Risk IQ™ is GymOS’s member attrition early-warning system. Using intelligent behaviour analysis, and powered by over seven years’ worth of behavioural data, Risk IQ is designed to provide gym owners and coaches with a “heads up” of those members who might be showing behavioural signs of disengaging with their membership.
The fundamentals
Risk IQ is designed to work for structured memberships with a weekly recurring offering. This is typically the case for memberships offered by GymOS clients, where a typical membership may offer 2, 3, 4 or whatever number of sessions per week.
Risk IQ does not provide alerts for PAYG members. Additionally, whilst they may be of value, alerts raised for members on non-structured memberships (E.g. open-gym) are likely to be more erratic – and thus can be easily filtered from the alert results.
Every day, Risk IQ will review, analyse and record a summary for each member (with a membership), nearly two dozen member behaviours. Using Risk IQ’s proprietary risk calculation algorithms, we then assess and determine members showing proven signs of disengagement. This is far more complex than x < y, but the result is that each member is given a daily “risk score”, and the attributing “concerns” are recorded. Risk scores are not linear, and certain behaviours or extremes of behaviour will see an increased score.
The chances are that those members with low scores (say 1-2), whilst showing signs that might be a concern, are likely not at major risk. However, if the score increases over time, greater heed should be taken.
The RiskIQ functionality provides a wealth of options to help gyms filter and evaluate their risk alerts.
Historically, member attrition warnings (including those in GymOS) have been predominantly focused on the last time the member was seen. Risk IQ goes to a whole new level and takes a radically more complex approach. Behaviours such as a member with good attendance suddenly cancelling all future sessions will now be flagged, whereas previously, it could have been a week or more before the issue with this member might have been noticed. Similarly, if a member was gradually dropping from 3 sessions per week, to 2, to 1 – this would likely have been missed entirely.
Being behaviour-based, Risk IQ is able to ‘iron out’ many of the discrepancies and variances between members. For example, some members have horrendously high cancellation rates. However, this is their ‘norm’. If they deviate from that norm (in a negative way), then Risk IQ will jump in and may raise an alert.
Our analysis has determined that as many as 20% of members may be at risk of attrition at any one time. Whilst this can result in a fair number of alerts on a daily basis, if these are managed effectively and efficiently, the gains massively outweigh the few minutes of review time required.
Risk IQ can be misled be natural behaviour changes caused by things like holidays, sickness, or membership pauses. To mitigate this, GymOS provides the functions to recorded scheduled absences and to manage membership pauses. These are then factored into the Risk IQ algorithms, to help reduce “false positives”.
Why is Risk IQ in BETA?
We have trialled and analysed Risk IQ across several months prior to launching it to GymOS clients. During this time, our analysis has shown that approximately 74% of the members flagged as concerns, went on to cancel within a 2-month period.
However, the Risk IQ algorithms are complex and they shall continue to be reviewed and tuned by the Quoox development and data teams over a period of time. These shall lead to further refinements and tweaks, as additional knowledge and “outcomes” become known.
What can I do with a Risk IQ alert?
Having reviewed a given Risk IQ alert, you have a number of options.
You can choose to mark the record as “not a concern”, or “might be an issue”. In either case you can optionally add comments to explain your conclusion.
For those you do determine to be a problem, you can optionally trigger a Risk IQ workflow. Each workflow is a GymOS trigger, and you can configure as many as you wish. For example, you might create different workflows for “of minor concern”, “of moderate concern”, and “of clear risk of leaving”. The choices and actions are entirely up to you, and you can use any of the actions available with member triggers.
What is the algorithm, and can I change it?
The algorithm is complex and the proprietary intellectual property of Quoox. The details or core aspects of it are not something we shall be revealing – although some elements will become self-evident from the flags that get raised.
Our studies of 7+ years of data showed that, for structured weekly recurring membership packages, member behaviour was the consistent across all clients. The algorithm does not need to be tailored on a per-client basis. However, gyms are able to control the alerts that they see, and can choose how to process them as they wish. Alerts for given members can also be suppressed for a period of time, or the member excluded on a permanent basis.
False positives and missed losses
Naturally, if a member’s behaviour changes (for whatever reason), this may result in a “false positive”. Risk IQ is “self righting”, in that its analysis is performed on a rolling basis. Therefore, if a member were to drop from 3 sessions per week to 2 sessions per week (and this was a known membership downgrade), Risk IQ would likely initially flag an alert or two, but these would ‘self correct’ over time.
Extreme behaviours may result in either false positives or missed losses. The Quoox data team are able to see these, and the data shall be used to further refine the algorithms (where possible) over time.
Why has x been flagged, or why was y missed?
The reasons a member has been flagged are listed on the Risk IQ anomaly alert.
The Quoox support team are unable to practically answer individual specific enquiries about why a given member might have been flagged (beyond the reasons given), or why a member that has subsequently cancelled might have been missed. The algorithms and volume of data analysed are simply too complex to be reviewed manually on a “per case” basis.
Risk IQ will not be 100% fool proof, but the concept is that “a member retained that might previously have been lost is a win”.
What if I don’t like the alerts that are raised, or if I don’t find them of value?
In the unlikelihood of you not finding any value from the data raised, or if you are a facility that does not fit the “structured recurring package model”, simply ignore the alerts.
Risk IQ alerts are stored in a separate section of Gym OS and, if you don’t want to see them, you can simply choose not to view them!
How can I see the historical Risk IQ alerts for a member?
You can see the recent historical Risk IQ alerts for a member by clicking on the filter button on one of their past alerts.
Alternatively, if the member has a current membership (and is not excluded from Risk IQ alerts), click the “View Risk IQ alerts” option in the Actions dropdown on the member record. This option will only display if the member has a current membership and is not excluded from alerts.