Making It Say Something

The brief arrived as one sentence from the director: a poster to raise awareness of blood donation. There was a budget for 2,500 A3 sheets, a week of visits to four colleges, and a target of 900 units.

The designer asked the seven questions and got through five of them. Question two stopped everything. What do they currently believe about donating here? The director said people did not understand that donation was safe. The communications officer said people were busy. Neither of them had asked anybody.

The second source cost one conversation

So she went to the counter and asked the two attendants what people say just before they change their mind and leave. They could recite it without pausing, because they hear it every day: how long is this going to take.

The register supported them. The bank’s own arrival and discharge stamps put the median donor in the building for one hour and fifty minutes. Almost all of that was waiting for the medical officer, who screened between eleven and one because that was when he was free between theatre lists.

The before state was not ignorance about safety. It was a specific memory held by people who had already donated once: I lost an afternoon. Awareness posters are written for people who have never heard of you. These people had heard of the bank and had decided.

The decision, and what it cost on each side

The honest line the brief now pointed at was a promise about time. Booked slot, forty minutes, out by ten past. That line could not be printed unless the bank changed how it worked on the two drive days: the medical officer screening from nine, a second counter staffed, and donors given a slot rather than a queue number.

That was two extra staff shifts and an argument with the pathology head, who said moving the screening earlier would collapse the sample-processing queue behind it and that the medical officer had a theatre list nobody could move.

The alternative was cheap and safe. Print a poster that said donate blood, save a life, over a photograph of a smiling donor. Nobody in the building would object to it. The designer could also predict its result, because it was approximately last year’s poster and last year’s result was the reason she had been called.

And there was a third option that was worse than both, which was the tempting one: print the forty-minute promise without making the change. That version would work once. The fastest way to destroy a claim is to get a lot of people to check it, and 900 people checking a false forty minutes would cost the bank the donors it still had.

Agreeing the measure before printing

Before anything went to press they wrote one line into the brief. Success was repeat donors, counted as arrivals whose donor identification number already existed in the register, over four weeks. The diagnostic measure was the gap between the arrival stamp and the discharge stamp, which the bank was already recording and had never looked at. A separate helpline number was printed on the poster and nowhere else, so a call could be attributed.

The designer also wrote the sentence she expected to need later: this fortnight also contains a college festival and a newspaper appeal, so we will not be able to separate the poster from either, and we are accepting that as rough. A limitation named before the fact is a professional position; the identical sentence afterwards is an excuse, and the difference is entirely in the date at the top of the document.

The forecast went into the same line, deliberately rough. The bank had been seeing about 40 repeat donors a fortnight. If the change worked, expect something in the region of 60 to 80, not 300. Writing that down cost a minute and put everybody’s expectations in one room before any money moved, which is also where the conversation about arithmetic happened: 2,500 posters reaching roughly 9,000 students at four colleges cannot produce 900 units on any conversion rate anybody has ever measured, so the drive needed the staff canteen and the two mills as well.

The flyer had been assembled the way flyers usually are, which is one reasonable request at a time. The hospital logo. The trust’s logo. The funding partner’s logo, because the grant agreement required acknowledgement. A photograph of the senior surgeon. A quotation from him about the right to sight. The dates. The eleven village names. A QR code. Two phone numbers, because one was the ambulance line. The word free, in red, because somebody in the meeting said red meant important.

Every item had a person behind it who could explain why it mattered. Nobody had looked at the sheet as a whole in a month.

Forty seconds of blurring

The communications assistant flattened a copy in a free browser editor and applied a twenty-pixel blur. Nothing survived. The sheet came back as an even field with three darker rectangles in it, and the darkest was the photograph of the surgeon’s face, which at that blur was the only identifiable object on the page.

That is the diagnosis stated precisely. The eye needs one entry point and finds it by looking for the strongest local contrast. There were four candidates at the same weight, so scan order was arbitrary, and an arbitrary scan is uncomfortable enough that people resolve it by leaving.

The decision, and the cost on each side

The fix was not deletion but ranking. In the proposed version, the two things that survived a blur were the dates and the words free eye check, taking the top third. The eleven village names went to level three at nine point in a single column. The three logos went to a row along the bottom at the size the grant agreement actually specified, which nobody had read and which was smaller than what had been used. And the surgeon’s photograph and quotation came off the flyer entirely.

The surgeon was also the hospital’s largest individual donor and the reason the camp had a theatre list at all.

So there were two real costs. Demoting him risked a relationship worth considerably more than one camp, and the administrator would be the person in that conversation. Keeping him risked eleven villages of people who could not find the date, which was the only fact the flyer existed to transmit.

The administrator did two things instead of choosing in the abstract. He had twenty copies of each version printed, and he took them to the outpatient corridor and ran a three-second test on staff and waiting relatives: look at this, then tell me what it was about and what it wanted you to do.

He also prepared a third option, which is the move that usually saves a cut element. The surgeon’s photograph and quotation were not thrown away; they were moved to a separate piece, an A5 card handed to every person screened, where the quotation is read by somebody who has already decided to come and has time.

Why the sheet could not simply be made louder

The review meeting had produced the usual instruction, which was to make the dates bigger. That was not going to work on its own, and the reason is the arithmetic of contrast. Emphasis is comparative: there is a fixed amount of perceivable difference on one surface, and every element that spends some of it leaves less for the rest. The dates had already been enlarged twice. What they needed was not more size but fewer competitors, which meant the other items coming down.

There was a second cost nobody had costed. The sheet had grown to eleven separate objects. Three elements carry three relationships to manage; eleven carry fifty-five, and each of them has to be positioned and weighted against all the others. Nobody had made a bad decision. The decisions had simply been made one at a time, by different people, over a month.

Grouping bought some of it back. The eleven village names, set in one column at one size with one alignment, are perceived as a single object rather than eleven, so long as the space around the block is clearly larger than the space inside it.

The founder’s brief was a list. The workshop began in 1978. Three generations have worked in it. It won a state award in 2009. A finishing machine was bought in 2011. It makes chairs, stools, screens, trays and lamp bases.

Written out as beats, every join between those sentences is and then. That is the diagnostic: each beat can be deleted without breaking the next one, and a viewer who senses that leaving costs them nothing will leave.

Two beat lists, seven sentences each

The film-maker wrote both versions as plain sentences before touching a camera, because at that stage a change costs thirty seconds.

The company version, honestly written out, had no change in it anywhere. At the end of ninety seconds the workshop existed, the family cared, the furniture was good, exactly as at the start. That is a description. Descriptions are legitimate objects and they have to earn attention through usefulness or beauty, never through suspense, and a description of a workshop set to music earns neither.

The second version came from a question asked in the interview: what nearly went wrong that you still think about. An architect in Guwahati had ordered sixty chairs for a cafe, on a five-week deadline, and had refused the first delivery because the joint finishing varied visibly across the run. The workshop remade twenty-two chairs at its own cost and changed how it batches finishing.

That is a state, an event, and a different state, and it joins with but and therefore at every seam.

The decision, and the cost on both sides

The founder’s son objected to putting the rejected batch in the film, and his objection was not vanity. The exhibition audience included a state procurement officer and at least one competing unit. A recorded admission that a delivery had been refused could be read at a later tender as a quality record rather than as candour.

The counter-argument is the one about statements against interest. A claim made by an interested party carries a discount, and the discount disappears when the statement damages the speaker, because the usual explanation for why they said it is unavailable. Twenty-two remade chairs is what makes every other claim in the film believable.

Both of those are true at the same time, and the choice had a price either way: silence buys a film nobody remembers, and candour buys a sentence that may be quoted back in a procurement room.

There was a second decision underneath it. The architect was a real person whose business continues after the film. She had to be asked what the piece was for, where it would run and for how long, and told she could ask for it to come down.

Who the film was about

The test that settled the structure took one minute. Delete every instance of the workshop’s name from each script and read what is left.

From the company version, nothing was left. Founded, grew, invested, won, makes: remove the name and there is no sentence with a person in it. From the second version, a story remained in full, because somebody wanted sixty chairs in five weeks, was let down, and ended up with a supplier she now uses for everything.

That is not an argument for hiding the workshop. It is a diagnostic for whether anything is being carried other than the workshop. The brand still appears, at the turning point, where it is relevant, and it is shown doing something specific rather than being described: the second batch going onto the bench, the new finishing jig, the number chalked on the wall.

Where the ninety seconds went

The first assembly spent forty-four seconds arriving: the road, the gate, the sign, the founder sitting down. The turn arrived at fifty and had twenty seconds left.

Rewritten to the beat list, the promise took three seconds, the situation and the problem seventeen, the turn forty, the new state twenty and the ask ten. The maker counted the narration at 231 words before recording anything, which at about 150 words a minute is where ninety seconds actually lands, and that count is why nothing had to be cut in a hurry at the end.

She agreed on one condition, stated plainly and more than once: she must not be identifiable. The organisation did what most people do. It changed her name, shot her in silhouette, and shifted the pitch of her voice.

Then somebody went through the rough cut detail by detail and counted, for each one, roughly how many people it still fits.

• The town: tens of thousands.
• A woman running a tailoring unit with two machines: a few hundred.
• Whose husband works at the sugar mill, which was named in a wide shot: perhaps a dozen.
• Whose daughter is in class nine: one, to anybody in two wards.

Not one of those details is her name. Together they publish an identity, and this is how anonymity actually fails. The name was never the identifier.

Three of the protections were weaker than they looked

The pitch shift is a reversible operation and undoing it is a menu item in free software. The silhouette held for most of the interview and slipped for two frames when she turned her head, and two clean frames in a file anyone can step through is a face. And the photographer’s stills, shot on a phone and sent for the accompanying article, still carried their EXIF block, which contains the coordinates of the room.

The robust protections are all subtractive. Crop the face out of the frame rather than covering it. Replace the audio rather than processing it. Remove the detail from the shot. What is not in the file cannot be recovered from the file.

The decision, and what it cost either way

Here is where it stopped being technical. The whole point of the film was a mechanism: the interest on her loan was being repaid in stitched garments rather than in cash, at a rate she had never been shown in figures. Nothing else in four hours of material carried that. It was the finding.

And the tailoring detail is what made her identifiable. Take it out and the film becomes a general piece about debt that a viewer has seen before, of no use to the people it was made for. Leave it in and the protection is decorative.

A third option was considered and rejected: rebuild the sequence as illustration. That removes the identification entirely and also removes the evidential claim, which in a film about a practice people deny exists is most of the value.

The organisation had also, at the planning stage, agreed a withdrawal policy and told her about it, which is the cheap decision that turned out to matter most later.

The release that would not have covered it

The form she had signed was the organisation’s standard one, and it named a documentary film. It did not name the accompanying article, the social cut, the conference screening, or the fact that her words would be edited into four minutes out of four hours.

Nobody in that situation denies signing. The complaints that arrive months later come from people who agreed to one thing and found another, so the part of a release that fails is the scope rather than the signature. The form was rewritten in ordinary sentences before publication and she was asked to agree to it again: what was recorded, what it is for, where it may appear, for how long, that it will be cut, and that she may ask for it to stop.

The audit that found the hole in the structure

The beats were listed down a page with the specific file that carries each one beside them, with a timecode, or the word MISSING. Two lines came back MISSING.

One was an asset gap: there was no shot of a ledger, and one could be filmed. The other was a structure gap. Beat six claimed a moment when she realised the arithmetic, and there was no such moment, because she had known for two years and had no alternative. No amount of filming repairs that. The beat was wishful, and it was deleted rather than propped up, which is the distinction the audit exists to make visible.

Last year’s subject line was: your child may be missing out. It opened at 41 per cent against a list average of 22, and the counsellor who wrote it wanted the same shape again.

The objection to it was not taste. It was the next email. That one opened at 9 per cent, against the list’s normal 19, and 84 people unsubscribed within two days. The cost of the first line did not appear in the first line’s number. It appeared a fortnight later, in a place nobody was looking, and the centre had recorded both figures without connecting them.

What the fold was doing to the honest version

The replacement subject was written to be complete rather than teasing: scholarship test, 14 December: who can sit it and what it covers.

Opened on a phone, most clients cut the subject somewhere around forty characters. That left scholarship test, 14 December: who can, which still carries the two facts that matter, because the specific words were at the front. The version the counsellor drafted first, how your child can compete for one of sixty scholarship places, cut to how your child can compete for one of, which carries nothing.

The preview line after the subject was being spent on view this email in your browser. That is a second headline of roughly the same length, available at no cost, and it was set to carry the fee waiver and the deadline.

The body opened with two sentences of throat-clearing about the centre’s commitment to accessible education. They were deleted and nothing was missing, which is the usual result.

The decision, and its cost on both sides

The honest subject was going to open lower this week. That was predictable and it was stated in writing before sending, along with the two numbers that would be used to judge it: opens on this email, and opens on the next two.

The cost of saying that out loud is that the admissions head is judged on this week, not on the following month, and an open rate that drops from 41 to something lower is a visible fact with the counsellor’s name on it. The cost of not saying it is that the list is spent a little further, and a list is the one route to these parents that nobody else can switch off.

There was a second, quieter piece of work on the body. Most of these parents read English as a second or third language. The draft used test, exam, paper and assessment for the same event, which reads as four things to somebody parsing carefully; it used phrasal verbs in the load-bearing sentences, including put off and sit for; and it wrote numbers as words. The whole draft was then cut against a stated limit of 260 words, from 410.

Cutting against a number rather than against a feeling

The limit did most of the work. Asking whether a paragraph is good has no stopping point; asking what earns its place in 260 words does. The first paragraph went and nothing was missing. There is a and it is openings were rewritten. Three nominalisations became verbs. Two sentences made the same point in different clothes, and the more concrete one was kept.

One cut was put back deliberately. The draft had explained twice, in different words, that the test is free to sit. A second statement of an idea already stated is usually the writer hearing themselves, but this one was doing a different job: a parent reading on a phone between shifts who missed the first mention catches the second, and the fee is the single fact most likely to stop somebody registering. Compression removes redundancy, and some redundancy is what lets a distracted reader catch up.

The test that found the rest

The counsellor read the final version aloud, standing up. She ran out of breath in the second paragraph, which is where the sentence was too long, and stumbled on a phrase that turned out to be an accidental rhyme. Both were fixed in four minutes. Nobody had needed to have an opinion about either.

The report was clean and well set. Out-of-range values were printed in red; in-range values in green. That was the whole signalling system, and it carried information in hue alone.

It came to light by accident. A phlebotomist’s father, a retired schoolteacher with a red-green deficiency, asked his son which of the numbers were the bad ones, because he could not tell. Nobody had ever complained, and the reason is worth stating: people who cannot read a document do not send feedback about it. They ask somebody, or they guess, or they take it to a doctor and say nothing.

The size of it, counted rather than estimated

The lab prints roughly 4,000 reports a month and about 55 per cent go to men. Taking the commonly cited figure of around 8 per cent of men with some form of red-green deficiency, that is on the order of 170 men a month receiving a document whose only signal they cannot read.

That is the smaller number. The larger one is everybody else, because the report is not only read in colour.

• Insurance claims are photocopied at the branch, in one colour. Every flag vanishes.
• Reports are photographed and sent on messaging apps, often at low screen brightness in daylight.
• The lab’s own PDF is opened by doctors on phones with a warm night filter running.

A screenshot of the PDF converted to greyscale lost 100 per cent of the flags. That single test, which took under a minute, was what moved the conversation from opinion to fact.

The contrast measurement was the second surprise. The red used, at the body size on the report, measured 3.9 to 1 against white, below the widely used 4.5 floor for body text. The green measured 1.9 to 1, which is near the bottom of the scale. Both had been chosen by eye on a good monitor.

The decision, and what it cost on each side

Changing the report was not a five-minute edit. The template sits inside the lab’s accredited document control, so a revision means a change request, re-approval, and a new version number across three branches. There was pre-printed stationery to replace. Fourteen staff read those colours daily and would have to be retrained. And the senior pathologist liked the scheme, on the reasonable ground that referring doctors recognised it at a glance.

Not changing it cost nothing visible and nothing measurable. That was the honest description of the alternative, and it was the strongest argument for acting, because a failure that generates no complaints generates no pressure either.

The proposal deliberately did not remove the colour. Colour was demoted from the only channel to the second one.

The typography was part of the same problem

The review picked up two more things while the template was open, both of which belong to readability rather than to decoration.

The patient’s name and the reference ranges were set in a light weight, chosen because it looked modern. Type carries its information in thin strokes, and thin strokes lose contrast first, so the light weight was a contrast decision disguised as a style decision, and it failed on a photocopy before anything else did.

The comments field at the foot of the report ran the full width of the page, about 110 characters a line. The return sweep at the end of each line is approximate, and the longer the line the more often the eye lands on the wrong one. Set to two columns of about 60 characters, the same text is read without anybody losing their place, and the page got shorter rather than longer.

None of this needed paid software. The template was rebuilt in a free layout tool, the contrast was measured in a free browser checker, and the greyscale test was a setting on a phone.

The footage was good. Daylight from a workshop door, a steady phone on a clamp, a real bench. Exported at 1080 by 1920, correctly. And the workshop’s best-performing reel produced two enquiries in a month.

Four separate failures were stacked on one file, and none of them was visible in the editor.

What the platform was drawing over the work

The compositions placed the hands and the joint in the lower third, which is where a person naturally puts the subject of a bench shot. On the platform, that strip carries the caption sheet, the account name and the audio label, and the right edge carries a column of buttons. The exported frame was correct and the visible frame was smaller, and nothing in the editing preview had ever shown the difference.

The cover was the second failure. The reel faded in from black, and the platform’s default is frame zero, so the still recruiting every play was almost black.

The third: the explanation of the joint existed only in the spoken audio, and most plays were muted, because autoplay is muted by policy rather than by preference. Silent, the opening was a stranger’s hands moving.

The fourth was inaudible by construction. The music bed had been chosen for its weight, which sat between 60 and 80 hertz. A phone speaker a few millimetres across returns almost nothing below about 300 hertz, so on the device nearly everyone used, the bed was a thin scratch and the drama built from it was not there.

What the numbers said, read as stages rather than as a score

Click-through was high. The thumbnail read HIDDEN JOINT over a tight crop, and people tapped it. Retention fell from 100 to 38 per cent within three seconds and then ran almost flat.

That pair has one reading. The cover attracted people and the opening did not pay it. A high click-through beside an immediate exit is the pattern every ranking system on the market is built to detect, so the two numbers were working against each other.

The decision, and what each option cost

Moving the work into the middle third was the change that mattered and there was no cheap version of it.

Re-shooting the build was not available; the piece was finished and sold. Cropping into the existing 1080-wide footage would reframe it at the price of resolution, and the grain of the wood was the thing the reel was about, so a softened crop that then went through the platform’s re-encode would lose exactly the detail the viewer came for.

The third option was to change how the next build was shot: record wide at the phone’s highest resolution, treat that as a master, and export crops from it. That cost storage the workshop did not have much of, and a noticeably slower edit on a five-year-old laptop.

Underneath all of it sat a structural problem nobody had raised. Nineteen thousand followers existed on a service that could change its interface, its ranking or its mind overnight, and the workshop had no route to those people that it owned.

The line they drew through the fixes

Every change on the list was a change to form: the aspect ratio, the safe area, the cover frame, the size of the on-screen text, the filter on the music, the specific noun at the front of the title. None of them altered what the piece claimed.

One proposed change was not form. Somebody suggested a thumbnail showing the finished cabinet from a build they had not filmed, on the grounds that it was the best photograph they owned. That is the promise rather than the delivery, and it is where a click is bought by spending the one asset that cannot be bought back. It was the only item on the list that was dropped.

The folklore was dropped as well. The workshop had been posting at a particular hour because a video had told them the algorithm preferred it. Ranking functions are unpublished, personalised and frequently changed, and what circulates as knowledge about them is assembled from the accounts of people whose piece happened to do well. The thousands whose identical experiment failed do not get a platform to explain it from, so the rule cannot be tested by anyone repeating it.

The request arrived on a Thursday afternoon, which is when these requests arrive. The centre’s marketing lead wanted a countdown timer reading seats closing, and a line saying only 3 seats left.

The designer had the brief document in front of her. The real cut-off was the board’s admission date of 30 June, six weeks away, and the timer was to be set to seventy-two hours and reset on reload. The figure of three had not been counted from anything; it had been chosen because it looked urgent.

The test that decides it

Everything on that page was persuasion, including the things she had already done: ranking the fee and the date above everything else, putting the specific noun first in the headline, and choosing a photograph of the actual classroom instead of a stock one. Persuasion is what a client hires a designer for.

The usable line is whether the technique survives being explained. We made the date the biggest thing so you would remember it still works when you say it out loud. We put a fake countdown there so you would panic stops working and makes the person angry, and the difference is not how exciting each one is. It is whether the reader is being given a reason they can evaluate or a process they cannot inspect.

A fake timer also fails a second, plainer test: the claim cannot be checked and is not true.

The decision, and the cost on both sides

Refusing was not free. The centre was about a fifth of her year’s income, and the page would go live with a timer on it whether she built it or not, because somebody else would.

Agreeing was not free either, and the argument that actually moved the room was about the centre rather than about her. Its whole position was we tell you your child’s real rank. The parents most likely to reload a page and notice a resetting timer are the parents who compare four centres before choosing, which is the audience the positioning was built for.

She had one piece of evidence and used it. The previous season’s page, which had carried a similar scarcity line, converted at 3.1 per cent, and a Google review naming the countdown was still the second review on the listing.

Then she offered the honest version of the same mechanism, which is the move that usually works better than a lecture. The 30 June deadline is real. The batch size of 60 is real. The current enrolment of 41 is countable and sits in a register that already exists. Printing the real count uses the same loss aversion and survives being checked.

That had a cost of its own, and it belonged to the director, not to her: a live count is a number that will look bad in a weak year, in public, and once it is up, taking it down is itself a signal.

How the argument was made

She used four sentences, in order, and then stopped talking.

The goal, restated from the brief: the one thing this page has to do is get comparing parents to book a visit. The mechanism: the position we agreed is that we tell people the real number, and this page is where that is either true or not. The evidence: last season’s page and the review that is still second on the listing. The cost of the alternative, said plainly and without drama: we can put the timer up, and the likely effect is that the parents who compare four centres, who are the ones we are for, find out that we do not.

Continuing to argue after the argument has been made reads as anxiety and invites a compromise nobody wants, so she stopped.

She also conceded everything else in the same meeting and conceded it quickly. The marketing lead wanted a different blue, the corner radius changed and the testimonials moved above the fee. None of those touched what the page claimed, and holding any of them would have spent the standing she needed for the one thing that mattered. A designer who defends every decision is not principled, and clients read it correctly.