We have all been sent on training for a new data analysis platform PaySense. Still, it is always good to add a new tool to my CV. The platform is already used by one of our US acquisitions for its state pay equity reporting, and it looks like we might use the tool globally. The training has been set up directly by the provider, so I log in early. It can take a few minutes to connect to another meeting app and find a way to obscure my background. I had expected the US comp team to be on the call too but no, this is a little private training session just for me, Big Bad Boss and my colleague Lazy Susan. That makes me feel a bit nervous, I cannot see Big Bad Boss paying any interest to the details, and the same is true of my colleague as, for sure, it will involve numbers to which she has an allergy. I had better pay attention then.
Data calculations
It starts by explaining how we would load our data. Our US colleagues have already loaded about 10 fields for several hundred rows of data for the purposes of this training session. The next 15 minutes or so are orientation around the various menus and views. The trainer shows some standard reports including EU pay gap reports. I look at the data fields loaded and ask how they derive the necessary hourly rates which I know to be really problematic. PayNonsense does not provide any calculations regarding hours; you need to upload any data you want to use. That is helpful then. I wonder if anyone picked up on that.
Finally, we get into the so-called analysis, as the tool can run and display pay gaps. The trainer demonstrates this one-click with a flourish. I am finding hard to get excited because I could do that with a pivot table; we do not need a platform for that. The trainer goes on to show how you can even convert all to dollars and show a global pay gap. This has my eyebrows shooting up, and I have to question the logic: doesn’t that mean you will be comparing across very different cost-of-living areas and tax regimes? The trainer has no answer for that. My verdict is that piece of analysis is invalid. Big Bad Boss and I start a side chat in our messaging system. Nonsense, he says. Utter nonsense, I agree, maybe they should call it PayNonsense. In a sudden panic, I double check we are not using the chat option in the meeting. It is one of those easy mistakes to make. A colleague once called out that the presentation was getting boring on a group meeting chat. Three hundred people read it including the speaker. Ouch.
Pay gap analysis
The trainer then demonstrates producing adjusted pay gaps using regression analysis. The system uses various ‘objective’ factors like tenure and job family to adjust gender pay gaps to an equivalent level. There are various statistical tests carried out to show whether there is a demonstrable relationship between the factor and pay. I am rediscovering a whole statistical vocabulary covered in university but never used since, terms like ‘linear multi-variate’ and ‘standard deviation’ are cast about like spells. I see Lazy Susan turn her camera off and I suspect it has got a bit too heavy for her now. I must admit it is heavy for me, and I have done this stuff before.
The upshot is the trainer demonstrates that by ‘adjusting’ the pay gaps for tenure and job family group, like magic, the pay gaps can be significantly reduced. Significantly, but not completely. The trainer goes on to show the budgeting functionality. Once you have found factors that best show a relationship with pay and you have adjusted your pay gaps accordingly, you can use that formula to create a predicted pay for each employee. The system will then use that formula to calculate a cost to bring everyone to a desired pay gap. Hey presto, that nasty pay gap problem has gone away.
Statistical snag
As with any seemingly simple solution, there is a catch. The formula derived from these statistics does not reflect how we actually plan to pay people. We have never and never will intend to pay a percentage of pay for each year of service. That would prevent us from paying more for people who bring relevant experience from outside the organisation. It would also mean that someone sitting in the same job for a very long time, and not necessarily doing it any better, would get paid significantly more than someone joining from the outside. In some cases that might be valid, but certainly not all. Take someone working in a very simple administrative role. After 30 years, an employee might well bring more to that role; they will know their way around the system and have developed internal networks to get things done, but I would argue that would only warrant a modest difference in pay, far less than the percentage derived from an apparent mathematical relationship to pay across the organisation.
I can just about get my head around using statistical nonsense to massage pay gaps to manageable levels, but I stop short of using them to decide how much people should be paid. The trainer shows that the budget is made up of suggested individual adjustments so there is the possibility to review each one and that the main benefit of the cost to fix analysis is for budgeting. I am still dubious. Surely the whole point of budgeting is to predict a future cost as accurately as possible. If we do not want to pay extra for each year of tenure, why would we want to create a budget based on doing so? I look to see what Bid Bad Boss thinks, but he has gone offline. I think I may be the only one to realise it is all smoke and mirrors.
Next time…Candid works on equity share grants.
