How Spicer Research works.

01Where the data comes from

Spicer is a mobile app for couples. Inside it, partners answer questions about intimacy to discover ideas they both want to try. The core mechanic is a private matching engine: each partner answers a question on their own, and only answers that both partners respond to positively become shared "matches." Anything one partner declines stays hidden from the other.

Because millions of these questions have been answered over the years, the aggregate of those answers forms a large body of signal about what couples want — at the level of populations, never individuals. Spicer Research mines that aggregate. We do not collect new survey data for these reports; we report patterns that already exist in how people have used the app.

02How answers are collected

Three things about the collection method shape everything downstream, and we state them plainly because they affect how our numbers should be read.

Answers are private and independent. Each partner answers a question without seeing how the other answered. This matters: it means a couple-level agreement is two people independently arriving at the same answer, not one partner influencing the other.

There are three options, not two. Every question offers Yes, No, or Maybe. The third option is deliberate — it lets people register openness or uncertainty without committing to a yes. We preserve all three throughout our analysis and never collapse maybe into yes.

Answering is voluntary. People answer the questions they choose to engage with. This is a source of self-selection we return to in the limitations.

03Two levels of analysis

Our reports contain two distinct kinds of statistic, and we always label which is which.

An individual-level statistic counts single answers — for example, the share of all people who answered a question who said yes. These samples are large, often well over a million answers per question.

A couple-level statistic is stricter. It counts only couples where both partners answered the same question independently, and it describes how the two answers combined. Couple-level samples are necessarily smaller than individual-level samples, because they require two linked people to have both answered. When we say something like "60% of couples," we mean couples in this stricter sense — both partners answered, independently.

04The couple-response taxonomy

When both partners answer a Yes/No/Maybe question, there are exactly six possible combinations. We report all six. They are exhaustive and mutually exclusive: every couple in a sample falls into exactly one.

We publish the full set rather than a collapsed summary, so a reader can always see exactly how a headline figure was built and can re-derive any grouping themselves.

05Groupings we use

For readability, our reports often describe three groupings built from the six categories above. We define them precisely here so the language is unambiguous across every report.

The both no category sits outside these three groupings and is always reported as its own outcome.

One point deserves emphasis, because it is easy to get wrong and we hold ourselves to it strictly:

A maybe is not a yes. "No explicit no" describes the absence of a refusal — not the presence of consent.

When we report that some share of couples had "no explicit no," we are describing the absence of a recorded refusal, not agreement. Maybe signals openness or uncertainty. We never present it as equivalent to yes, and we never frame a result in a way that implies consent where the data only shows the absence of objection.

06How we report numbers

A few conventions govern how figures appear on the page, all aimed at one thing: a reader who does the arithmetic themselves should never find a contradiction.

• One number, everywhere. A given sample size or percentage appears in exactly one form across a whole report — the same figure in the headline, the charts, the methodology box, and the structured data. We do not mix exact and rounded versions of the same number.
• Percentages to one decimal. Category percentages are rounded to one decimal place. Because of rounding, a set of categories may sum to between 99.9% and 100.1%; we note this rather than adjusting a true value to force a round total.
• Gaps reconcile with what's shown. When a report shows two percentages and the gap between them, the gap is the difference of the displayed (rounded) figures — so the number a reader calculates from the page matches the number we state.
• Counts are exact. Where we show a count alongside a percentage, the count is the actual figure, not an approximation, so the two always correspond.

07Statistical vs. practical significance

Our samples are large — often hundreds of thousands of answers per group. At that scale, almost any difference that is not exactly zero is statistically significant, because statistical significance largely reflects sample size. That makes the test of significance close to meaningless on its own here.

So we report effect sizes and direction, not p-values, and we distinguish two ideas the word "significant" tends to blur:

• Statistically significant — the difference is unlikely to be an artifact of sampling. At our sample sizes this is almost always true and rarely interesting on its own.
• Practically meaningful — the difference is large enough to matter for how a real couple might think or act. This is the bar we actually care about.

When we describe a small gap as negligible, we mean practically negligible — not that it is statistically undetectable. We try to keep the line between "detectable" and "meaningful" visible to the reader rather than hiding behind a significance test.

08Privacy & data ethics

Privacy is a design principle of the Spicer app and a non-negotiable rule of Spicer Research. The whole product rests on people being able to answer honestly and privately; our research cannot undermine that.

These rules are not a footnote to the brand; privacy is the brand. A reader, a journalist, or a couple deciding whether to trust Spicer should be able to verify that nothing here could ever point back to a real person.

09Standing limitations

The following limitations apply to every Spicer report. Individual reports add limitations specific to their topic, but these are always true and we state them every time.

10How to cite Spicer Research

Spicer reports are free to cite and reference. We ask only that citations name the specific report and its sample size, and link to the report page so readers can see its full methodology and limitations. Each report carries its own sample size and snapshot date; use those, not a generic figure.

For data requests, embargoed access to upcoming reports, or questions about a specific figure, the press contact is on the press page.