Working for Yourself

Ritika had been employed for five years designing report covers, price lists and pharmacy shelf cards for a chain of diagnostic labs across Nashik and Dhule. She wanted to leave and do packaging for small food producers, of which the district has several hundred.

She did the arithmetic first, which is unusual. Take-home needed: ₹38,000 a month. Annual costs, written down honestly: ₹16,000 of software she intended to replace with Inkscape and Scribus in year two, ₹24,000 of laptop amortised over three years, ₹22,000 of internet, power and phone, ₹10,000 for an accountant. Ten earning months at eighty billable hours gave her a floor of about ₹790 an hour, or roughly ₹52,000 for the kind of pouch-and-carton job she meant to sell. Her spendable cash, after subtracting the tax she owed and one month she refused to touch, was ₹2,40,000. Against a bad-month survival cost of ₹31,000 that was seven and a half months of runway, which her ten-week cash cycle reduced to about five months of decision-making room.

Then she started the test: three paid jobs, three strangers, at or near ₹52,000, inside three months, without resigning. By the end of month two she had one. A turmeric and spice brand in Sinnar had paid ₹48,000 for a four-SKU pouch family, found through a printer she had used at work. Two other conversations had gone quiet.

In week nine the offer arrived. A packaging vendor who supplied her employer wanted her on a nine-month retainer: ₹35,000 a month for what he described as “roughly two days a week” of artwork adaptation: resizing existing designs for new SKUs, correcting label panels, preparing print files. Regular money, from a person she already knew, starting the day she resigned.

She worked out what it actually was. Two days a week is about sixty-four billable hours a month, which put the rate at ₹547 an hour against a floor of ₹790. Worse, the scope was availability rather than output: “roughly two days” with no ceiling written anywhere. And it was adaptation work on other people’s designs, which would go out under the vendor’s name and produce nothing she could show.

The cost of refusing was not abstract. ₹35,000 a month covered her survival cost with something left over. Taking it meant she could resign in a fortnight instead of waiting, and it meant her runway stopped falling.

The cost of accepting was that the test would end. Two days a week of contracted availability, plus the third day the scope would quietly expand to, would leave her the remainder of the week for the outreach that was already the slowest part. The one thing she had learned in two months was that the route to a client took her about eleven hours per genuine conversation. There is no version of the next quarter in which she finds two more food brands at ₹52,000 while adapting somebody else’s labels three days a week.

She also noticed the concentration. At ₹35,000 a month against her target income, the vendor would be roughly seventy per cent of everything, in month one, before she had a single direct client of her own.

There was one more thing she checked before replying, and it is the thing most people in an overlap skip. Her employment contract had a clause headed Intellectual Property that assigned to the company anything created “in connection with the business of the employer”, and a second under Outside Activities requiring written notice of paid work elsewhere. Diagnostics and packaged turmeric are not the same business, so the first clause was survivable, but she gave the written notice anyway, in four lines, and kept the reply. The vendor was also a supplier to her employer, which is the one category of client an overlap should not touch, and that was a second and quieter reason the retainer was a bad idea.

She wrote down, for each of the three test jobs, five lines: how they found her, what they asked for, what she quoted, what they paid, and how many hours it actually took. The hours column was the uncomfortable one. The Sinnar job had taken 71 hours against the 45 she had imagined, which put her real effective rate on it at ₹676 rather than the ₹1,067 she had assumed, below her floor, on a job she had considered a success.

Anand edited video for seven years, the first three at a wedding studio and the last four alone. His site carried forty-one pieces. Twenty-six were wedding films and highlight reels. Eleven were corporate: factory walkthroughs, a hotel group’s room tours, two training videos for a cashew processing company. Four were what he actually wanted more of: long-form interview edits for a Malayalam podcast and two documentary shorts for an NGO working on coastal erosion.

His enquiries were almost entirely weddings. In the previous twelve months, sixteen of nineteen enquiries had been weddings or engagement shoots, and they were about sixty per cent of his income. His effective hourly rate on them, once he counted the sixty hours of logging and the three rounds of family feedback that arrived through a WhatsApp group, was around ₹330. On the podcast work it was about ₹1,100.

He understood the mechanism, which is that a portfolio is a menu and people order from menus. He had been telling clients for a year that he was “moving toward documentary work” and nothing had changed, because the declaration was in his conversations and the evidence was on his page.

So the decision in front of him was whether to remove the twenty-six wedding films.

The argument against was straightforward and not stupid. Weddings were ₹1,08,000 a quarter, arriving reliably, in a season he could predict. He had a two-year-old daughter and an EMI. Nobody in Kochi was commissioning documentary shorts at volume, and the coastal erosion film had been paid for by a grant that had ended. Removing the wedding work meant removing the only thing on his site that anyone had ever paid him for at that frequency, in a market where the alternative demand was thin and unproven.

The argument for was equally concrete. He had run the numbers at the end of the financial year and found his utilisation was seventy-one per cent (very high) while his income was flat for the second year. He was busy and not moving. The wedding work consumed the weeks in which he would otherwise have approached the four Malayalam-language production houses and eleven NGOs he had listed and never written to. And every wedding film on the page produced another wedding enquiry, which produced another wedding film, which he was then too busy to stop taking.

There was a third option he considered and rejected: keep both, and put the documentary work first. He had tried a version of this already by reordering the grid. The enquiries had not changed, because the wedding work was still visible, still numerous, and still the thing a reader scrolled past most of.

Cutting the page was only half of it, and the smaller half. What the seven remaining pieces needed was four sentences each: what the client actually needed and why, the constraint that shaped it, one decision he could defend, and what happened afterwards. He had none of that written down for any of them, because he had never had a reason to write it while the files were open, and reconstructing it two years later took him a full weekend of scrubbing through timelines.

The erosion film was the one that taught him what the words are for. The images are a fishing village and a receding shoreline; they do not say that the NGO needed it to work silently on a phone in a Delhi meeting room, that the entire narration therefore had to survive as burnt-in Malayalam and English subtitles at a size legible on a five-inch screen, or that he cut the interview footage to eleven minutes when the grant expected thirty because eleven was what a funder would actually watch. A viewer cannot infer any of that, and it is the only part of the project that demonstrates judgement rather than software.

He had no outcome numbers for any of it and refused to invent one. What he had instead were three honest substitutes: a quotable sentence from the NGO’s programme head, asked for in the delivery email while she was pleased; the fact that the podcast had come back for a second season, which is a real outcome fully within his knowledge; and a delivery fact (three films, eleven days, first cut approved without changes) which is small, true and exactly what an anxious buyer wants to know.

Devika had been posting illustration on Instagram for eight months. She had 4,300 followers, mostly other illustrators and art students, and a reel about her brush textures that had reached 90,000 people. In that time she had earned ₹19,000: one book cover for a small Hindi publisher at ₹12,000 and two portrait commissions from followers at ₹3,500 each.

Her landlord had given notice of a rent increase effective in eight weeks. Her spendable cash was ₹34,000 against a survival cost of ₹19,000 a month. She needed a paying job inside six weeks, and the honest position was that she had no pipeline at all: no proposals out, no conversations live, nothing agreed.

Her instinct was to post harder. The reel had worked; a second one might work better; the account was clearly growing. Every piece of advice she had absorbed in eight months said that consistency compounds and that she was close.

The problem with that instinct is a lag she could have measured. Inbound arrives on a clock of a year or more. Her eight months of posting had produced two commissions from followers, which is a conversion she could extrapolate: roughly one paid job per four months of posting, at ₹3,500. Six more weeks of it, at the same rate, produces about ₹1,300 of expected income. The channel was not failing. It was behaving exactly as inbound behaves, and it was the wrong instrument for a six-week problem.

Two channels can move inside six weeks, and neither is posting. Direct outreach lands in two to eight weeks. Subcontracting lands within about a month of getting onto a roster, because the intermediary already has the client, the brief and the budget, and has only run out of hands.

So the decision was how to spend forty-five days. Posting was not merely low-return; it was expensive, because a piece she was proud of took her two days and the reels took an afternoon each. Six weeks of outreach meant six weeks of not posting, and she had watched what happened to other people’s accounts when they stopped. Reach decays; the algorithm forgets; the eight months of accumulation she had paid for in unpaid evenings would be partly wasted.

She listed what she actually had that could be sold to somebody with a budget, rather than admired by somebody without one. Editorial illustration for Hindi and English magazines. Children’s book interiors. Spot illustration for annual reports, which nobody enjoys and which every NGO and cooperative bank in the state commissions once a year.

She also had something she had never counted: twelve conversations in the comments with other illustrators who were busier than she was.

Two channels she did not use are worth recording, because she considered both and the reasoning matters.

She looked at marketplaces. The arithmetic was not the obstacle, though it was worse than it looked: a commission, then a withdrawal fee, then an exchange-rate margin of two to four per cent that is not presented as a fee at all, so a quoted job arrives about a fifth lighter. The structural problem was the one she could not fix. On a listing page the buyer sees a row of interchangeable profiles with star ratings and prices, cannot assess illustration because they are not illustrators, and therefore chooses on the two axes the interface displays. Competing on being better does not work in an interface that does not show quality. She noted one exception she might use later: a narrow, technically specified service such as preparing press-ready artwork, where the buyer knows exactly what they want and can tell whether they got it.

She also did not ask for referrals, for the honest reason that she had almost nobody to ask. Referral is unsuitable as a primary channel for anyone who has not yet done much work, which is precisely when it is most wanted, because the whole mechanism depends on a stock of satisfied clients. What she did instead was build the stock: at the end of the chapter-head job she asked the commissioning editor, at handover and not six months later, whether she knew anyone publishing illustrated non-fiction in Hindi this year: a named client type and a named trigger, which is a question somebody can answer in seconds rather than a search of a category with no edges.

Sabari quoted ₹86,000 for an export catalogue: thirty-two pages, forty-one garment styles, product photography by a studio he would brief but the client would contract directly, layout in Scribus, press-ready PDF, one printed proof check, and two rounds of revision. He had costed it from his own records (the last two catalogues of similar size had taken him 104 and 118 hours including the press visit) at a floor of ₹720 an hour plus a margin.

The client, a two-person export house shipping cotton knitwear to Germany and the Netherlands, came back within a day. They liked the work. They wanted thirty per cent off.

This is the most ordinary request in the trade and it is where rates die, because the obvious responses are both wrong. Refusing outright loses a real project from a client with real money. Agreeing loses ₹25,800 and, more expensively, tells them the first number was not the real number: an inference that is permanent and that will be applied to every future quote and to anybody they refer him to.

So he asked a question instead of making a concession. Which part of it is the problem: the total, the timing of the payments, or the scope?

The answer took two exchanges to extract and was not what he expected. The total was not the problem. The timing was. Their German buyer paid them 60 days after shipment, they had just paid a yarn supplier, and they could not release ₹43,000 as a deposit in the same fortnight as the catalogue had to start. They had reached for a discount because it was the only lever they knew existed.

That changed what was on the table. A cash-flow problem is solved by payment stages. A scope problem is solved by removing something. Only a genuine total-budget problem needs the price to move, and this was not one.

He also had a second constraint he had not raised. The forty-one styles were not all needed. Fourteen of them were variations in colourway of six base garments, and the buyer in Düsseldorf ordered by base style. Presenting them as forty-one separate entries was going to make the catalogue longer, slower to produce, and (he thought) worse to use.

Which gave him a real option to put beside the original: twenty-four pages, twenty-seven entries with colourway swatches rather than separate spreads, same photography, same two rounds. Fewer pages, less layout, a genuinely smaller job. ₹64,000. Not a discount: a different, complete, honestly delivered product.

Before any of that he had to be sure the quote itself would survive the conversation, because a number in an email is not a quote. His was two pages and carried eight things: the problem restated in their words, what was included with a count of each item, what was explicitly not included, the number of revision rounds, the schedule with dated dependencies, the price, the payment terms, and how long the quote stood.

The exclusions list was doing more work than any other paragraph. Photography was excluded because the studio would be contracted by the client directly, which also meant Sabari would not be paying a photographer out of his own account and waiting to be reimbursed. Copywriting in German was excluded. Translation was excluded. Any style added after the layout was approved was excluded, priced at ₹1,400 a spread, stated in advance so that the answer to a late addition is a number rather than an argument.

The dated dependency was the one that saved the schedule later. “Photography files by 4 March; delivery moves by the same number of days they are late” is not a defensive manoeuvre, it is the only honest description of how a schedule works, and clients accept it immediately because it is obviously true. The photographs arrived on 19 March. Because the consequence had been written down in week one and agreed, the fifteen-day slip was a fact both sides already knew the meaning of rather than a dispute about whose fault it was.

The enquiry was good on paper. A 320-bed district hospital was rebuilding its outpatient block and needed wayfinding: twenty-two signs in Marathi and English, four floor directories, ward numbering, and a pictogram set for the departments, since a large share of the people using the building could not read either language comfortably.

Meera quoted ₹1,40,000 with two rounds of revision. Then, in the qualifying call, she asked the question she had learned to ask: besides you, who else has to be happy with this?

The answer was five people. The medical superintendent, who had raised the budget. The civil engineer managing the block. The head of nursing, who would have to live with it. A trustee of the society that partly funded the building. And the public relations officer, who had commissioned the hospital’s existing logo and had opinions about its typeface.

None of them was named as the decision-maker. Asked directly who signed off, her contact (the civil engineer) said it would be “a consensus”.

She knew exactly what that meant, because groups do not produce five opinions, they produce a relationship between five people. Comments are made in response to what a colleague said, to what a senior person appeared to think, and to the need to be seen contributing. Agreement is easiest on what nobody actively dislikes, which is how committee review removes exactly the clarity a wayfinding system exists to supply. And five reviewers with no consolidator does not mean five comments; it means several rounds of contradictory comments arriving in fragments over weeks.

So the decision was whether to say this out loud before the project started.

Saying nothing was tempting. It was a ₹1,40,000 job in a year that had been thin, and raising process objections before a contract exists is the kind of thing that makes a cautious client choose somebody less difficult. There was a real chance that naming the problem would lose her the work.

Saying it meant quoting for what would actually happen rather than for what she hoped would happen. On her own records, projects with a committee had run between four and six rounds, not two. At roughly eighteen hours a round, the three extra rounds were about fifty-four hours (₹43,000 of work at her rate) which was the difference between a good project and one she would resent by March.

There was a third element she could ask for that cost the client nothing: one named consolidator who would resolve internal disagreements before anything reached her, and one named person who decided.

The kickoff took ninety minutes and produced a one-page note the same evening. What we are making. What is excluded: no exterior signage, no ward furniture labels, no ambulance bay, all named because a hospital is an environment where scope grows sideways. The stages. Who signs off. And the dependency list: the architect’s floor plans as editable drawings rather than a scan, a decision on whether the existing hospital blue was mandatory, the Marathi department names in their official spellings, and access to the outpatient block after five in the evening to photograph the sightlines. Each with a date and a stated consequence.

She also asked the four kickoff questions, and one of them changed the project. “What would make this a failure?” produced an answer nobody had put in the brief: the head of nursing said that if people still stopped at the lift lobby to ask a ward boy for directions, it had failed. That is a measurable outcome and it is not the same as legible signs. It moved the whole job toward decision points rather than toward a beautiful directory, and it is why the pictogram set ended up larger than quoted and the floor directories smaller.

The production questions came before anything was drawn. The fabricator’s maximum panel was 1200mm, the vinyl they used would not hold a stroke under 4mm at the viewing distance, and the corridor lighting was 4000K fluorescent, which is the fact that later settled the argument about the hospital blue. Ten minutes on the telephone, free, and expensive to discover afterwards.

The brief was ordinary and the deadline was real: a coaching centre with four branches wanted a refreshed identity, a prospectus and six hoardings, ready before the admissions push in March. The budget was ₹2,10,000. Karan had been recommended by their printer.

The client sent a folder on day one. It contained their existing logo as a 900-pixel PNG, a display typeface as a TTF file, eleven photographs of students in classrooms, and a scan of last year’s prospectus.

He asked the question he asks at every kickoff, framed as a scoping question rather than a legal one: are these photographs ones you commissioned, or bought from a stock library, or from somewhere else, and if they are stock, could you send the licence so I know what we can use them for?

The answers, over two days, were a problem.

The logo had been drawn in 2011 by the founder’s nephew, who was now a software engineer in Pune. There was nothing in writing. Under the default rule the nephew might still own the copyright in the artwork, and the centre might have only an implied licence to use it. Karan was about to redraw and extend that mark across hoardings the size of a wall.

The typeface had been downloaded from a font aggregation site. Nobody knew which licence. A search found it listed as free for personal use, with a commercial licence sold separately by a small foundry. A hoarding is about as commercial as use gets, and the face was detectable in any published PDF.

Six of the eleven photographs were of identifiable students, taken by a staff member at a previous year’s result celebration. No releases. Three more had been downloaded from a stock site by a former marketing executive who had left; there was no receipt and no account. Two were genuinely theirs, taken by the current administrator, of empty classrooms.

So Karan had a decision with a cost on both sides.

Saying nothing was entirely possible and extremely common. The material had been supplied by the client, his terms carried an indemnity clause for client-supplied material, and the probability of anyone noticing was, honestly, low. Raising it meant delivering bad news in week one, adding cost to a budget that was already agreed, and risking the deadline the whole project was built around.

Saying it meant being the supplier who caused a problem before the work had started. He was also aware that an indemnity from a coaching centre is worth what the coaching centre is worth, and that if a demand letter arrived it would go to the client and be forwarded to him, as the person who put the image in the layout.

There was a second conversation he had to have, and it was the one clients find most surprising. The centre believed that commissioning a logo meant nobody else could use it. That belief merges two entirely separate systems. Copyright arises automatically in the artwork and stops somebody copying the drawing; a trade mark protects a sign used in trade, by class and by territory, and only the client can obtain one through a registration process. Karan supplies the first and cannot supply the second.

Their proposed name was also a common Hindi word used by at least two other coaching businesses he found in ten minutes of searching free public trade mark databases and ordinary image search. He put the recommendation in writing, in the quote and again at handover: the design is original work and not copied from any existing mark; he is not able to carry out a clearance search or advise on registrability; and before they spend on six hoardings and four branch fascias they should pay a trade mark attorney for a search. That sentence is honest, it is genuinely useful to them, and it is what he would want to be able to point at if a conflict emerged in two years.

The last item was the one with the money on it. Six hoardings and eight thousand prospectuses is production at volume, which is where a typographical error stops being a correction and becomes a reprint. His terms capped his liability at the fee and excluded consequential loss, and he sent the final artwork with an explicit request: check every detail including the phone numbers, the branch addresses, the fee table and the spelling of every teacher’s name, and reply confirming this is approved for print. He read the numbers aloud backwards himself anyway. One branch telephone number was wrong.

Shalini made explainer videos: ninety seconds to three minutes, script to final render, mostly for software companies and two medical device firms. She charged between ₹95,000 and ₹1,40,000 and delivered about fourteen a year. Her effective hourly rate, measured properly, was ₹1,250.

She was tired of quoting. Every job began with a call, a proposal, two follow-ups and usually a scope conversation, and on her own records that cycle was averaging eleven hours per won project and about four hours per lost one. Across a year it was close to a month of unpaid work.

She had been building a course for seven months: “Motion Graphics for Product Teams”, about forty per cent recorded, in a folder she opened roughly twice a month. She had 1,900 newsletter subscribers, collected over three years, who opened at around thirty-eight per cent.

In October a client postponed and she had six clear weeks. She could finish the course, or she could do something else with them.

The course arithmetic, done honestly, was not encouraging. Revenue is audience times conversion times price, less the platform’s cut. At 1,900 subscribers, a launch conversion of two per cent (the upper end of what creative digital products do to an engaged list) and a price of ₹4,500, that is 38 sales and about ₹1,71,000 before fees, call it ₹1,55,000 after. Six weeks of work. She would also inherit permanent obligations: support email, re-recording when the software’s interface changed, and the fact that revenue in month six depends entirely on whether anybody is still pointing at it. Set against six weeks of client work at her current rate, the course was slightly worse in year one and possibly better over three, with all the uncertainty on the course’s side.

The alternative was the thing she had not considered because it was not exciting. Nine of her last fourteen projects had been near-identical: a ninety-second product explainer, one round of script, one storyboard, one round of animation changes, delivered in three weeks, for a company launching a feature. Her hours on those nine had ranged between 62 and 78. She knew the process, the file structure and the objections.

A fixed scope, fixed price, fixed timeline version of that (₹1,10,000, ninety seconds, one script round, two animation changes, fifteen working days from receiving the brief and the assets) would remove the quoting cycle entirely for that category. And every hour she saved by getting faster at a job she had done nine times would be hers rather than the client’s, which is not true of anything she bills hourly.

She had done the ceiling calculation once, a year earlier, and it was the reason any of this was on her mind. Twenty-five billable hours in a good week, forty-two working weeks after holidays and the fortnight in August when nothing moves, is about 1,050 billable hours a year. At ₹1,250 that is a ceiling of roughly ₹13,00,000, and the only two ways to raise it are to charge more, which her market limits, or to work more, which her body limits. Both limits were visibly approaching.

What she had not appreciated until she wrote it down is that the strongest argument for non-hourly income is not the money. A freelancer with a fifth of income arriving from somewhere else negotiates differently: she can refuse a badly-run project, hold a price through a silence, and wait for a better client. That changes the quality of the client work, which raises the rate, which raises the ceiling anyway. Judged that way, a second channel earning a modest monthly figure is buying the ability to say no, which is the most expensive thing in the trade.

She also set a stopping condition before she started, in writing, because judging afterwards is unreliable and sunk effort makes everything look almost-there. The condition was twelve sales of the productised explainer within twelve months. If it reached twelve she would keep it and start the next thing. If it did not, she would stop, take the salvage (the process notes, the dependency list, the add-on price list) and not spend a second year finding out.

Nandita did her first proper annual review in January, half a day with the records, because a peer group of four freelancers she met monthly had all agreed to do one and compare.

Turnover was ₹18,90,000, up nine per cent on the previous year. That was the number she had been watching, and it looked fine.

The other figures did not.

Her effective hourly rate (everything invoiced, divided by every hour including quoting, admin, revisions and the two afternoons a month she spent at a printer) was ₹640. Two years earlier it had been ₹710. Utilisation was 74 per cent, which is very high, and the high figure was not good news: it meant she was billing three days in four and doing almost no business development, which is how next year empties.

And one client, a mid-size advertising agency in Indore that subcontracted her for packaging artwork, was 68 per cent of income.

She had not chosen that. Nobody does. The agency had been pleasant, paid within forty days, and sent enough work to fill her calendar. Because she was full, she had stopped writing to anyone. Because she stopped writing, her other clients had drifted; two of them had not heard from her in fourteen months. Each individual month had been a sensible response to the month in front of it.

Three costs had already arrived and she could see all of them in the record now she was looking. Her rate had not moved in two years, because raising a price on the client who is two-thirds of your income is a conversation you cannot afford to lose, so she had never had it. She had absorbed four unpriced scope changes in eleven months without a single change note. And her portfolio had narrowed to one category, one market and one style, which was a problem precisely because it was the evidence she would need if the agency stopped.

She also recognised the disguised version of the problem. The agency’s work came from eleven different end clients, which had made it feel diversified. It is not. If the agency loses its largest account or restructures, all eleven go at once, and she has no relationship with any of them and nothing she is permitted to show.

So the question at the end of the half-day was not whether the concentration was a risk. It was what to do about it in a year when nothing was wrong and she was busy and reasonably content.

The costs of acting were real. Two hours a week of pipeline work, taken out of billable time, is about ₹66,000 a year of forgone income at her rate: spent on something that returns nothing in the week she spends it. Raising her rate with the agency risked the relationship that paid for most of her life. And she was, genuinely, enjoying the work.

The half-day had a second part she nearly skipped, which was five questions answered in writing rather than in her head, because writing catches the answers that thinking slides past.

Which project this year would she want three more of? A small packaging job for a spice cooperative, at a third of the agency’s fee, done directly, in six days, at an effective rate of ₹1,100. Which would she refuse? Two of the agency’s, both of which had run to five rounds with no change note. What had she learned that she could not do last year? Nothing she could name, which was the answer that bothered her most, because a year with no development costs nothing that year and a little of the next, and five of them cost a career.

What had she spent the most time on that nobody paid for? Artwork adaptation that fell outside every written scope she had. And the last question: if next year were identical to this one, would that be acceptable? She answered no, which is the only reason the three decisions happened. A year that is fine is easy to repeat indefinitely, and the repetition is the drift.

She noticed one more thing in the records that she had been reading as tiredness. She had begun referring to clients as a category, with contempt, in conversation with the peer group. That is the third of five recognised burnout signals and the point at which the remedies are still short, and almost all of what was producing it (absorbed scope, a rate that could not move, no control over which work arrived) was downstream of the concentration she had just measured.