A week of tracked screen-outs: the honest log

I built this log because I kept seeing the same complaint in survey forums and in our own reader mail: people feel like they spend more time getting rejected by surveys than actually taking them, but nobody has numbers to back that up. So I put together a simple tracking sheet, the kind anyone can copy in five minutes, and used it to structure how a week of survey activity across a handful of panels actually gets measured.

This isn’t a payout report. I’m not going to tell you I earned some specific dollar figure by the end of the week, because that number depends entirely on your demographic profile, your region, and which studies happen to be open when you log in. A single week from one person tells you almost nothing useful about your own results. What the log is good for is different: it turns “surveys feel like a waste of time” into a rate you can calculate for yourself, on your own panels, in your own week, using your own numbers instead of mine.

The finding worth writing up is that the screen-out rate itself is the number to track, not the payout. Once you have a real screen-out rate and a real average time-to-screen-out, you can work out your effective hourly rate before you’ve cashed out a single reward, and that number tells you more about whether a panel is worth your time than any review does. I’ve covered the mechanics of why screening happens at all in why you keep getting screened out of surveys; this piece is about measuring it.

the setup

The log is a plain spreadsheet, nothing fancier than Google Sheets. One row per survey attempt, with these columns:

date and time of the invite
panel name
source of the invite: email, in-app notification, or panel dashboard
quoted length in minutes, as shown before you click start
start timestamp
outcome: completed, screened out, or quota full
end timestamp
actual minutes spent, calculated from the two timestamps
reward, recorded in the panel’s native currency (points, SB, or cash)
category tag, meaning your best guess at what the screener was filtering on: income bracket, homeownership, employment status, a specific product category, a health condition, parenting status

I ran this across four panels that between them cover most of what readers ask us about: Prolific, Swagbucks, YouGov, and Toluna. I’ve compared two of these head to head in Toluna vs YouGov if you want panel-specific detail; here I only care about what the timestamps show, not which panel “wins.”

Cost was zero beyond time. No paid tools, no browser extension for timing, just each panel’s own on-screen clock where it exists and a manual timestamp typed into the sheet where it doesn’t. That manual step turned out to be the weakest link in the whole setup, and I’ll get to why below.

The category tag column is the one people skip, and it’s the one that makes the log worth keeping past week one. A screen-out rate on its own tells you time is being lost. A screen-out rate broken out by category tells you where, which is the only version of the number you can actually act on.

what worked

A few things made the log usable instead of just a pile of rows.

timestamping the moment you click into a survey and the moment you land on a screen-out or completion page, rather than relying on the panel’s own estimate. Quoted lengths are frequently wrong, usually on the low side, and I go into why in why a ten-minute survey takes twenty-five.
separating the pre-screener, meaning the two or three quick questions a panel dashboard shows before it even routes you to the client’s survey, from an in-survey screen-out that happens five or ten minutes in. These cost very different amounts of time, and lumping them into one “screened out” bucket hides the actual problem.
tagging the likely screening category on every attempt, even when it’s a guess. After a week you can sort by tag and see which categories, whether that’s income bracket, a specific product you don’t own, or a health condition, are burning the most attempts without a single completion.
checking each panel’s published screen-out and payment policy before assuming a screen-out always means zero reward. Prolific, for example, documents in its participant help center how it handles cases where a researcher removes a participant partway through a study, which is worth reading before you assume every screen-out pays nothing.
keeping the reward column in the panel’s native unit instead of converting to cash on the fly. Swagbucks and Toluna pay in points, YouGov pays in points redeemable for cash or sweepstakes entries depending on the market, and converting each row by hand mid-week just introduces errors. I converted once, at the end, and I’ve written separately about why that exchange rate matters more than panels let on in points, coins, and the exchange rate panels do not show you.

what broke

Three things broke in the first pass at this log, and I want to be upfront about them since they’ll break for you too if you copy this format.

some panels don’t show a screen-out message at all. They just redirect back to the dashboard with the survey greyed out, or show a generic “no surveys available” page. There’s no clean way to timestamp a rejection you can’t see happen. The workaround was to note the URL pattern each panel uses for a genuine screen-out versus a technical error, and to mark anything that didn’t match a known pattern as “unclear” rather than guess and pollute the data.
quoted survey length turned out to be close to useless as a standalone column, because actual time so often ran long on the surveys that didn’t screen you out early. I kept the quoted-length column for reference but stopped using it in any calculation, and switched entirely to timestamp-derived duration.
double-counted invites. Several panels send an email and also surface the same study inside the dashboard, and early in the week I logged both as separate attempts, which inflated the total count. The fix was adding a survey ID, or a truncated version of the study title, as a dedupe key checked before a new row goes in.

the numbers

I’m not handing you a real payout total, because a single week from one person tells you nothing about your own rate. Your region, your demographic profile, and which studies happened to be live that week all move the number more than anything you personally control. What follows is the math the log is built to produce, worked through on a hypothetical week so you can see how to run it on your own data.

Say your log for the week shows 40 total attempts across four panels. Of those, 14 end in a screen-out, 3 hit a quota-full before you finish the pre-screener, and 23 complete. That’s a screen-out rate of 35 percent (14 divided by 40), which is already a useful number: more than a third of your invites are dead ends before you see any reward at all.

Now bring in time. Suppose the 14 screen-outs average 4 minutes each before you’re kicked out, for 56 minutes total. The 3 quota-fulls average 1 minute each, for 3 minutes. The 23 completions average 14 minutes each, for 322 minutes. Total time for the week is 381 minutes, or 6.35 hours. If the 23 completions paid a combined $38 across the four panels, your effective hourly rate isn’t $38 divided by the 322 minutes you spent actually completing surveys, it’s $38 divided by all 6.35 hours, screen-outs included. That works out to roughly $6 an hour, not the $7.08 an hour the completions alone would suggest. For comparison, the current US federal minimum wage is $7.25 an hour, and several state minimums sit well above that, so the gap between the two numbers is worth sitting with before treating survey time as equivalent to paid work. I’ve broken down the same calculation without the screen-out time folded in, in how much paid surveys actually pay per hour.

There’s a scaling ratio hiding in that math too. If you could cut the screen-out rate from 35 percent to 20 percent while completion pay and duration stayed roughly the same, the same 40 attempts would cost noticeably less total time for the same or better total reward. That’s the entire argument for tagging categories and being selective about which invites you open, rather than clicking every one that lands in your inbox and hoping it works out. Cost per qualifying attempt, meaning total time divided by completions rather than total attempts, is the closer analogue here to a customer acquisition cost, and it’s the number that improves fastest once you start skipping categories your own log shows you rarely qualify for.

lessons

track time, not just reward. A panel that pays well per completion can still have a poor effective hourly rate if its screen-out rate is high, and you’d never see that from the payout figure alone.
the screen-out rate is panel-specific and profile-specific, not a fixed property of “surveys” in general. Research industry bodies like ESOMAR publish standards around sample and panel quality, but even those don’t tell you your own rate. Your own log, for your own demographic, is the only version of this number that means anything to you.
read a panel’s screen-out and payment policy before assuming the worst. Some panels partially compensate a screen-out under specific conditions, and whether that applies usually depends on how the individual study was configured by the researcher, not on the panel as a whole.
quoted survey length is closer to a marketing number than a planning number. Survey methodology guidance, including Pew Research Center’s notes on writing survey questions, points to how easily estimated completion time drifts once real respondents are involved. Treat the quoted minutes as a rough floor, not something you can build a schedule around.
batching attempts by category, based on what tends to screen you out, cuts wasted time faster than switching panels does. If health-condition screeners rarely qualify you, skip them rather than hoping this one’s different.
redemption terms change the real value of a completion after the fact. A points-based reward isn’t worth its face value until you’ve checked the actual exchange rate and whether cash-out options like PayPal, gift cards, or crypto change what you net, which I cover in PayPal vs gift cards vs crypto.

would I do it again

Yes, with a caveat. The log did what I wanted: it turned a vague feeling of “these surveys screen me out constantly” into a number I could look at and act on, and it made obvious which categories of study weren’t worth opening in the first place. That’s a genuinely useful habit if you spend real time on panels regularly, especially outside the US, UK, or a handful of other markets that get the deepest pool of studies. Our note on which panels tend to serve members in Southeast Asia at reasonable volume, panels that serve SEA members, came directly out of noticing category and geography patterns like this while running the log.

The caveat is that maintaining it by hand is tedious, and tedium is exactly what causes a log to quietly stop being accurate around day four. If I were running this for longer than a week, I’d automate the timestamp columns with a simple browser-based time tracker instead of eyeballing the clock, and I’d build the dedupe check into the sheet as a formula instead of relying on memory. The format holds up fine. The manual data entry is the part that needs fixing before you’d trust a month of this data the way you can trust a careful week of it.

If you’re weighing whether a panel is worth your time at all, start with the review breakdowns on the blog, including the Swagbucks earnings review, and then build your own version of this log before you decide anything. A week of your own numbers will tell you more than any review, including this one, ever can.

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 2026-09-15.