1. Behaviourally-informed emails prompting employees to provide their sociodemographic data will increase the rate of sociodemographic data provision relative to the status quo email and a passive control group receiving no emails.

2. Relative to the status quo email, behaviourally-informed emails will increase email engagement, as defined by email open and click-through rates.

3. Both status quo and behaviourally-informed emails will be more likely to increase sociodemographic data provision when sent closer to the dates on which employees confirm their monthly attendance within the same HR platform (usually at the end of a month or the beginning of the next month). A parabolic relation between the email sent date and sociodemographic data provision is hypothesized, whereby sent dates closer to the transition between months should result in higher levels of sociodemographic data provision (inverted ‘U’ shape).

4. No a priori hypotheses are defined regarding the effects of sent date on email engagement, as defined by email open and click-through rates. Potential effects will be explored as part of the data analysis.

Five conditions in a randomized controlled trial. Participants will be randomly assigned to either the passive control condition (receiving no email during the trial) or one of four email conditions: (1) status quo (text adapted from initial promotional material for the launch of the sociodemographic initiative), (2) simplified information, (3) implementation intention, and (4) active choice. Emails will be sent out evenly across conditions over a 15-day trial period, starting approximately a week before the end of the month and ending approximately a week after the beginning of the following month. Sociodemographic data provision and email engagement will be compared across conditions and sent dates.

Primary outcome
• Entry of sociodemographic data into the HR platform. To preserve the privacy of employees, the data will be reported to the research team as aggregate counts per condition and sent date.

Secondary outcomes:
• Open rate of email message as recorded by the email platform.
• Click-through rate of email message as recorded by the email platform.

Logistic regression analyses will be conducted to analyze differences in the effects of conditions and sent dates on sociodemographic data provision, email open rates, and click-through rates. Note that only aggregate data will be reported to the research team to preserve employees’ privacy. The aggregate data will be in the form of frequency tables showing rates of sociodemographic data provision by experimental condition and sent date. However, because only categorical and ordinal variables are being examined in the analyses, it will be possible to reconstitute anonymized individual-level data by weighting the variable levels by observed frequencies. As such, both chi-square and logistic regression models may be used to analyze the data.

To compare the effects of different email conditions on sociodemographic data provision and email engagement, the following planned orthogonal contrasts will be used:

Contrast 1: Passive control vs. all 4 email conditions.

Contrast 2: Status quo email condition vs. all 3 behaviourally-informed email conditions.

Contrast 3: Active choice email condition vs. simplified information & implementation intention email conditions.

Contrast 4: Simplified information email condition vs. implementation intention email condition.

To test for an effect of email sent date, a logistic regression will be run using sent date as an ordinal variable to predict the likelihood of sociodemographic data provision. The hypothesized parabolic relationship between sent date and outcome likelihood will be tested by determining whether a logistic regression model including both x_date and x_date squared as predictors explains significantly more variance than a model with only x_date as a predictor.

The total sample size for this RCT is approximately 70,000 with 15,000 employees in the email conditions and 55,000 in the passive control condition.

Data of employees who leave the OPS during the trial period will be excluded from the analyses. Similarly, data of new employees joining the OPS during the trial period will be excluded.