December 05, 2016
At Polar, we believe an innovative organization drives innovative products and services. That’s why we constantly analyze how we work—like in examining how we communicate via Slack—and experiment with new business processes and organizational structures.
At the beginning of this year, following a time audit by our CEO showing he spent a surprising 33 percent of his productive hours in internal meetings, we experimented with eliminating internal meetings on Tuesdays and Thursdays—known as “Meeting-Free T-Days”. The motivation was that we should spend more time actually making things rather than meeting. Batching meetings together on the remaining weekdays, or simply reducing meetings altogether, would help provide more blocks of uninterrupted time to focus on creating.
Easier said than done given how ingrained meetings are in business culture.
So, in the several months since we began this Meeting-Free T-Days experiment, did we actually eliminate or at least reduce meetings on Tuesdays and Thursdays? If so, did we shift those would-be meetings to other days of the week or did we reduce the total number of meetings (or, heaven forbid, did we increase the total number)? Finally, were any changes due to a collective effort or due to a few ardent proponents of the new policy?
To address these questions, I gathered calendar data from the Google Calendar API between July 2015 and July 2016—roughly 6 months prior to the beginning of the experiment on January 4th, 2016, to 6 months after—on 24 individuals whose tenure spanned the entire time frame. ¹ (The R code used to gather the data and produce the following analysis can be found here.)
The following plot shows the average number of internal meetings per T-day (Tuesdays and Thursdays) and non-T-day by week. There’s a clear and sizable drop in the average number of meetings per T-day once the Meeting-Free T-Days policy came into effect at the beginning of 2016, and this decrease seems to have persisted in the several months since. Meetings on non-T-days don’t seem to have changed much, however.
After applying an intervention statistical model to the data, which tries to determine whether and how large the impact of a policy change is, the policy was found to lead to a significant 55 percent drop in the average number of T-day meetings, moving from about 10 meetings prior to just under 5 meetings after. ² The plot below shows how the model fits the data, which looks quite reasonable (the gaps indicate when the policy was suspended during our quarterly company-wide conferences).
With regard to non-T-days, there was no statistically significant change in the average number of meetings once the policy came into effect, which the model fit below illustrates. Rather than rescheduling those would-be T-day meetings to Monday, Wednesday or Friday, it seems we just cut them out.
The decrease in T-day meetings in aggregate reflects a broad-based effort by individuals, as displayed by the following plot. The left panel indicates that all (of the top 10) meeting organizers reduced the average number of meetings they initiated on T-days. On non-T-days, while we previously saw that meetings didn’t change much in aggregate, the majority of staff did increase the number meetings they initiated, only to be offset by large decreases from a couple individuals.
While we didn’t completely eliminate meetings on Tuesdays and Thursdays, we significantly reduced them as a result of this experiment, driven by the efforts of many individuals throughout the organization. This doesn’t come as too big of a surprise given how much support staff have professed for the policy. Moreover, we haven’t compensated for the forgone meetings on T-days by scheduling additional meetings on the other days of the week. This suggests we may be more efficient with our existing meeting time on those days or realize many meetings just aren’t that necessary. Either way, I don’t see those meetings creeping back soon.
¹ This was to mitigate the potential bias arising from calendar data being unavailable on former employees.
² The intervention analysis used here follows that developed by Box and Tiao (1975). The T-day series was assessed to contain an AR(1) process while the non-T-day series did not display significant temporal dynamics. An abrupt intervention functional form was found to be the most appropriate model for the policy’s impact.
This brand new guide brings clarity to the publishing and advertising industry by defining the three distinct shades of branded content: editorial sponsorships, custom content and advertiser content – as well as further flavors contained in each shade.DOWNLOAD FREE COPY