Does Your Film Need an Archivist, a Historical Adviser, or More Reporting?
Does Your Film Need an Archivist, a Historical Adviser, or More Reporting?
There is usually a sentence in a pitch that nobody on the team can check. Sometimes it's the opening image: we begin on the workshop's last working day. Sometimes it's a claim about how people lived: the apprentices were practically family. Sometimes it's a cause: the place closed because the owner retired. Everyone has heard something along those lines. Nobody can put the evidence on the table.
The reflex is to add a line to the budget marked "research" and go looking for a person whose title sounds right. That's where the trouble starts, because a title tells you very little about what you'll get back in six weeks.
Turn "we need research" into a decision
Take the uncertain sentence and ask what changes if it's wrong. Not "would we like to know more" — what in the film is different?
Three separate problems hide under one sentence, and they call for different work.
It might be contradicted. The account you have is one of several, and the others left documents.
It might be thinner than the pitch implies. Something happened, but not in the form or at the scale you've described, and the claim has to shrink or soften.
It might be undocumentable. Not false — just unavailable. Nobody photographed the last day, nobody wrote it down, and no one living can describe it. That isn't a lie in the pitch. It's a shot you can't have.
Only the first of those is what people usually mean by fact-checking. All three change the film.
So rewrite the topic as a question with an edge. "Find anything about the workshop" is not a commission; it's an atmosphere. "Establish whether the proposed final working day can be documented, and if so, by what" is a commission, because it can come back no.
Two rules for writing it down.
Write it so the answer can go against you. "Confirm that the workshop closed the way we've described" is not a question. It's a request for a signature, and anyone willing to supply one cheaply is telling you something about themselves. If you've already decided and want a second pair of eyes, say that instead, and pay for that.
Don't buy what you won't use. If both answers leave the film exactly as it is, you don't have a research question; you have curiosity. That's fine, but it's a different purchase. A topic absorbs money forever, because there's always one more folder. A question closes.
Separate the work before choosing the person
Five tasks travel together under the word "research," and they answer different questions.
Locating material. Does a thing exist, and where is it now? A photograph, a reel, a minute book, a planning file, an interview that was recorded and never transcribed.
Establishing provenance. What is this thing, when was it made, who made it, and how did it end up here? Finding a reel does not date it. Labels get written by hand, years later, by someone repeating what they were told.
Interpreting context. What did a word, a role, a payment, or a custom mean at that time and in that place? What "apprentice" meant in one workshop's wage books may not be what it meant in the shop across the road, and almost certainly isn't what it means in a pitch written this year.
Gathering new testimony. Finding people who were there, interviewing them, and reporting what they say, including where they contradict each other.
Reviewing a portrayal. Commenting on a script, a cut, a sequence, a costume, a line of dialogue. "That's the wrong decade for this technique" is a different remark from "this is not how people in that trade talked to each other."
The failure mode isn't hiring the wrong person. It's handing one person a question from each column and assuming the answer comes back joined up.
Someone who finds the reel may not be able to date it. Someone who can explain a term from wage books won't produce a witness. A witness can tell you what happened on a Tuesday; they usually can't tell you whether it was typical, and asking them to is asking for something they don't have. A person who has read everything about a region may be exactly right to tell you your kiln is the wrong decade, and exactly wrong to say whether a scene rings true to the people it depicts.
Titles won't sort this out, because the same words can cover different work in different places and institutions. One "historical adviser" spends a production writing script notes; another spends it chasing archive permissions. What matters is demonstrated work. Ask a prospective collaborator to describe a job they finished: what they were asked, what they brought back, what they couldn't get.
Three questions are worth asking before you commission:
- What would you do first, and what would you need from us to do it?
- What would you tell us if the trail went cold?
- Is any part of this outside what you normally do?
A good answer to the third is a relief, not a problem. Someone who tells you where their competence ends knows where it ends.
Agree on a return another person can use
A brief that produces something reviewable has six parts. The order matters less than the completeness.
The question. One sentence, with an edge in it.
What's already been looked at. Not to control the search — to save the first week, and to let the searcher tell you whether the earlier look was adequate. Sometimes the useful reply is, "You searched the catalog, but this accession isn't cataloged yet."
The access you can provide. An introduction, a letter, a membership you hold, a site visit arranged with a family. Be honest about the ones you don't have. If the whole assignment depends on someone being let into a loft, say so in the brief, because it changes what can be promised.
The deliverable. Not "a report." Name the shape:
- item-level leads — holder, collection, reference, what the catalog claims, what nobody has checked;
- an attributed account — who said what, when, and what sits uneasily beside it;
- comments tied to specific pages or timecodes.
Exclusions. What this is not. Not a rights clearance, not an interview series, not a review of the whole script, not a statement about how an entire trade worked.
How a dead end reaches you. A failed search is a result. Agree how and when it's reported, and set a checkpoint on long searches — not to supervise, but because two empty repositories may mean you should redirect the money.
Then the practical matters, which should follow the work rather than the title. A two-week provenance note and a six-month interview program are not the same job, and the fee, the credit line, and the number of revision rounds should track the job. If you want the option to ask follow-up questions, say how many and whether they're included. Expanding a dating check into "and could you also talk to the family" without agreeing it first is how a working relationship goes bad.
One workshop, three briefs
What follows is invented — a paper exercise, not a case study. No workshop, researcher, or finding here is real. But it's built the shape a real one would be.
Imagine a documentary pitch about a pottery workshop that shut down. Three sentences carry the risk:
- "We open on footage of the workshop's last working day."
- "The workshop trained its apprentices by hand, one at a time, over years."
- "It closed in 1994 when the last owner retired."
Here's the brief the team writes first.
Research the workshop — history, any photographs or footage, how the training worked, why it closed. Six weeks. Also check the script for accuracy.
You can't accept this return or reject it. If it comes back thin on footage, you won't know whether that's because no footage exists or because five of the six weeks went into script notes. You can't hand it to two people and compare answers. And it quietly asks one person to sign off three separate sentences.
Three scoped briefs instead.
Brief one: locate.
Question: Is the workshop's final working period recorded on film, video, or stills, and if so, where is the material now?
Access provided: permission to search the regional archive's uncataloged accessions; an introduction to the family that bought the site.
Return expected: item-level leads. For each — the holder, the reference where one exists, what the catalog or the holder says the item shows, and separately what has not been established. Plus a list of what came up empty and what a further search would require.
Outside this assignment: dating the material, viewing and assessing it, identifying people in it, clearing any right to use it. Locating is the first link in a chain; this brief buys the first link and says so.
Brief two: context.
Question: In the workshop's own records, what did the word "apprentice" cover?
Access provided: the wage books and minute books, on site, with the archivist's supervision.
Return expected: an attributed account of how the term was used here — who it was applied to, what they were paid or not paid, how long they stayed — with the limits stated plainly: which years the records cover, which are missing, and where the usage is loose enough that no single answer holds.
Outside this assignment: producing a witness, generalizing to other workshops, supporting any line that speaks for a whole trade.
Brief three: testimony.
Question: How did work at the workshop end, according to people who were there?
Access provided: introductions to four former workers who have agreed to be approached; two more have not answered.
Return expected: attributed testimony — each account tied to a named person and a date, disagreements reported as disagreements rather than averaged into one story, and a note of who declined or couldn't be reached.
Outside this assignment: speaking for former workers as a group, and establishing the cause of the closure as a settled fact.
Now watch what each return can move.
Brief one lands on sentence 1. If the leads turn into located, datable material, the opening can stay as written — provided someone takes on the dating and the rights, which this brief didn't. If the return is unconfirmed leads and nothing located, the opening changes: the film starts on a photograph, or on a voice, or on the building as it stands now.
Brief two lands on sentence 2. It may support the pitch as written. It may show the wording has to change, because the term covered something broader or narrower than the pitch implies. Or it may come back saying the records don't answer the question, and the sentence rests on weaker ground than you assumed.
Brief three lands on sentence 3. If the accounts disagree — one describing a retirement, another describing a sale — the disagreement is filmable, and probably better than a single tidy cause. What the assignment can't do is pick one and call it the reason.
These three jobs can be one person, two, or three. That's a scheduling and money question, not a professional-title question. If one person genuinely does all three, write it down and pay for three jobs. Combining them because a title sounds broad is how you end up with a long email and nothing usable.
Treat the finding as work, not a seal of approval
When the work comes back, read it against the question you asked rather than the sentence you hoped to keep.
Ask three things of every return.
What can the pitch now change? If the answer is nothing, either the question was decoration or you haven't absorbed the return yet.
What still needs a different assignment? A located reel still needs dating, identifying, and clearing. One interview isn't a picture of how something ended. A contextual account of a single workshop can't carry a claim about a whole region.
Where does the uncertainty have to stay attached? If the return says "probably 1994; the label is inconsistent," the film can't say "the last working day, 1994." It can keep the "probably," or it can buy the dating work. What it can't do is keep the number and drop the hedge.
Three confusions are worth naming, because they turn up on nearly every project.
A portrayal review is not a community's approval. If you show a scene to one former worker and they don't object, you have one person's response. That's useful. It isn't consent from anyone else, and no assignment can manufacture consent you'd have to obtain another way.
A credit is not an endorsement. A name in the research line means that person did that work. It doesn't mean they stand behind every scene, and it doesn't mean they reviewed the final cut.
Research help is not clearance. That a reel exists, that it's from the right year, and that you're allowed to use it are three separate findings, and locating the reel produces only the first. Rights work is a different job with different people. A researcher who tells you where something is has not told you what you may do with it.
What you're buying is work: a question pursued, an answer that can come back negative, a set of uncertainties you now know by name. What you're not buying is authority over the story — not from a title, and not from a name on the team page.
Write the assignment before you hire
The useful thing to write down, once, before anyone starts, is short:
- the unresolved question;
- the decision it changes;
- the evidence being sought;
- the return, in the shape you'll actually read it;
- what stays outside.
Five lines. If you can write them, you know which work to commission, and you'll recognize a good return when it arrives. If you can't write them, no amount of title-shopping will help, because you'd be asking one person to decide the film for you — and nobody can do that from the outside.
Frequently asked questions
How do I turn a vague need for research into a decision?
Take the uncertain sentence in the pitch and ask what changes in the film if it is wrong. The body separates three problems: the account might be contradicted by other documents, it might be thinner than the pitch implies, or it might be undocumentable. Rewrite the topic as a question with an edge, one that can come back no. If both answers leave the film unchanged, it is curiosity, not a research question.
What kinds of work hide under the word research, and why can one hire go wrong?
Locating material, establishing provenance, interpreting context, gathering new testimony, and reviewing a portrayal. The failure mode is not necessarily hiring the wrong title but handing one person a question from each column and assuming the answers join up. Ask a prospective collaborator to describe a finished job: what they were asked, what they brought back, what they could not get.
What should a research brief contain?
The body lists six parts: the question, what has already been looked at, the access you can provide, the deliverable in a named shape, exclusions, and how a dead end will reach you with a checkpoint on long searches. Practical terms such as fee, credit, revision rounds, and follow-up questions should follow the job rather than the title.
In the invented workshop example, how do the three briefs relate to the pitch?
Locate lands on the claim about the last working day, context lands on what apprentice meant in the workshop records, and testimony lands on how the work ended. If accounts disagree, the disagreement is filmable and probably better than a single tidy cause, but the assignment cannot pick one cause as settled. The three jobs can be one, two, or three people, but if one person does all three, write it down and pay for three jobs.
How should I judge a research return?
Ask what the pitch can now change, what still needs a different assignment, and where uncertainty has to stay attached. If a return says probably a date and the label is inconsistent, the film cannot keep the number and drop the hedge. A portrayal review is not a community's approval, a credit is not an endorsement, and research help is not clearance.