Survey logic and the screener

A survey decides who takes part in its Screener box, then sends people different ways through the Study box using logic cards: If / then, Quota, Random split, Terminate and End, plus Show only if on any single question. When you buy responses, screen-outs and quotas on the panel's own demographic questions turn into panel targeting by themselves, so people who don't fit are never sent at all.

The Screener and Study boxes

A new survey opens with two boxes on the map: the Screener on top and the Study below. The Add menu on the left follows the box you have selected, with a Screener / Study toggle at the top. Click an item to add it to the selected box, or drag it into either box. To move a card across later, drag it into the other box or use the move button in the right-hand panel. A question, block or loop moves on its own (Move to the Study puts it first in the Study, Move to the Screener last in the Screener). If the move would break the flow, such as a question that screens people out going into the Study, it is refused and the reason shows. A logic card can't be lifted out of its lines, so its button reads Put in the Screener (with everything before it) or Take out of the Screener (and what follows it there).

Two rules keep the shape sound, and the Map checks tell you if either is broken:

Only the screener can screen people out. A screen-out in the Study is an error. (Ending people as quota full or on quality is fine anywhere.)
The survey starts in the screener, and nothing after it leads back in.

Inside the screener, each card carries a badge. Screens out means an answer can turn someone away, which lowers incidence (the share of people who qualify) and so sets the panel price. Profiling means it is asked in the screener but nobody is turned away there, so it doesn't lower incidence.

Demographics

With the Screener selected, Who can take part lists the panel's standard demographics (age, gender, region and so on), each linked to the panel. More panel profiles holds the rest of the panel's library, searchable by name or wording ("pets", "job title"). With the Study selected, the same questions appear under Demographics (profiling): asked in the study, they describe people without screening anyone out.

Logic cards

If / then checks its conditions from the top and sends each person down the first one that matches. Anyone who matches none follows Otherwise. In the right-hand panel, give it a Name, click + Add a condition (up to 50), give each condition a name of its own if you like (it names that way out under Where people go), and remove one with its bin button. Each condition gets its own connector to wire up.
Show only if (on any question, in the right-hand panel): only people who match see the question; everyone else skips it and carries on. Click + Show only to some people to add a rule, or Show to everyone instead to remove it. It can only read questions that come before it. On a page, a Show only if that reads another question on the same page updates as people answer.
Random split places each person in one cell at random, in proportion to its weight, and always the same cell if they come back. Name each cell and set its Weight; the panel shows the share each weight works out to. + Add a cell adds another, up to 20, and a split always keeps at least one.
Terminate ends the survey for that person. Its Reason is Screened out (doesn't qualify), Quota full, Failed a quality check or Other, and is recorded with the respondent and sent to the panel.
End marks the survey complete.

A Terminate or End card's Name is for you; respondents see a standard thank-you screen.

You can also screen out straight from an answer: connect that answer's line to a Terminate card. A No to a consent question isn't counted as screening, since it isn't about who qualifies.

Conditions

A condition is one or more rules joined by All of these or Any of these, up to 50 rules. Click Add a rule (then And another rule) to add one, and the cross beside a rule to remove it. Each rule picks an earlier question and a comparison: is or is not an answer (on a grid, a row and a scale point), a number comparison (=, ≠, ≥, >, ≤, <) on rating, slider and number questions, or was answered / was skipped. A condition with groups inside groups (which only an import can make) can't be edited in the panel for now; import the questionnaire again to change it.

Quotas

A Quota card caps how many of each group complete. Each cell has a name, a share in percent and a condition, and the cells must add up to 100%. Click + Add a cell to add one (up to 50); the Cells total line under them turns red until they reach 100%. Each person fills the first cell that matches them. The right-hand panel shows how many completes each share works out to at your target.

Their cell still has room: they carry on.
Their cell is full: they follow Cell full, or are ended as quota full if nothing is connected.
They fit no cell: they follow Fits no cell, or are screened out if nothing is connected. Because that is a screen-out, such a quota must sit in the screener.

A place is held from the moment someone reaches the quota. It is given back if they are later screened out or their response is rejected, and after four hours of inactivity if they walk away.

How the screener becomes panel targeting

When you buy responses, we work out from your map what to ask the panel for. A screen-out or quota on one of the panel's linked demographics becomes targeting: the panel only sends people who fit, so nobody is sent, paid for and then turned away. The question is still asked, so the answer is in your data.

This happens when the question is asked of everyone (not in a loop, no Show only if) and is screened by answers connected straight to a screen-out, or by an If / then straight after it that reads only that question. A quota becomes a panel quota when everyone passes it, each cell reads one linked pick-one question, and nothing is connected to Cell full or Fits no cell.

Anything else stays in the survey: people are sent and then screened out by your questions, which lowers incidence and raises the price per response. An answer the panel can't target, such as Prefer not to say, also keeps a screen in the survey. The Audience tab lists exactly what the panel is asked for and why anything stays (see launching a survey).

Quality checks you set

Attention check (in the Study's Quality checks menu): a question asking for one stated answer. Anyone who gives another is rejected.
Made-up options: add an answer that doesn't exist, such as a brand nobody has heard of, and tick it under Made-up options. Anyone who picks it is rejected.
Believable range and Must agree with: a number outside the range, or two answers that can't both be true, hold the person for review.

Every other check runs by itself; see respondents and quality checks.