Collecting gender information

Gender information is collected by a variety of groups for a variety of reasons. Here are some things to question, pitfalls to avoid, and alternatives when asking for gender information.

Why ask?

What is the purpose of gathering the information? What will be done with it? When possible, share the reason with the group who are being asked - this will increase the chance of them giving accurate and relevant information when disclosing their gender involves a personal risk.

Good reasons

Knowing the starting point to measure the effectiveness of gender diversity initiatives
Caring about your people and using their information responsibly

Sufficient reasons

As part of legally required reporting
Validating government-issued identification (why?)
Gender-segregated bathroom proportions - however, individual private complete bathrooms need not be gender specific, are inclusive to every gender identity, and can be equipped to support a variety of needs (wheelchair transfer, baby diapers, menstrual supplies, audio privacy, extra space).
Best guess medical preset ranges - however, any medical software should be able to cope with things like hormone replacement therapy, a man who requires regular pap smears for his cervix, etc.

Bad reasons

Pronoun choice - ask the user for what pronoun set you should use instead (note: framing the user's pronouns as "preferred" can be dismissive of their importance to the user; framing the question as a global question such as "what pronouns do you use" gets complicated for people who use multiple sets or who aren't public everywhere about a transition; "what pronouns should we use for you" narrows the scope to a single organization/website which is more manageable)
Honorific choice - ask the user what honorific you should use instead, and include non-binary choices such as Dr., Rev., and Mx.
Mandatory display of gender to other users - display of gender to other users should be strictly opt-in
Placeholder avatar choice - give the user options (including non-gendered), or generate a non-gendered unique placeholder
Avatar body type - human bodies come in all sorts of shapes; let the user pick what suits them best
Avatar clothing choices - in an ideal world, all avatars would have access to all styles of clothing
Blue/pink swag count - don't default to blue and pink! Ask preferred color directly, or ask a representative small group for preferences and extrapolate
T-shirt size/style - ask for preferences after giving size and cut information, or ask a representative small group for preferences and extrapolate
Small/large non-shirt swag count - ask preferred size directly, or ask a representative small group for preferences and extrapolate
To support stereotype-driven marketing - form a more nuanced picture of the audience
Seated vs. standing bathroom facilities - don't make assumptions about people's bathroom needs from gender
Changing table vs. no changing table - don't make assumptions about people's childcare needs from gender
Anatomical assumptions - this is very likely to be too personal for anything other than medical reasons
Unbreakable medical assumptions - bodies get complicated, and medical software should break less often when confronted with reality. If you think there are only two medical sexes, start by looking up about intersex people.

Two isn't enough

Avoid unnecessary binary male/female gender questions.

At minimum, include some form of other/none/decline to disclose if legally possible (some government forms, particularly US government, have a mandatory binary male/female choice).
The options for "other" and "decline to disclose" should be separate.
People who have been harassed based on their gender may feel unsafe disclosing their gender unnecessarily.
Poor design practices (such as unnecessarily gendered placeholder avatars) can leak ostensibly private gender information.
Not all gender systems are limited to two choices.

A mandatory male/female gender question may cause distress to people who don't fit those categories neatly (such as people who are intersex, two-spirit, genderqueer, questioning their gender identity, in the middle of transition, have lost reproductive organs for medical reasons, and more). Items which are subject to research ethics review should assess this risk before allowing a mandatory male/female binary choice as part of the study.

Instructions to "pick whichever fits best" and similar are unlikely to be helpful. In a highly gendered society, this is not an unfamiliar question, and does not need this instruction. In that context, this is patronizing rather than helpful. A helpful response would provide more suitable choices or explain why the choice is necessary. ("Because the advertisers don't think other genders exist" may be honest, but not particularly useful.)

The most nuanced method of asking for gender is write-in. This is also most difficult to automatically sort, if only because of different terms for the same thing, or typographical errors. A reasonable compromise between nuance and ease of use is a gender collection form with the following four options:

Male
Female
Other: ________
Decline to state

Allowing a write-in option for "other" acknowledges that not all gender-diverse people are the same while limiting the complexity of the form. Some groups maintaining longitudinal gender statistics choose to include any option which had above a threshold percentage of response as an independent choice in subsequent surveys.

Government identification

If your purpose for collecting gender is to validate against government-issued identification, state this, but know that there are some people whose appearance may have changed significantly since their identification was last issued.

Some reasons why people might not have identity documents with a gender marker matching their gender:

Expense
Large amount of paperwork/hassle to change
Still in the process of gender exploration
Recent changes
Their government does not issue non-binary gendered identification

Some governments do issue non-binary gendered identity documents, or identity documents which lack gender markers. This means that when there is software which forces a binary male/female choice, people with non-binary gender markers on their identity documents will be forced to make a choice that is doubly incorrect. Be aware of the current status of gender on identity documents in every country your system will work with.