Case Study: Chat Usage & Performance
Senior Product Analyst @ Docplanner Tech
Executive summary
Docplanner’s Chat is a high-impact but still underutilized feature.
While ≈60–65% of doctors log into the platform on a weekly basis, fewer than 30% actively seek to use it. Those who chose to do so perform visibly better: doctors using both the core platform and Chat achieve 2× higher median bookings and 2.8× higher average bookings when compared to platform-only users.
The Chat functions as a revenue amplifier, not just a mere communication tool. It can be used by doctors to broaden their patient portfolio whenever the opportunity arises: a quality answer to a simple patient query can result in a long-lasting business relationship.
The biggest opportunity lies in activating the large “Platform Only” segment [those users who use the core platform but are not yet using Chat].
The strategic focus around this feature should be to create ways to drive Chat activation and embed it into weekly workflows of doctors, as well as making it tool that patients use in a routinely manner. There’s room to improve in either side, for both doctors and patients.
- Improving its onboarding and awareness:both doctors and patients can benefit of an extra nudge when navigating the app. A simple call-to-action can go a long way in in breaking the ice, reducing the friction in the first step. Ideas:
patients can ask using structured prompts about symptoms or therapies
expose the response-time (“typically answers within 2 hours”) to create a bigger incentive to reach out to a top responder;
- Stimulating a regular usage, leveraging the preparation of appointments but also eventual follow-up questions and therapy instructions; this can be particularly relevant for the most relevant specializations in the platform such as psychology;
- Reducing patient unread rates (currently ≈13% of messages): this can be achieved by rewarding the most assiduous doctors through a reputation system that encourages further engagement (e.g., exposing the overall response rate).
Exploratory analysis
The volume of doctors using Docplanner software in the reference period is comparable between both countries (Brazil and Mexico). Mexico’s “golden cohort” (i.e., spike on the number of doctors subscribing to the software) indicate that the service became available at a later date for this country. Most doctors offer medical practice, with paramedical representing a very small market share.
The active doctors offer services across almost 60 listed specializations. In the above chart we can see the most popular (in terms of number of available doctors). Those specializations where the supply is rather small are grouped at the bottom under ‘Others’ (these 41 specializations account for less than 20% of all doctors subscribed to Docplanner).
The Docplanner Chat is one of the many services provided by Docplanner, allowing doctors and patients to exchange messages (questions, answers, booking reminders, file sharing, etc).
A few observations:
This feature seems to be way more popular in Brazil, accounting for ≈58% of all messages (against 42% in Mexico); although this difference is sizable, it should be no reason for concerns since we previously comfirmed Brazil has also more doctors and patients (thus more messages will be generated).
Doctors seem to be more active in Chat than patients, sending 56-61% messages (patients only send 39-44%); country differences are once again probably related to the difference in the number of doctors and patients.
As expected most messages are texts - only a minority of the Chat interactions pertains to files sharing (e.g., medical records, prescriptions, etc).
The trend in total Chat volume seems to be stable with small uplift in read messages (from both doctors and patients) - this could be a seasonal effect, with higher demand for medical services leading to more Chat usage.
The amount of unread messages per week is about ≈12-13% of all messages sent: from those, the overwhelming majority is sent by doctors. I assume a large portion of those messages are appointment reminders sent by doctors and that go unnoticed (or ignored) by patients.
It’s reasonable to assume that this feature might help doctors engaging with patients, driving more successful bookings and retaining recurrent clientele. But in order to measure the value of this feature for clients (doctors), we first need to understand its overall discoverability and usage.
Note: I’ll perform this analysis combining both countries since up until now there seems to be no reason to be concerned with regional segmentation.
Chat discoverability
From all the subscribed doctors at least 59% ever receives a message and around 57% ever reads a message (about ≈2% doctors have received messages and never read them).
Around 56% of the subscribed doctors sent at least one message in this period.
In sum, it’s fair to say that about 56-57% doctors had at least one interaction with the Chat (they know of its existence and have tried it, either reading or sending messages).
Chat usage
Across the first 10 weeks of 2021, the share of doctors subscribed to Docplanner who log-in at least once per week varies between 60-65%. However, the share of doctors who’s actively using the chat (i.e., sending and/or reading received messages) on a weekly basis is <30%.
The above chart illustrates the distribution of weekly bookings per doctor. Observations for each doctor-week pair are classified by activity tier as follows:
Inactive: the doctor did not log into Docplanner during that week;
Active (Platform only): the doctor logged into the Docplanner platform during that week but did not read/send any patient messages.
Active (Platform + Chat): the doctor logged into Docplanner and was active in chat (read/sent messages to patients) during that week.
Doctors active in both platform and chat generate 4.45 bookings per week on average (and 2 median weekly bookings): significantly higher than Platform Only doctors, who booked an average 1.61 weekly appointments (and median = 1).
The four correlation coefficients quantify the linear relationship between chat metrics and weekly bookings:
Active days vs Bookings (r = 0.23) is the weakest link among all studied correlations; even though it shows a positive relationship, it suggests that simply being active on the platform does not guarantee additional bookings;
Number of messages read/sent vs Bookings (r = 0.33-0.34) signals a stronger relationship between doctor engagement and a higher rate of successful bookings;
Unique patients messaged vs Bookings (r = 0.39) is the strongest relationship in the pack, supporting the hypothesis that widening the chat communication can lead to higher patient conversion (and thus more bookings).
A solid retention rate can be a great symptom of engagement and customer satisfaction. By providing a seamless user experience, Docplanner can facilitate patient conversion (and retention), which in turn should improve doctors’ satisfaction and their own retention. Let’s first take a rough look at that metric.
Doctors’ retention fluctuates between 90-60% within the first 4 weeks of data, showing strong stickiness with the platform;
Disclaimer: this is a proxy metric; I considered doctor’s first active week from data as a proxy for subscription; while this is not a true cohort start and it creates an upward bias for doctors who subscribed long time ago, it should be fairly accurate forlate adopters;
The lack of older data makes some cohorts quite small (samples containing between 40-100 records leading to wide 95% CIs), reflecting uncertainty around the true value of doctor’s retention.
Next it’s time to compare how doctors within each activity tier perform in terms of retention rate.
Retention is clearly higher in the segment of doctors who use chat at least once than for those who only use the core platform. This fits the general pattern that users engaging with more interactive and social features tend to stick around longer.
Note: with this data we can only establish correlation, not causation. Chat usage and retention reinforce each other in a self-feeding loop, where more doctors who are willing to engage with their patients are both more likely to try the Chat and more likely to remain customers. A proper experimental design would be necessary to establish causation.