Did you know that nearly half of all job-seekers who use our chatbots use them outside of recruiter office-hours?
Last year, we made this observation for one of our chatbots. Recently, we designed a bigger experiment, validating this finding.
Look at 5 of the most high-volume chatbots, some with hundreds of unique, multi-turn conversations* per day, and some that had been live for almost a year. Each of the 5 employers are different in size and industry, and use different messaging platforms (Messenger vs Smooch). So it was surprising that the relative amount of the evening and weekend conversations was relatively consistent.
Based on insight on when recruiters are responding to messages, by interacting with our B2B product, we changed the definition of working hours, from 9:30 to 5, to 8 to 6. This is a longer window, so it is even more interesting that, still, almost half of all traffic comes from outside of working hours.
Before doing this experiment, I was expecting more of the conversation to take place on the weekend; from my own experience, that is when I’ve chosen to look and apply for jobs! However, it turned out that quite a lot of the job-seeker traffic happens during the week. According to almost every major productivity study since 1987, Tuesday is the busiest day of the week (Workopolis), and the most common day of the week to apply for jobs (Workopolis). However, it’s not the most successful day to apply: In fact, the probability of getting to the next stage of the process if you apply on Tuesday drops by 10% relative to Monday!
When I analyzed the traffic by day of week, I was shocked.
There can be significant variation from one day to the another; for example, for one of the highest-volume chatbots, there is a significant difference** in the number of conversations (with unique users) between Thursday and Friday. Most of the time, though, there are other factors that influence the traffic more: for example, seasonal job offerings, marketing campaigns by the employer, and so on.
In the future, I would like to explore more deeply the different job-seeker scenarios, like, for example, seasonal job offerings, because these hidden factors are the ones that no one would tell us, but they actually influence what information we could get from the data. In the meantime, we have validated a vital previous observation:
Almost half of all traffic from job seekers comes outside of working hours - independent of messaging platform, or employer size or industry.
So, if you’re wondering, is a chatbot right for me - ask, do I want to be able to answer the right candidate at the right time… including those 45% when I’m not at the office, but they are approaching me anyway?
* All analysis reflects unique conversations with unique users, rather than messages. In the Messenger platform, this means a conversation can last days; in Smooch, it lasts only as long as the user is on the page
** (t=2.44 p=0.04) In paired t-test comparing relative traffic on Thu vs Fri. The traffic was normalized by overall conversations that week. For the other four chatbots, there was too much variation from week to week for statistical significance.