TL;DR
- The core decision: whether the headcount plan is a list of manager requests or a model with named drivers, so that disagreement with finance lands on an assumption rather than on a role.
- When doing nothing is right: when finance already accepts the current process, your last cycle's plan was not cut by surprise, and the cost of rebuilding the model exceeds the cost of being overruled again.
- What has to be true: someone on the HR side can read the plan aloud and explain why every number in it's what it's, without asking the requester.
- How the options split: top-down envelopes, aggregated bottom-up requests, ratios to revenue or customers, driver-based models, zero-based rebuilds, and rolling re-forecasts. Each is honest for a different reason and dishonest for a different reason.
- Decision rule: if a change in revenue or volume changes the headcount, you've a model. If it doesn't, you've a list, and the list will lose.
- Outcome to expect: finance stops asking how many and starts asking which assumption you're using. Some of those conversations you win. The ones you lose are smaller than the cuts you take today.
The Cut That Came Back
A compensation director at a mid-market professional services firm opens her inbox on a Tuesday in late October. The CFO has sent back the headcount plan for the new financial year with a single line of commentary and a number. The number is the count of approved roles. The line is that it's the number the firm can afford. She looks at the plan she spent six weeks building and realises she can't defend it. Every line in it's a request from a manager, with a justification written by the manager, in the manager's own language. When the CFO asked her why a particular role existed, she had to call the manager to find out.
She walks into the next meeting with the spreadsheet and a stack of papers, and the meeting lasts forty minutes, most of which is the CFO pointing at rows and asking why. She answers some of the questions and defers on others. By the end of the call, the count has moved. The CFO is happy because the number is lower. The managers are unhappy because three of their roles disappeared without a reason they were given. She is unhappy because she has lost control of the conversation and will lose it again next year.
But the real issue isn't that finance cut the plan. Finance will always cut the plan. The real issue is that the plan she brought to the meeting had no internal logic, so the only available response to a cut was to accept the cut, and the only available defence of a role was a sentence written by the person who wanted it. A plan built as a model behaves differently under pressure. When the driver changes, the number changes with it, and the conversation moves from how many roles to which assumption you disagree about. That's a conversation HR can win, or at least survive intact.
Best tools for HR Strategy
When the Current Setup Is Genuinely Fine
Some HR functions don't need to rebuild anything. The first honest case is the function whose last plan was approved close to submitted, with the changes traceable to named disagreements, and where the same process has worked two cycles in a row. In that company the head of people walks into the finance review with the spreadsheet and a short list of asks, finance approves most of it, and the people who disagree about specific roles talk to each other afterwards. The model is in the room, not on paper, and it works because the same two people have done this three times and understand each other's defaults. If that describes your situation, write down what works, why it works, and who has to stay in the room for it to keep working, because the model lives in two people's heads and one of them will leave.
The second case is friction that's real but not structural. A team of sixty in a regional retailer has a plan that gets cut every year by a fixed amount, and the cut lands close to the same departments each cycle. Operations takes more, support takes less, and the people director has stopped fighting the pattern because the pattern is predictable. This isn't a good plan. It's a stable one. If the cuts are survivable and the manager base has accepted them, rebuilding the model would consume political capital that the function needs for other fights. The honest move is to leave it, document the implicit rule, and rebuild only when the cut pattern changes or the organisation does.
The third case is genuine risk that has not yet surfaced. A scale-up that grew from ninety to three hundred people in eighteen months is still submitting plans the way it did at ninety. Finance tolerates the imprecision because the company is hiring into revenue growth that's easier to read than to forecast. The day growth slows, the plan will be the first thing finance re-examines, and HR will discover that it has no model to defend. This case looks like the first one from inside. It isn't. The signal is the gap between the size of the company now and the size of the company when the current process was built.
The fourth case is the edge. A holding company with four business lines, each on a different planning cycle and a different finance contact, runs four versions of this fight in parallel. There's no single plan to rebuild, and no single finance team to argue with. The work here's to pick one line of business where the model rebuild has a chance of being read, run it as a pilot, and decide whether to extend. Anyone who promises to fix all four at once is selling something.
What an HR Leader Asks Themselves at 11pm
Why did finance cut the plan and not the marketing budget? Because the plan had no internal logic, so the cut was free. Marketing budgets carry assumptions about pipeline, conversion and revenue that finance can interrogate. Headcount plans carry manager preferences. The reasoning behind it's that a budget line defended by a number is harder to cut than a budget line defended by a sentence. If you get it wrong, finance treats your function as a cost to be reduced rather than an investment to be shaped.
Can I name the three roles I would cut if I had to? If you can't, you don't own the plan. You're the postman. The reasoning is that a plan you can't cut yourself is a plan you're asking finance to cut for you, and they will do it more bluntly than you would. If you get it wrong, you spend the next cycle explaining to managers why a role they cared about vanished, and you've no framework to point to.
What changed between when I built the plan and when finance reviewed it? Usually nothing, which is the answer finance uses to take the plan apart. The reasoning is that a plan without a sensitivity analysis is a plan that assumes the world is exactly as it was on submission day. If you get it wrong, you treat a forecast as a fact and defend it as such.
Do my managers know what their request looks like inside the company total? Almost never. The reasoning is that a request written in a vacuum will be written optimistically. If you get it wrong, the plan carries an embedded inflation that finance reads instantly and discounts aggressively, and the real conversation about real roles happens in a smaller, angrier room.
What is the smallest rebuild I could do that would change the next conversation? Usually a one-page model that links three drivers to three role counts. The reasoning is that perfection is the enemy of a defensible plan, and a small model next cycle is worth more than a complete one next year. If you get it wrong, you spend the year building a cathedral and walk into the finance review with the same list of wishes.
Who in finance is the actual reader of this plan? Often someone junior who has been told to deliver a number. The reasoning is that a plan aimed at the CFO reads differently from a plan aimed at the controller, and most plans aim at the wrong reader. If you get it wrong, your best argument lands on someone who can't act on it.
Three Honest Categories the Approaches Split Into
The aggregation play. Take what managers want, normalise the job titles, add a layer of judgement about which requests are credible, and present the result as a plan. This is the default in most mid-sized firms, and it's honest when manager judgement is good, when the manager base has a shared view of what a role is worth, and when finance is willing to accept the aggregation as the starting point. It fails the moment manager judgement diverges from finance judgement, because there's no shared framework to argue inside. A regional services firm that ran this approach for four cycles had a plan that grew every year while revenue flattened. The HR business partner who carried the plan into the finance review had to defend each role on its own merits, and the conversation became a sequence of micro-fights she could not win.
The financial play. Take a target cost or a ratio and divide it down. Finance sets the envelope, HR distributes the slots, and the conversation is about placement rather than existence. It's honest when the cost target is right, when the work the organisation does is roughly stable in shape, and when the function doesn't need to reshape the workforce through the plan. It fails the moment the cost target is wrong, because every role in the plan inherits the wrong assumption. A manufacturing group that ran headcount as a ratio of revenue for three cycles cut its way into a workforce that could not staff a product launch, and the post-mortem took longer than the rebuild would have.
The structural play. Take the work the organisation has to do, model the volume and the time, and derive the headcount from the model. This is the rebuild people describe when they say they want to do workforce planning. It's honest when the model reflects the work, when the volume assumptions are owned and reviewed, and when the function has the time to build and defend the model. It fails the moment the model is owned by one person who leaves, or the volume assumptions are owned by finance and never revisited. A software company built a beautiful driver-based model that mapped every role to a feature, and the model broke the day a feature was killed. The next cycle took three months to rebuild and the cycle after that never finished.
Five Diagnostic Questions You Can Answer About Your Own Organisation
Can you explain why every role in the plan is in the plan, without phoning the requester? If yes, you've a model. If no, you've an inbox. To answer it for real, open the plan, pick three roles at random, and write down the justification for each without looking at the request. If you can do it for two out of three, you're closer than most. If you can't do it for any, the plan is the manager's, not yours.
Does a change in the revenue or volume assumption change the headcount number? If yes, you've a model. If no, you've a list. To answer it, take last year's plan, halve the revenue assumption, and see what changes. If nothing changes, the plan doesn't depend on the assumption finance cares about, and finance is right to disregard it.
Where in the plan is the inflation? Every plan carries inflation. The question is whether you can see it. To answer it, take each manager's request, mark which roles would survive a moderate cut and which wouldn't, and ask whether the surviving roles are the ones the company needs most. If the answer is yes, the manager is being honest. If the answer is no, the manager is hedging, and you're carrying the hedge into finance.
Who in your function can defend the plan if you're not in the room? If the answer is nobody, the plan lives in your head and your function is one resignation away from losing the argument. To answer it, hand the plan to a peer for an hour and ask them to walk it through the finance review. If they can, the plan works. If they can't, the plan is fragile in a way finance will eventually exploit.
What did finance cut last year, and why? If you can answer that question with a specific driver, finance has told you what model they want. If you can't, finance has cut without explaining and you've accepted the cut without challenging. To answer it, go back to last year's email trail and find the cut. If there's a reason attached, use it. If there's not, the rebuild starts by asking finance what they would have asked for.
Six Ways to Build the Number, Reviewed
Top-Down Envelope Set by Finance and Divided Up
The envelope is the cost. Finance sets the total, often as a ratio of revenue or a growth assumption, and HR distributes the slots. The plan is fast to build because the hard number is given. It's right in a downturn, when the question is what to cut, and when the work of the organisation is stable in shape and the workforce is mature. It falls short the moment the envelope is wrong, because every role inherits the wrong assumption, and the function has no argument when finance chooses a ratio that doesn't fit the business. It also falls short when the function needs to reshape the workforce, because the envelope is sized to last year's shape and locks in last year's cost structure.
Bottom-Up Manager Requests Aggregated
Each manager submits what they want. HR aggregates, normalises titles, and applies judgement about which requests are credible. The plan is right when manager judgement is good, when the manager base shares a view of what a role is worth, and when finance is willing to treat the aggregation as a starting point rather than a ceiling. It falls short when manager judgement diverges from finance judgement, because there's no shared framework to argue inside. The aggregation also carries embedded inflation, because managers hedge by asking for more than they expect to get, and finance reads the hedge and discounts it, and the conversation becomes about who is more cynical rather than who is right.
A Ratio to Revenue or to Customers
The headcount is a function of a single number, usually revenue, sometimes customer count. The plan is right when the business is steady, when the relationship between revenue and headcount is real, and when the function needs a quick check rather than a full rebuild. It falls short the moment the ratio is wrong, which is most of the time, because the ratio is a snapshot of one moment in one company and assumes the moment is typical. It also falls short when the function needs to reshape the workforce, because the ratio locks in the shape.
Driver-Based Modeling From Volume Assumptions
The plan is built from the work. Each major volume driver is named, its assumption is owned, and the headcount is derived from the driver. The plan is right when the work of the organisation is measurable, when the volume drivers are owned by people who will defend them, and when the function has the time to build and maintain the model. It falls short the moment the model becomes the plan, because the model is a representation and the representation drifts. It also falls short when one person owns the model and leaves, and the next cycle starts from nothing.
A Zero-Based Rebuild From the Work Itself
Every role is questioned, every role has to be re-justified, and the plan is built from the work the organisation has to do in the next cycle. The plan is right when the organisation is reshaping, when the existing workforce is no longer the right shape, and when the function has the political capital to ask every role to defend itself. It falls short the moment the rebuild becomes a project, because the project becomes a target for the people whose roles are at risk. It also falls short when the function can't run the rebuild and the day job at the same time, and the rebuild lands late and under-resourced.
A Rolling Re-Forecast Revisited Each Quarter
The plan is a living document. It's built once a year and revisited every quarter, with changes traceable to named drivers. The plan is right when the business is volatile, when the function wants the conversation with finance to be continuous, and when the work of revising is cheaper than the work of starting over. It falls short when the function doesn't have the discipline to revisit, because a re-forecast that never happens is a plan that goes stale. It also falls short when finance wants a single number for the year, and the rolling approach is read as indecision.
The Decision Table
| Situation | Scale | Setup | Primary Pain | Recommended Starting Point |
|---|---|---|---|---|
| Last cycle was approved close to submitted, with traceable disagreements | Any | Process works in the room between two people | Process depends on continuity of those two people | Document the implicit model, identify the assumption finance reads |
| Plan is cut every year by a fixed amount, pattern is predictable | Mid-sized | Cuts land close to the same departments each cycle | Implicit rule has not been written down or challenged | Write the implicit rule down, decide whether to defend it or redesign it |
| Scale-up that grew fast and still plans the way it did at small size | Growing fast | Single planning rhythm, single finance contact | Plan has no internal logic, will be exposed when growth slows | Build a one-page driver model on the largest function first |
| Holding company with multiple business lines on different rhythms | Multi-line | No single plan to rebuild, no single finance team | No place to start | Pilot one line of business, run a full cycle, decide whether to extend |
| Workforce needs reshaping and current plan locks in old shape | Mid-sized | Existing workforce no longer fits the work | Plan protects roles the business no longer needs | Zero-based rebuild, scoped to the functions that need reshaping |
| Revenue is uncertain and the year will be re-forecast several times | Volatile | Plan is a forecast of a moving target | Finance wants a single number and the business cannot supply one | Rolling re-forecast with named drivers, owned by one person |
| Function is mature and work is stable, but no model exists | Mid-sized | Same shape every year, no driver-based logic | No defence for any specific role | Ratio to revenue as a starting point, with an honest note that the ratio will not hold |
The Conversation With Finance
The point of presenting a plan as a model is to make disagreement land on an assumption rather than on a role. The CFO can disagree with a volume assumption. The CFO can't disagree with a role without disagreeing with the work the role exists to do. Move the conversation upstream.
| Assumption | Who owns it | What would change it | What happens to the number if it moves |
|---|---|---|---|
| Revenue growth rate for the relevant business line | Finance, with the commercial lead | A revised forecast, a new contract, a lost customer | At a ratio-based company the headcount number moves with revenue by the same ratio, automatically |
| Productivity assumption per role | The function that owns the role | A tooling investment, a process change, a new system | A change in productivity assumption changes the headcount needed for the same work |
| Scope of work in the next cycle | The senior leader who owns the function | A new product, a new geography, a deprioritised initiative | A change in scope changes the work and therefore the headcount, independent of the revenue assumption |
| Cost per role | HR, with finance | A change in the salary band, a change in location mix, a change in the role shape | A change in cost changes the number of roles the envelope can afford |
| Manager judgement about which requests are credible | The HR business partner | A change in the manager, a change in the manager's context | A change in judgement changes which roles survive the cut |
Bring this table to the finance review. Walk it through, line by line, and ask which assumption finance disagrees with. If finance can name one, you've a conversation. If finance can't, the cut is free and you've lost.
When the Plan Meets a Freeze
A freeze lands in the middle of a cycle. The plan exists, some roles are in flight, some offers are out, and finance has stopped signing off new hires. The work is to decide which commitments you honour and which you unwind, and to do it visibly.
Roles with a signed offer and a start date are the hardest to unwind. The candidate has resigned from a previous role and the goodwill cost of pulling the offer is real. If the role is genuinely discretionary, the cost of unwinding is a financial number and a relationship cost. If the role is essential to a regulatory or contractual obligation, the cost of unwinding is exposure. These aren't the same conversation and should not be merged.
Roles in flight, with an offer drafted but not signed, are easier. The candidate has not resigned, the goodwill cost is smaller, and the function can pause for two weeks while the freeze is clarified. The honest move is to delay, not to withdraw, and to communicate the delay with a date by which the function will decide.
Roles that are approved but not yet recruited are the easiest. Pause them, document the pause, and revisit at the end of the freeze. The cost of pausing is mostly internal, and the function can revisit with a clear process.
Manager commitments that were made informally are the trap. A senior leader who told a peer their role was approved has created an expectation that's now broken. The work is to surface these commitments early, decide which ones the function will honour and which it won't, and communicate the decision in the same conversation rather than letting each one surface separately.
The freeze ends in one of two ways. Either the function resumes the plan with the drivers it had, in which case the model is intact and the conversation resumes where it stopped. Or the freeze becomes a permanent cut, in which case the model has to be rebuilt against the new envelope and the conversation with finance restarts. Either way, the work is to keep the model alive through the freeze, because the model is what you walk back into the room with.
| Commitment | Honourable to unwind | Harder to unwind | What it costs |
|---|---|---|---|
| Signed offer, accepted, start date set | No, unless role is genuinely discretionary | Candidate has resigned, goodwill cost is real | Financial and relational cost |
| Offer drafted, not yet signed | Yes, with a delay and a clear decision date | Smaller goodwill cost | Internal only |
| Role approved, recruitment not yet started | Yes, with a documented pause | No external cost | Internal only |
| Manager commitment made informally | Yes, but surface it now | Relationship cost with the requesting manager | Internal political capital |
| Role tied to a regulatory or contractual obligation | No | Exposure to the regulator or counterparty | Material exposure, take local advice |
What to Put in Writing
A good decision without a paper trail is a decision that has to be re-made by whoever inherits the function. The artefacts below are the minimum set, and each one has an owner because the owner is the person who has to be able to produce it under questioning.
| Artefact | Who owns it | When it is written | What it prevents |
|---|---|---|---|
| The plan itself, with assumptions named | The HR business partner for the function | Before the finance review | A cut without a reason, a debate without a frame |
| The justification for each role, in one sentence | The requesting manager, signed off by HR | At submission | A defence that has to be reconstructed under pressure |
| The drivers and their owners | The HR lead for the function, agreed with finance | At submission | A disagreement that has to be re-litigated every cycle |
| The sensitivity table, showing what happens to the number if a driver moves | The HR lead | At submission | Finance asking the question you should have asked first |
| The cut history, showing what finance cut last year and why | The HR lead | After each cycle | A repeat of the same cut without challenge |
| The decisions taken during a freeze, with the reasoning | The HR lead | During the freeze | A decision that has to be re-explained after the freeze ends |
| The implicit rules, where the plan is built on judgement rather than a model | The HR lead, agreed with finance | At submission | A plan that depends on the continuity of two people |
Write these down. The writing is cheap and the absence of writing is expensive. The cost of a paragraph is small; the cost of not having the paragraph when finance asks the question is the cost of the cut you take on top of the cut you would have taken anyway.
Questions to Ask Before You Commit
The model itself. What assumptions does the plan depend on, and who owns each one? A bad answer is silence or a list of assumptions without owners. A good answer is a table with each assumption, its owner, and the date it was last reviewed.
The drivers. Which three drivers explain most of the variation in headcount across the last three cycles? A bad answer is a vague reference to growth. A good answer is a specific driver, a specific range, and a specific reason it moved.
The inflation. Where in the plan is the embedded hedge, and how did you account for it? A bad answer is that there's none. A good answer is a description of the hedge, where it sits, and what was done to it before the plan went to finance.
The cut. What did finance cut last year, and what was the reason given? A bad answer is that the cut was unexplained. A good answer is a specific reason, written down, and a note on whether the reason is still true.
The reader. Who in finance is the actual reader of the plan, and what do they need from it? A bad answer is the CFO by default. A good answer is a name, a role, and a description of what they have asked for in previous cycles.
The freeze. What is the procedure when a freeze lands mid-cycle, and who owns it? A bad answer is that the procedure is to ask the HR lead. A good answer is a written procedure with named owners for each type of commitment.
The rebuild. What is the smallest change to the plan that would change the next conversation with finance, and what would it cost? A bad answer is that the rebuild would take a year. A good answer is a one-page model that could be in the next submission.
The continuity. Who else in the function can walk the plan through the finance review, and what would they need? A bad answer is nobody. A good answer is a name and a one-page summary that would let them do it.
The legal shape. Where does this decision touch employment law, and where is local advice required? A bad answer is silence. A good answer is a list of the exposures, with a note that the specifics depend on local rules and need confirmation.
The Cost of Getting This Wrong
The first cost is the cut you take today. A plan that's a list of wishes gets cut by an amount finance chooses, and the function has no argument. The cut lands on the roles that are easiest to remove, which are the roles that are newest, or the roles whose value is hardest to see, or the roles that belong to managers with the least political weight in the room. The cost of the cut is the cost of those roles, plus the cost of the work those roles existed to do, which is a cost that doesn't appear on the invoice but shows up in the next cycle.
The second cost is the cycle that follows. A function that takes a cut without a model has to rebuild the next plan inside a smaller envelope, with the same managers submitting the same requests, and the same finance review landing on the same argument. The cycle gets harder, not easier, because the manager base has learned that asking for more works, and the finance team has learned that cutting without explanation works, and the pattern is now embedded in both sides. So the cost of getting it wrong once is the cost of getting it wrong twice, and the cost of getting it wrong twice is the cost of a function that has lost the argument about its own shape.
The third cost is the function itself. A people function that can't defend its own plan is a function that has lost the right to be in the room when the workforce is being shaped. Other functions will step in. The work will be done by someone who doesn't know the work. And the next time the organisation needs to reshape, the conversation will happen without the people who knew the work best. So what does it cost to keep arguing from a list, and is the cost of building the model really higher than the cost of the function that loses the argument?
When You Are Ready to Go Further
If the case above describes your function, the next step is to look at how others have rebuilt the model without rebuilding the function. HROpsLab's independent comparison work covers the planning tools, finance systems and analytical approaches other organisations have used, with the strengths and the trade-offs written down by people who have watched the implementations, not by the people who sold them. The comparison is independent, which means we don't sell any of the tools and we don't get paid by any of the vendors. The work exists to help HR leaders walk into the finance review with a model they didn't have to build alone.
If you want a second opinion on the shape of your own rebuild, our research desk will read your current plan and tell you which assumption is the one finance will read first. The reading is free, and it's the same reading we do for the comparison work, which means you get the same answer we would write for a publication. The desk doesn't sell software, doesn't sell advice, and doesn't take referral fees. It's a desk because that's what it's.
The last step is yours. The model you build this cycle is the model you walk into the room with next cycle, and the room is the same room, and the CFO is the same CFO, and the only thing that changes is what you bring.
Frequently Asked Questions
What does headcount planning actually involve?
It involves turning the work the organisation has to do in the next cycle into a number of roles, with a cost, with assumptions that can be named and defended, and with a paper trail that explains every line. The work isn't the spreadsheet. The work is the conversation between the manager who wants the role and the function that has to defend it, and the conversation between the function and finance about the assumptions behind the role. The spreadsheet is the artefact. The work is the conversation.
Who should own the headcount plan?
The HR business partner for the function, signed off by the senior leader of the function and agreed with finance. The owner is the person who has to walk the plan into the finance review and defend it under questioning. If that person isn't the HR business partner, the plan is owned by someone who isn't in the room, and the plan will lose. The owner is also the person who has to keep the plan alive between cycles, which means revisiting the drivers and updating the assumptions.
How far ahead should a headcount plan look?
The plan is built for the next financial year and revisited every quarter. A plan that looks further ahead than a year is a plan that depends on assumptions no one can defend, and a plan that looks less ahead than a year is a plan that doesn't allow managers to recruit. The work is to build a year-long plan and a rolling re-forecast that updates the plan every quarter, with the drivers as the link between the year-long plan and the re-forecast. The shape of the year depends on the business, and the specifics need to be confirmed locally.
How should backfills versus new roles be handled?
Backfills are easier to defend because the work the role does is already known and the cost is already in the budget. New roles are harder because the work the role will do has to be justified, and the cost is incremental. The honest move is to separate the two, defend the backfills on continuity and the new roles on the work they will do, and ask finance to consider them separately. If they're merged, the backfills subsidise the new roles and the new roles undermine the backfills.
What do you do when managers inflate their requests?
Mark the inflation, adjust the request to the level you can defend, and tell the manager what you did and why. The conversation is harder the first time and easier the second time, because the manager learns that the inflated request gets adjusted and the honest request gets approved. The work is to be consistent. If some managers get their inflation and others don't, the function loses the argument about its own judgement.
How do you plan when revenue is uncertain?
Build the plan from the work, not from the revenue, and revisit the plan every quarter. The drivers are the volume of work the organisation has to do, not the revenue the organisation expects to earn. If the revenue moves, the work moves, and the plan moves with the work. The specifics of the planning rhythm depend on the business and the finance team, and the shape of the exposure needs local confirmation.
How often should the plan be re-forecast?
Every quarter, with the drivers as the link between the year-long plan and the re-forecast. A re-forecast that doesn't have drivers is a re-forecast that has to be re-argued every time. The work is to keep the drivers alive, with named owners, and to make sure the re-forecast changes the plan when the drivers change and not when the manager changes.
How do you defend a plan that gets cut?
Bring the model to the next cycle and walk finance through the assumptions. The cut is a signal that the plan was not a model, and the rebuild is the defence. The work is to show finance which assumption they cut, and to ask which assumption they would have used instead. If finance can name one, you've a conversation. If finance can't, the cut was free, and the rebuild is the way to make the next cut cost something.
HROpsLab is the independent review publication for HR operations, payroll and workforce planning. We don't sell software, payroll services or advice. We write what we find.
