Fiverr's annual active buyers fell to 2,676,000 in the year to 30 June 2026, down 21.9 percent from 3,425,000. Annual spend per buyer went the other way, up 15.6 percent to $368 (Fiverr Q2 2026 results, Form 6-K exhibit 99.1, filed with the SEC on 29 July 2026). Upwork reported the same shape three weeks later: 763,000 active clients, down 4 percent, and gross services volume per active client at a record $5,230, up 5 percent and growing sequentially for the eighth quarter running (Upwork Q2 2026 press release, 10 August 2026).
Two marketplaces that compete with each other, moving in the same direction. Fewer buyers, each one spending more. That is the market this question lands in, and it makes most of the advice about getting clients wrong in a specific way. The tactics people teach are aimed at the half of the market that is disappearing fastest.
What follows is the order I would work through the channels, what each one costs before it pays, and the numbers behind that claim. Every figure was checked on 7 September 2026 against the filing or the survey that owns it, and I name the source each time. Where a source is vendor research rather than a filing, I say so.
# Where a first client comes from, ranked by
# how fast it works.
#
# Fast, and the pool is tiny
# People who already know you
# A business in your own city
# Somebody's broken project you fixed
#
# Slow, and the pool is huge
# Marketplaces (28.0% cut on Fiverr)
# Search and writing
# Communities and open source
#
# Where the money is
# The second project from client one
#
# Nobody's first client comes from list two.
# Almost everybody's fifth one does.The work everyone teaches you to chase is the work that is shrinking
The transactional end of the market is contracting, and the marketplaces say so in their own filings. Fiverr cut full-year guidance to $356 million to $372 million, down 17 to 14 percent on 2025, and named the cause: "AI-related demand and traffic headwinds observed in recent weeks that have continued into the third quarter," plus "persistent weakness across categories most exposed to AI automation" (Fiverr Q2 2026 results, 29 July 2026, https://www.sec.gov/Archives/edgar/data/1762301/000117891326003624/exhibit_99-1.htm).
The same release describes what is growing. "Clients completing $1,000+ projects grew 13% y/y on a trailing twelve month basis." Upwork's numbers point the same way: gross services volume from AI-related work "increased more than 22% year-over-year," and inside that, AI strategy and consulting "grew over 50% year-over-year" (Upwork Q2 2026 press release, 10 August 2026, https://www.sec.gov/Archives/edgar/data/1627475/000162747526000046/upwork2q26-pressrelease.htm).
Buyer counts falling, project sizes rising. A $50 gig for a landing page sits in the shrinking half. A fixed-scope application build sits in the growing half. Both are hard to win, and only one of them is getting easier every quarter.
Getting a first client is a trust problem. Every channel below is a different way of borrowing trust you have not earned yet, and they differ mostly in how much of it they lend.
Nobody can find a full-stack developer
The supply side is the number that should decide your positioning. GitHub counted more than 180 million developers on the platform and more than 36 million new ones in a single year, 23 percent up on the year before. India added more than 5.2 million of those, over 14 percent of all new accounts worldwide, and now holds 21.9 million against 28 million in the United States (GitHub Octoverse 2025, published 28 October 2025, covering 1 September 2024 to 31 August 2025, https://github.blog/news-insights/octoverse/octoverse-a-new-developer-joins-github-every-second-as-ai-leads-typescript-to-1/).
Hold that against 763,000 active clients on Upwork. The two counts are not the same kind of thing, since a GitHub account is not a freelancer looking for work, and the mismatch is the point anyway. The buyer side of the largest marketplace is measured in hundreds of thousands and falling. The side you are standing on is measured in tens of millions and rising.
Which is why full-stack developer is not a position. It describes 21.9 million people in one country. What a client can hold in their head is the kind of business you build for. My studio lists five: e-commerce and retail, EdTech, healthcare, real estate and proptech, early-stage products (https://amitkumarraikwar.dev/studio). None of those is a stack. Each is a set of problems I have seen before, which is the only claim a first-time client can check.
Here is the test for whether your positioning works. Say it once, then see whether the person can repeat it to somebody else. "I build listing platforms for property dealers" survives that. "I do React, Node and MongoDB" does not, because it describes what you type instead of what the other person gets.
# The same person, described two ways.
#
# Findable by nobody
# full-stack developer
# React, Node, MongoDB
# I build websites and apps
#
# Findable by somebody
# listing platforms for property
# dealers
# attendance and fees for coaching
# institutes
# patient portals for single-doctor
# clinics
#
# The second list is not a smaller market.
# It is the same market, described in the
# words the buyer already uses.One live URL beats a portfolio of screenshots
A client cannot read your code and will not try. What they can do is open a link on their phone and decide in ten seconds whether you are serious. So the most valuable thing you own is one product that is live, that you can point at, and that has somebody other than you using it.
Ours is Finalist, at https://finalist.live/. It runs esports registration, slot lists, room details, results and leaderboards, with a Discord bot mirroring all of it into the server the players already sit in. Why the hard part turned out to be the clock rather than the feature list is at https://amitkumarraikwar.dev/blog/an-esports-platform-is-mostly-a-clock. The write-up matters less than the link. The link is what gets sent to somebody deciding whether to reply to you.
If you have nothing live, build for a business you can walk into. A coaching institute keeping attendance in a notebook. A dealer running enquiries through WhatsApp. A clinic with a paper appointment book. Those are small builds with a named user attached, and a named user is what turns a side project into evidence.
Then keep the demo cadence honest once you are being paid: something openable and clickable every week rather than a status report. It surfaces drift early in both directions, which is the only reason a fixed price stays fair (https://amitkumarraikwar.dev/blog/a-demo-you-can-click).
The channels, and what each one costs before it pays
Ranked by how fast a first client arrives: people who already know you, businesses in your own city, marketplaces, search, then communities. That order says nothing about which channel is best. It tracks how much trust each one lends you on day one.
# What each channel costs before it pays.
#
# Channel Cost First result
# People you know nothing days
# A local business your travel weeks
# Marketplace bid 28.0% cut weeks
# Search, writing 9 to 12 mo slow, compounds
# Open source months unpredictable
# Referral nothing after project one
#
# The 28.0% is Fiverr's marketplace take rate
# for the twelve months to 30 June 2026, up 40
# basis points on the year before.On marketplaces, read the take rate before deciding how long to stay. Fiverr's marketplace take rate was 28.0 percent for the trailing twelve months to 30 June 2026, up 40 basis points year on year (Fiverr Q2 2026 results). Upwork's Q2 2026 release states no take rate; dividing its revenue of $191.7 million by gross services volume of $966.4 million gives roughly 19.8 percent, and that arithmetic is mine rather than theirs.
A quarter of the invoice is a fair price for your first three reviews and a bad one for your fifth year. Treat a platform as somewhere to buy proof rather than as a channel. Both platforms forbid moving a contract you found there off-platform, so the exit is not that client. It is the next one, who came to you directly because the reviews existed.
On search, the ground has moved and it is worth knowing by how much before committing a year to it. Pew Research Center tracked the browsing of 900 US adults across 68,879 Google searches in March 2025. On result pages carrying an AI summary, people clicked a traditional search result in 8 percent of visits. On pages without one, 15 percent. Clicks on links inside the summary happened in 1 percent of visits, and browsing sessions ended on 26 percent of pages with a summary against 16 percent without (Pew Research Center, 22 July 2025, https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/).
Writing still brings work in. It brings it in slower than it did, and it now rewards being the source a summary has to name over being the tenth blue link. What I would compete on instead is at https://amitkumarraikwar.dev/blog/how-to-rank-a-website-on-any-keyword, and the short version is to pick the keyword nobody with a budget wants.
What the first call is for
Twenty minutes, free, and I am not pitching in it. I want three answers: what business goal sits behind the software, who touches it daily, and what happens to the client if it never exists. That last one sounds unkind and it is the most useful question on the list, because it separates a project with a deadline attached to something happening from a project somebody is thinking about.
Then a written proposal within three days. Scope listed item by item, fixed price, timeline, and a section for what is not included. That last section is the one that prevents arguments. If the scope changes the price changes and we both know it changed, because the original list is on paper. The long version of how I price a job is at https://amitkumarraikwar.dev/blog/the-quote-is-the-hard-part.
One more thing wins work with no skill attached: reply within one business day. Every enquiry, including the ones that are obviously never going to become projects. Most of the competition does not, and a client comparing three developers compares response times before they can compare code.
The mistakes that have cost me money
Twice I quoted a project on the assumption that the client had their content ready. Both times the build finished and then sat waiting on copy and images for weeks. The code was done and the project was not, and that time is billable to nobody. Content readiness is now a line in the proposal with a date against it.
The second one is integrations. A payment gateway I have shipped four times takes a day. A gateway I have never touched, in a country whose rules I have not read, can take a week, and I do not find out which until I am inside it. Those get quoted as a separate line now, with the reason written next to them.
The third is what this post is mostly about. Taking cheap work because it is there puts you in the half of the market both marketplaces just cut their guidance over, against people whose living costs less than yours.
Cheap work does not stay cheap work. It becomes the number the next client hears when they ask what you charge.
What to aim at, in numbers
A workable target for a first year is two fixed-scope builds and one retainer. At my rates that is an application from ₹1,50,000, usually up to ₹4,00,000 depending on scope, a marketing site from ₹40,000, and maintenance from ₹25,000 a month (https://amitkumarraikwar.dev/studio). Copy the structure rather than the numbers. The structure is what makes a year predictable: lumpy project income with a floor underneath it.
For where that sits against everyone else, 13.9 percent of respondents to the Stack Overflow Developer Survey 2025 described themselves as independent contractor, freelancer or self-employed, across 48,178 answers to that question (https://survey.stackoverflow.co/2025/work/). In India the figure was 9.3 percent, against 10 percent in the United States and 12 percent in the United Kingdom. Client work is a minority path even among developers who are already working.
The top of it is not small money. MBO Partners counted 5.6 million US independents earning above $100,000 in 2025, up from 4.7 million in 2024, in a survey fielded in April 2025. Read that as directional rather than exact. It is vendor research from a company selling services to independent workers, the summary page states no sample size, and it does not define what counts as independent (https://www.mbopartners.com/state-of-independence/).
A job is a channel too, which is worth saying because the choice usually gets presented as one or the other. The US Bureau of Labor Statistics puts software developer employment at 1,717,800 in 2025, projected to reach 1,892,600 by 2035, a rise of 10 percent, with a median wage of $135,980 in May 2025 (Occupational Outlook Handbook, page last modified 27 August 2026, https://www.bls.gov/ooh/computer-and-information-technology/software-developers.htm). Employment and client work fund each other in the first two years more often than either camp admits.
If the missing piece is paperwork rather than clients, that part is cheaper than people expect and I wrote out what is free and what has a bill attached at https://amitkumarraikwar.dev/blog/how-to-start-a-company-in-india-with-minimum-money. If the missing piece is which income streams to stack, the routes I have used are at https://amitkumarraikwar.dev/blog/five-ways-to-earn-money-full-stack-developer-2026.
The order, once more
Pick a kind of business instead of a stack. Get one thing live that somebody other than you uses. Ask the people who already know you before you ask strangers, then the businesses within an hour of your house. Use a marketplace to buy reviews and price that decision at 28 percent. Start writing in month one knowing it pays in month nine. Then treat the second project from your first client as the thing you optimise for, because that is where the money in this business has always been, and it is where both marketplaces are now pointing as well.
If you have a project and want the number in writing, the first call is twenty minutes and free, and the proposal follows within three days: https://amitkumarraikwar.dev/contact. I reply to everything inside one business day, including a no.
Frequently asked questions
- How do I get my first client as a software developer with no experience?
- Build one small thing for a business you can walk into, then use it as evidence. A coaching institute keeping attendance in a notebook or a dealer running enquiries through WhatsApp is a named user with a problem you can see, and a live URL beats a portfolio of screenshots while nobody can vouch for you. Ask the people who already know you before you ask strangers, because trust you already have is the cheapest kind to spend. Marketplaces work for a first review, at a price: Fiverr's marketplace take rate was 28.0 percent for the twelve months to 30 June 2026.
- Are Upwork and Fiverr still worth it in 2026?
- For buying your first few reviews, yes. As a permanent channel the numbers argue against it. Fiverr's annual active buyers fell 21.9 percent year on year to 2,676,000 as of 30 June 2026 while spend per buyer rose 15.6 percent to $368, and it cut full-year guidance citing "persistent weakness across categories most exposed to AI automation." Upwork's active clients fell 4 percent to 763,000 while volume per client hit a record $5,230. Both are consolidating into fewer, larger buyers, so the low-priced transactional work that is easiest to win there is the work shrinking fastest.
- How much should I charge my first client?
- A fixed price, for a scope written down item by item, higher than the number that feels safe. What you take from client one becomes what client two hears when they ask your rate, and moving a price up later is harder than starting there. My own floor is ₹1,50,000 for an application and ₹40,000 for a marketing site, with maintenance from ₹25,000 a month. If you cannot defend a number yet, defend the scope instead: fewer features at a price you can live with, delivered on the date you promised.
- How long does writing or SEO take to bring in clients?
- Months, and slower than it was two years ago. Pew Research Center found that on Google result pages carrying an AI summary, users clicked a traditional result in 8 percent of visits against 15 percent on pages without one, from tracked browsing of 900 US adults across 68,879 searches in March 2025. Writing still compounds and is still worth starting in month one, but plan for it to pay around month nine and fund that year with work from people who already know you.
- Do I need a registered company before taking client work in India?
- No. You can invoice as a sole proprietor under your own PAN with nothing filed at the Registrar of Companies, take the free Udyam (MSME) registration, and open a current account in the business name. GST registration comes in when the threshold for your supply type and state applies, or when a client requires it. What a proprietorship does not give you is limited liability, shares to hand anyone, or DPIIT startup recognition. Those are the reasons to incorporate, and looking established is not one of them.
Written by Amit Kumar Raikwar, full-stack engineer & product designer in Indore, India. If you want something built, start here.
Working on something?
Let's build it together.
I take projects from an empty Figma file to a live product. Fixed scope, weekly demos, code you own.