The market research supply chain, and why geography sets your pay
Prolific publishes a pay floor. When I last read their researcher documentation it was £6.00 or $8.00 an hour, with £9.00 or $12.00 an hour recommended. As I read it, the floor applies to every participant regardless of the country they log in from. Check the current figures on their site before you rely on any of this, because a company can change a number like that whenever it likes. Most of the other panels whose terms I have read publish no hourly floor at all.
That gap is what this article is about. On a panel with no floor, nobody has promised you a rate, so your rate is an outcome. It gets decided several steps upstream of you, by a client with a budget, an agency with a questionnaire, a marketplace matching supply to demand, and a panel that decides how much of the price to pass down. Country shows up at nearly every one of those steps.
I covered the symptom in why singapore sees fewer surveys than the us. This piece goes upstream to the cause, and it is written for people who already know what a screen-out is. One limit up front: no panel publishes its margin and no marketplace publishes its rate card, so I cannot tell you the split on your panel. What I can do is lay out the structure from public documents, show a small model with every assumption labelled, and give you a script to measure your own rate. I am not going to promise anyone a payout. Any number below that does not come from a published term is a toy input, and I say so each time.
background and prior art
Online access panels replaced most phone and mail recruitment for consumer research from the late 1990s onward. The original model was simple. One company recruited members, profiled them, paid small incentives and sold completed interviews to agencies. Then buyers wanted more reach than any single panel could give, and sample marketplaces appeared to route one project across many suppliers. Cint runs one of the better known ones, and Dynata and Toluna are examples of large panel owners. Alongside them sit survey tools like Qualtrics and SurveyMonkey, which let a researcher order respondents from inside the tool, and academic-first marketplaces like Prolific, where the researcher deals with participants directly and the platform sets a floor. So it is not one supply chain but several stacked on top of each other, and the same person can be reached through more than one of them.
The rules of the road are mostly published. The ICC/ESOMAR code is the ethics baseline for the industry, and ISO 20252 is the quality standard some suppliers certify against. Neither says what a respondent should be paid, which is why pay is a market outcome and not a regulated one. On the economics side, the tool for comparing prices across countries has been around for decades: purchasing power parity. The World Bank publishes GDP per capita and PPP conversion factors by country, and those two series are the quickest way to see why a flat global reward would make no sense to a panel operator. Then there is quality. Pew Research Center has published work showing that online opt-in samples can produce misleading results for some groups, with bogus respondents as part of the reason. That research is a big part of why buyers now pay for verification, and why panels care where you really are.
the core mechanism
the layers
A survey payout is the last leg of a relay. This is the chain as I understand it from published material and from how the incentives have to work.
- the client: the brand, the government department, the campaign or the professor with a grant. They own the question and the budget, and they are the only party who pays for the research itself.
- the agency or insights team: turns the question into a questionnaire, picks a sample size, sets quotas by age, gender and region, and orders sample. Sometimes the client is its own agency.
- the marketplace or aggregator: matches the order to supplier inventory and takes a fee.
- the panel or traffic source: owns the relationship with you. It recruits, profiles, pays, and decides what share of the price reaches your balance. Rewards sites such as Swagbucks sit at this layer, and many of them fill their survey lists with a mix of their own and partner projects.
- you.
Every layer has to make a margin, and the money runs the opposite way to the questions. The client pays a price per complete, usually called the CPI (cost per interview), and by the time it reaches you it has been split at least three times. The splits are private and they differ by deal. If someone tells you the exact percentage that reaches respondents on a given panel, ask where the number came from.
Your hourly rate is nobody’s KPI.
what sets the price
Buyers and suppliers quote a CPI from a short list of inputs.
- length of interview (LOI), in minutes
- incidence rate (IR): the share of people who enter a screener and qualify
- target difficulty: a general adult panel is cheap, a hospital IT director is not
- sample size and quota design, because the last few cells of a quota are always the slow ones
- country or region
- fielding speed, plus whatever quality checks the buyer insists on
I go through the buyer’s side of that quote in how panels price a study. For the respondent’s side I use a simple cost model, and it is mine, not any vendor’s formula.
cost_per_complete = R + (1/IR - 1) * T + (1/IR) * E
CPI_floor = cost_per_complete / (1 - margin)
R reward paid to someone who completes
T consolation paid to someone screened out
E cost of each entrant: recruiting, retention, tech, fraud checks
margin the supplier's target margin on the price
It is good for one thing. It shows that low incidence pushes cost up faster than length does, because every complete needs 1/IR entrants and 1/IR - 1 of them are screen-outs somebody has to fund. At 50% incidence that is one screen-out per complete. At 10% it is nine.
where geography enters
Demand. Clients buy research where their customers are, and customers are concentrated. A global consumer study usually starts from a shortlist of large markets, and smaller ones get added as an extra cell or left out. In-home product tests are the extreme case, since a physical product has to reach a physical address and logistics decide who is asked at all. I went through that in in-home product tests, how they work and what they pay. Demand also has a clock. Fielding follows the buyer’s working day and quota cells fill from the front, and the timing side has its own write-up in when new surveys actually appear and whether timing matters.
Pool size against quota. Singapore has about 6 million people and the US roughly 340 million. A brief for 300 completes is the same size in both places, so it takes a far bigger bite out of the small pool. Suppliers see this in feasibility checks before a project starts, and a project that looks slow or costly gets a higher CPI or gets declined. Example 2 below puts numbers on it.
Local reward norms. Panels keep reward tables per country, and the anchor is local cost of living and what competing panels in that country offer, not the client’s home price. PPP conversion factors are the standard tool for that comparison, and it is fair to assume operators reason in the same spirit even if they never say so. That is inference, not a published policy. One thing does surprise people. Singapore’s nominal GDP per capita in the World Bank series has been in the same range as the US, so income alone does not explain a thin Singapore feed. Pool size and client demand do more of the work.
Fraud pricing. Buyers pay for clean data and cleaning is expensive. Suppliers check IP location, device, language and sometimes documents. When a region has produced clusters of duplicate or low-effort accounts, the filters against it get blunt, and real people in that region absorb the friction. Verification is the price of staying in a pool buyers trust, and why panels ask for id and what verification protects goes through it from the panel’s side.
Payment rails. Your reward may be points, then a dollar balance, then a PayPal payout, then a conversion into whatever currency your bank uses. Each step can carry a fee or a spread, and thresholds and expiry rules are set per panel. The processor’s own fee schedule is the source of truth on cost, and I would rather you read it than trust a figure from me.
the respondent side
Everything above sets what the panel is paid and what it can pass on. What you actually earn is a different quantity.
effective_rate = total_reward_received / total_hours_spent
Hours spent includes the screener, the router, the login, the screen-outs and the over-quota kicks, not only the completes. The “10 minutes” on a survey card is what the panel expects a typical completer to take. It says nothing about the four minutes you spent being screened out of the previous one.
worked examples
example 1: prolific’s published floor
This is the one place in the article where I can give you numbers from the operator’s own terms. As last published, the minimum is £6.00 or $8.00 an hour and the recommended rate is £9.00 or $12.00. A researcher sets a reward and an estimated time, and Prolific checks one against the other. The minimum reward for a study is therefore minutes times the hourly figure.
| study length | £6.00/hr | £9.00/hr | $8.00/hr | $12.00/hr |
|---|---|---|---|---|
| 5 min | £0.50 | £0.75 | $0.67 | $1.00 |
| 15 min | £1.50 | £2.25 | $2.00 | $3.00 |
| 30 min | £3.00 | £4.50 | $4.00 | $6.00 |
| 60 min | £6.00 | £9.00 | $8.00 | $12.00 |
Two things to notice. First, the floor does not move with where you live, so geography changes the invitation and not the price. Whether you see a study depends on the researcher’s filters, which are usually country and language first, and on how many active participants from your country exist. That is where prescreening comes in, and how prolific prescreening works covers it. Second, as I understand it the floor is checked against the researcher’s estimated time. If the estimate is low, your real rate is lower, and timing yourself is the only fix.
example 2: a 300-complete brief in two markets
Take one brief. 300 completes, 12 minute LOI, a consumer target with 10% incidence in both countries. Every figure below is a toy input, chosen to show direction. None is a quote from any supplier. The only assumption that separates the columns is E, the cost of each entrant, which I set higher for the small pool because more of each invite wave goes to people who were already invited last month.
| large pool | small pool | |
|---|---|---|
| IR | 10% | 10% |
| R (reward to completer) | $1.00 | $1.00 |
| T (screen-out consolation) | $0.05 | $0.05 |
| E (cost per entrant) | $0.10 | $0.25 |
| cost per complete | $2.45 | $3.95 |
| CPI floor at 30% margin | $3.50 | $5.64 |
Work the large-pool column: 1.00 + 9 x 0.05 + 10 x 0.10 = 2.45, divided by 0.70 gives 3.50. The small pool comes to about 60% more for the same brief, and notice that R, the number you see, is identical in both columns.
Now the pool arithmetic. 300 completes at 10% incidence needs 3,000 entrants. If one in five invited members opens and starts (another toy input), that is 15,000 invites. Suppose a supplier has 50,000 active members in the small market, again an assumption. One project would invite 30% of everyone they have. Scale that pool by the population ratio of roughly 57 to 1 and you get about 2.8 million members, where the same 15,000 invites touch about 0.5% of the pool. That scaling assumes Singapore has the same panel penetration as the US, which is generous to Singapore.
So what happens to a brief like that? I can think of three outcomes. The buyer pays more, the buyer skips the small market, or the supplier trims R to fit a fixed budget. Only the first leaves the respondent whole. The other two are consistent with the thin feed described in why singapore sees fewer surveys than the us, though I cannot prove which one applies to any given project.
example 3: your own attempts log
The only rate that matters is yours, and you can measure it with a spreadsheet and a stopwatch. Log every attempt, including the screen-outs, with the minutes it cost and what it paid. Use the currency that lands in your account.
panel,minutes,outcome,reward
panelA,12,complete,1.10
panelA,4,screenout,0.00
panelA,9,overquota,0.00
panelB,20,complete,2.40
Those four rows are made up to show the format. This script turns a real file into a rate per source.
import csv
from collections import defaultdict
totals = defaultdict(lambda: {"minutes": 0.0, "reward": 0.0, "n": 0, "complete": 0})
with open("attempts.csv", newline="") as f:
for row in csv.DictReader(f):
t = totals[row["panel"]]
t["minutes"] += float(row["minutes"])
t["reward"] += float(row["reward"])
t["n"] += 1
t["complete"] += row["outcome"] == "complete"
for panel, t in sorted(totals.items()):
hours = t["minutes"] / 60
print(f"{panel}: {t['reward'] / hours:.2f}/hr over {t['n']} attempts, "
f"{t['complete'] / t['n']:.0%} complete")
On the fake rows, panelA works out to 2.64 an hour because two of its three attempts paid nothing, and panelB to 7.20. The values are meaningless. The shape is the lesson: panelA’s card shows a decent reward per complete, and its rate does not. Keep one file per week, keep sources separate, and never average across panels.
edge cases and failure modes
the points illusion
Points, coins and balances in a panel’s own unit tell you nothing until you convert them into the currency you finally receive, after the processor takes its share. The counter is to log rewards in that final currency and to read the payment processor’s fee page yourself. Whether panel income is taxable where you live is a question for a tax professional. This is not legal or tax advice.
the late screen-out
Being told you are over quota after eight minutes is the most expensive event in your log, and it is more common where pools are small and quotas fill quickly. Track outcome and minutes for every attempt and drop sources where late screen-outs dominate. Complete your profile fields so the front-end targeting can work, and on Prolific use the prescreeners so the platform filters before you spend any time.
location mismatch on a real account
Expats, people who travel, people on a corporate VPN and people on mobile roaming can all end up with an IP in one country and a profile in another. Anti-fraud systems flag exactly that mismatch. The counter is boring: turn the VPN off, update your residence in your profile when it changes, and write to support before you take surveys from abroad. The flip side needs saying plainly. I will not tell anyone to fake a location to reach a better-paying market. It breaks the terms, it is the pattern these systems exist to catch, and why survey accounts get banned shows how that tends to end. In the terms I have read, a ban can forfeit the balance.
verification and one person, one account
Some panels ask for ID or proof of residence, and expats often hold documents issued elsewhere. Use the panel’s documented route and ask support what they accept. One person means one account per panel, and keeping one account clean covers the habits that keep it that way. Borrowing or buying someone else’s identity is fraud, and it is outside what I will write about.
dormant balances and payout thresholds
In a low-volume market a balance builds slowly, and thresholds, inactivity clauses and expiry rules can quietly eat it. Read those terms for every panel where you hold money. If a threshold is reachable, cash out at it rather than letting the balance sit as a claim on a company you cannot audit.
what we learned in production
The thing I keep coming back to is that reward per survey is the wrong unit. A card that says $1.50 for 10 minutes reads as $9 an hour, and it means nothing until you know your completion rate and your screen-out time. That is why I distrust “best countries for survey income” lists. They rank published reward tables and ignore who actually gets invited, and invitations are what decide your week. You may want to argue with me here, and that is fine. I would rather someone in Singapore spend their effort on one clean account, a complete profile and sources that publish a floor than on hunting for a better country. I would also treat exchange-fed panels as a top-up I do not plan around, and spend the first month measuring. That is reasoning from incentives and from reading terms. I am not holding up a spreadsheet of my own payouts to back it, and I would not trust anyone who did without showing the log.
What I have not seen is the inside of a marketplace. Everything in the mechanism section is inference from public documents and from how the incentives have to work, and the toy model is a teaching device, not a benchmark. If you work at a sample supplier and I have a step wrong, I would like to hear it. The rest of the site is indexed at the blog.
references and further reading
- World Bank, GDP per capita (current US$): the nominal series for comparing countries, including Singapore against the US.
- World Bank, PPP conversion factor, GDP (LCU per international $): the standard price-level comparison behind local reward tables.
- Pew Research Center, online opt-in polls can produce misleading results: why buyers pay for verification and care where respondents really are.
- ESOMAR, codes and guidelines: the ethics baseline for market research, including the ICC/ESOMAR code.
Written by Xavier Fok
disclosure: this article may contain affiliate links. if you buy through them we may earn a commission at no extra cost to you. verdicts are independent of payouts. last reviewed by Xavier Fok on 2026-09-26.