once is not the same as current; footer years prove nothing.
Judge whether a page is maintained before trusting its facts.
A stale signal means corroborate somewhere fresher before recording.
A date-seen column turns findings into claims you can always defend.

Discovery Versus Verification

~5 min

Two task shapes, two methods

Discovery asks you to find something that may exist: owner emails, competitor prices, a contact name. Verification asks whether something already claimed is still true: does this profile still match this role, is this firm still at this address. The shapes differ. Discovery starts wide, using secondary sources to generate candidates, then narrows to confirm. Verification starts narrow, at the claim itself, and goes straight to the primary source to test it. Using the discovery method on a verification task wastes hours; using the verification method on a discovery task finds nothing.

Discovery: wide, then confirmed

In discovery, secondary sources are your map and primary sources are your proof. Cast a wide net across directories, search engines with varied phrasings, and the entity's own site, then confirm the candidates that matter against primary sources. The first failure mode is stopping at the map: delivering the first directory hit without confirmation. The second is not knowing when to stop. A real search has an endpoint, and once the reasonable avenues are exhausted, unavailable is a finding. The next lesson covers exactly where that line sits.

Verification: the claim is the starting point

In verification, each row is a claim to test, and your deliverable is a status: confirmed, changed, or could not confirm. Go to the source that owns the claim, the profile itself or the company's own page, and compare. Record what you saw and when. Resist two temptations. Do not quietly overwrite changed data unless the brief asks for updates; the client asked whether their data is still true, and changed is the answer they are paying for. And do not stretch could not confirm into confirmed because most other rows were fine. Could not confirm is a real status, not a failure.

Remember

Discovery starts wide and narrows; verification starts at the claim and tests it.
In discovery, secondary sources map and primary sources prove.
Verification deliverables are statuses: confirmed, changed, or could not confirm.
Never quietly overwrite changed data; changed is the answer being paid for.

Unavailable, Honestly

~6 min

Unavailable is a finding, not a failure

The standard says every field is populated or explicitly marked unavailable after a real search. That last clause is a license: unavailable is a legitimate deliverable. Some owners publish no email. Some companies hide pricing behind a sales call. A file with honest unavailable rows is complete; a file with guessed rows is not. What makes unavailable honest is what stands behind it, meaning the searches you actually ran. What makes it dishonest is using it to skip hard rows, or avoiding it by inventing something to fill the cell.

What a real search looks like

Before writing unavailable, you should have tried the reasonable avenues: the entity's own site including contact and about pages, a search engine with several phrasings, the obvious directories for that industry, and the platform where the fact would naturally live. For a person's role, that includes their professional profile. Keep a short note of what you tried; three or four search descriptions per stubborn row is enough. When the operator sees checked site, searched three phrasings, checked two directories, no owner email published, the unavailable mark is credible and defensible. A bare empty cell is neither.

The extrapolation line

Here is the line that matters most. You know forty owner emails follow the pattern of first name at domain. Row forty-one has no published email, but you know the owner's first name. Typing the pattern into the cell feels like insight. It is fabrication. No source told you that address exists; you did. A plausible inference presented as a finding is an invented finding under the standard. If a pattern seems worth mentioning, put it where it belongs: in your delivery note to the operator, labeled as a pattern, never in the data. The cell gets unavailable. The note gets your reasoning. The operator decides what to do with it.

When the brief itself is unclear

Uncertainty about the brief is handled the same way as uncertainty about a fact: flag it, do not guess it silently. If a term could mean two things, pick the reading that best fits the task, apply it consistently, and say so in your delivery note: which definition you used and which rows it affects. The operator can accept it or request a revision with the definition corrected, and either way your record shows judgement rather than concealment. A delivery that hides its assumptions gambles the whole payout on a coin flip. A delivery that states them is safe even when the assumption was wrong.

Remember

Unavailable after a real search is a complete, legitimate answer.
Log the searches behind every unavailable mark; a bare empty cell defends nothing.
Patterns go in the delivery note as patterns, never in the data as findings.
State assumptions in the delivery note; hidden assumptions gamble the whole payout.

Research

After this course

Everything, and what is in it.

What Verification-Grade Means

The standard you are judged against

Variance is forgiven, fabrication is not

Why the bar is this high

Primary and Secondary Sources

What primary means here

What secondary means, and its uses

When each is enough

The Two-Source Rule

Two independent sources for facts that matter

Independence is the whole rule

Capture evidence as you go

Current, or Published Once

The web does not age visibly

Reading freshness signals

Your delivery is a snapshot

Discovery Versus Verification

Two task shapes, two methods

Discovery: wide, then confirmed

Verification: the claim is the starting point

Unavailable, Honestly

Unavailable is a finding, not a failure

What a real search looks like

The extrapolation line

When the brief itself is unclear

The courses are free. The work is real.

The curriculum