Why a five-dollar survey is not five dollars an hour

A five dollar survey listed at “about 15 minutes” reads like $20 an hour. Time it yourself with your phone’s own stopwatch, not the platform’s estimate, and the real number is usually lower, sometimes by half, sometimes zero if a screener cuts you at question four.

That gap is not an accident and it is not really about slow typing. It is baked into how market research panels price a “complete”: someone estimated the length of interview during a pilot test, the panel published a flat reward based on that estimate, and nothing in the system corrects for the respondents who get bounced out halfway, restarted by an attention check, or funneled through three studies before one of them takes them. None of that time gets paid, because none of it produces a completed survey.

Readers of how much paid surveys actually pay per hour already know headline hourly rates on these platforms sit in a wide range. This piece is the arithmetic underneath that range: why the advertised time and the real time diverge, how incidence rate and routers turn “screened out” into unpaid labor, and how to measure your own effective rate instead of trusting the platform’s clock.

background and prior art

Market research has run on two numbers since long before the web existed: length of interview (LOI) and incidence rate (IR). LOI is how long a study is expected to take, set from a small pilot batch before the study opens to the full panel. IR is the percentage of the general population who will actually qualify once they hit the screener questions, also set from that pilot. A sample provider prices a study using both, and the pricing convention across the industry (ESOMAR’s guidelines for sample and data quality describe the same LOI and IR framework used by most sample providers) treats low-IR studies as more expensive per complete, because more people have to be screened to find one qualifying respondent. That screening cost gets built into the researcher’s budget line, not into the respondent’s paycheck.

None of this is new. Phone and mail panels ran the same LOI and IR math for decades, they just moved slowly enough that nobody noticed the unpaid screening time as a distinct cost. Online panels compressed the whole loop to minutes and made the mismatch visible in real time. A respondent can now hit four screen-outs in the time it used to take a phone interviewer to place one call.

The clearest documentation of what that mismatch does to actual pay comes from crowd work rather than consumer panels. A widely cited 2018 study presented at the ACM CHI conference by Kotaro Hara and coauthors pulled real Amazon Mechanical Turk task logs and found the median hourly wage across the platform sat far below any US minimum wage, because MTurk has never imposed a platform-wide pay floor and leaves per-task pricing entirely to requesters. Consumer survey panels aimed at the general public sit in a different spot. Prolific responded to that same body of research by requiring studies to clear a minimum hourly rate before they’re allowed to launch, while most mass-market panels (Swagbucks, InboxDollars, Branded Surveys, YouGov and similar) publish a flat reward per survey and a cash-out minimum, with no hourly guarantee attached to either number. Survey panels versus micro-task platforms covers that split in more detail. The gap this article is about lives inside that middle ground: a published reward that assumes a clean run, applied to a session that rarely is one.

It’s also worth naming the base rate everyone’s mentally comparing against. The US federal minimum wage has sat at $7.25 an hour since 2009, and that floor applies to employees, not to independent respondents doing piece-rate survey work. There is no legal requirement that a survey pay minimum wage per hour, because a survey taker isn’t an employee of the panel. Whatever floor exists is one the platform chooses to impose on itself, or doesn’t.

the core mechanism

Strip out the marketing and a survey payout reduces to one formula:

effective_rate = payout / (actual_minutes / 60)

The advertised rate uses the panel’s estimated minutes. Your effective rate uses the minutes you actually spent, and those two numbers only match on a clean run. Four things reliably drive them apart.

The first is the screener itself. Before the LOI clock effectively starts, you answer age, income bracket, employment status, product ownership, whatever qualifies you for the study. That takes real time and, on most consumer panels, pays nothing if you don’t qualify. Filling prescreeners properly covers the mechanics of answering these without tripping a quota that’s already full, which is its own separate time sink layered on top of this one.

The second is incidence rate. If a study’s IR is 20%, the panel’s own pricing assumes roughly one in five attempts qualifies. The number of attempts before a qualifying hit follows a geometric distribution, and its expectation is simply 1 divided by IR:

expected_attempts = 1 / incidence_rate
expected_screen_time = expected_attempts * seconds_per_screener

At a 20% IR and a 90 second screener, that’s 5 expected attempts and 450 seconds, 7.5 minutes, of screening layered on top of whatever study you eventually land in. None of that shows up in the panel’s advertised “15 minutes,” because the panel’s estimate is for the study, not for the path to reach it.

The third is the router. Many consumer panels don’t send you straight into a named study, they send you into a router that tries to match you against several studies in sequence, bouncing you to the next one each time you fail a quota or a screener. Survey routers and why you get bounced between studies covers the mechanism in full, but for the hourly math the router does exactly what a bad incidence rate does: it inserts unpaid time between you and a paid complete, and it can do that two or three times in a single session before you either land a study or give up.

The fourth is the attention check, and it’s the one that costs the most because it strikes late. A trap question, a straightlining detector, or a speeder flag can void a submission after you’ve already answered forty questions, at minute eighteen of a twenty minute study, for a $0 result. Panels build these checks to protect data quality against bots and low-effort respondents, which is a legitimate need on their side. From the respondent’s side, though, a rejection at minute eighteen is strictly worse than a screen-out at minute two. Same zero dollars, sixteen more minutes gone.

None of these four are bugs. They’re the natural result of panels being paid by researchers per completed survey, not per respondent-hour. The panel’s incentive is to maximize completes per unit of its own operating cost, and unpaid respondent time isn’t a cost that shows up on its books at all. Put together, the real formula for a session isn’t payout over advertised time, it’s:

effective_rate = payout / ((screener_time + router_time + study_time) / 60)

where payout is often zero and the other three terms are rarely zero.

worked examples

Say a study lists $5 and “about 15 minutes.” You open it, answer a two minute screener, qualify, and the study itself runs long because of a grid question that reloads slowly on your connection. You finish at 22 minutes total. $5 over 22 minutes works out to $13.64 an hour, not the $20 the listing implied. That’s still a perfectly usable rate. The version to watch for is the one where a mid-survey attention check flags you as a speeder at minute 18 and the platform voids the whole submission: same 20-plus minutes spent, $0 paid, an effective rate of $0/hr for that block, and depending on the panel’s policy, a strike against your account that account under review explains the consequences of.

Now run the incidence rate math from the last section as a session rather than a single study. You open a panel, hit a router, and get bounced through three screeners at roughly 90 seconds each before landing in a study that pays $4 for an advertised 10 minutes. Screening cost you 4.5 minutes, the study takes its full 10, total time is 14.5 minutes for $4, which is $16.55/hr, not the $24/hr the “$4 for 10 minutes” listing implied on its own. Land in a study with a worse incidence rate, or fail to qualify for any of the three and end the session with nothing, and the number gets worse fast. A week of screen-outs is what that compounding looks like across many sessions instead of one.

Points-based panels add a cost that has nothing to do with per-minute pay at all: liquidity. Swagbucks, for instance, has long converted points at roughly 100 SB to one US dollar and won’t process a PayPal cash-out below $25. Say you earn 250 SB a day from surveys that took a genuine 25 minutes combined. That’s $2.50 for 25 minutes, an ordinary $6/hr. But you can’t touch that money until you’ve accumulated $25 worth, which at that pace is ten days out. Your nominal hourly rate is fine. Your effective rate, if you count the time value of the money at all, is worse, because a chunk of every day’s earnings sits frozen behind a threshold you haven’t cleared yet. This is the part of the math flat per-survey reviews tend to skip, and it’s worth checking a panel’s published minimum cash-out before judging its per-survey rate in isolation.

edge cases and failure modes

incidence rate mismatch: a listing’s advertised time assumes the panel’s own IR estimate holds for you specifically. It won’t always. If you fall into a demographic that fills quotas fast (a common age bracket, a common job title, a common income band), your personal effective IR runs lower than the panel’s blended average and your screening time compounds across a session. Track your own qualify rate over a week instead of trusting the panel’s headline number, and deprioritize categories where you keep getting bounced.
prescreener answers that read as inconsistent: quotas fill fast on common demographics, so a screener that looks two minutes long can end the moment you tick the wrong box, sometimes by design, to protect a quota that’s already full. Filling prescreeners properly and why you keep getting screened out of surveys both cover the specific patterns worth watching for. Answer honestly and consistently across panels, since flip-flopping demographic answers between studies is itself something some panels quietly flag.
attention check rejection after a long completion: the failure mode from the first worked example. Read grid and matrix questions properly even when they feel repetitive, since that’s exactly where straightlining detectors sit, and keep an eye on account status after a rejection since repeated voids can trigger a fuller review rather than a one-off zero.
cash-out minimums that lock in earned value: covered in the points example above. Before committing real time to a points-based panel, check its published minimum cash-out and conversion rate, and weigh that against how often you’ll realistically clear the threshold.
listings that promise a flat hourly rate with no panel behind them: a small number of “surveys pay $30/hr” ads circulating outside the major panels are closer to work-from-home scams than research studies, especially any that ask for payment up front or banking details before you’ve done anything. The FTC’s guidance on job scams is a reasonable baseline for what to check before handing over information. This isn’t legal advice, just a sanity check: legitimate panels pay you, they don’t charge you to start.

what we learned in production

Building out the panel breakdowns on this site means reading a lot of terms of service and payment policy pages rather than running every panel personally end to end. The pattern that shows up most often isn’t dishonesty, it’s omission. Panels are generally accurate about what they’ll pay for a completed study, and close to silent on what happens on a screen-out, a void, or a slow-loading question that pushes you past the estimate. If a policy page doesn’t say what happens when you fail a screener, assume the answer is $0 and plan your time around that assumption rather than the advertised one.

I’d rather see a panel publish a flat, boring per-minute rate than wrap piece-rate pay in points, badges and streak bonuses that make the actual dollar-per-hour harder to see, not easier. That’s a minority view in an industry that treats gamification as a retention feature rather than a pricing fog, and I’ll take the argument on it, but the math above is the reason I hold it: every layer between time spent and dollars received is a place the effective rate can quietly drop without the respondent noticing.

None of the figures in this piece are guarantees. Panels change reward tables and cash-out minimums without much notice, and we build these breakdowns from what’s published plus patterns readers report, not a live feed into every platform’s backend. Treat any specific number here as a snapshot from around publication, not a promise it holds next quarter. For the wider spread of what pays across formats, see paid interviews and focus groups at the top end, and the blog index for everything in between.

references and further reading

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 2026-09-18.