ADP – Data aggregation done well

When explaining the people analytics market we tend to classify firms into one of 5 categories based on their dominant approach. Over the last few years a group which has grown in importance has been the Data Aggregators. OrganizationView have advised several firms interested in entering this market.

Data Aggregators

Data Aggregators is what we call firms who, typically by running HR transactional systems for clients, now have large quantities of individual level data. We’re now seeing these firms building products which aggregate and anonymise this industry data and provide it back to clients to provide context to reporting of their own data.

There are two advantages to this compared to traditional benchmarking: first, the unit of analysis is the individual, not the organization. From an analyst perspective this means you’re aggregating – throwing away information – at the last possible moment. For the purchaser of the data it’s therefore possible to get far greater detail. It may also be easier to ensure you’re comparing apples with apples.

Second, the data is likely to be of better quality than self-reported surveys. For many of these data aggregators it can be close to real-time comparisons. What you need to do this well is a lot of data. Ideally you want a decent percentage of the market covered, and that the other companies in the pool are representative of the market as a whole.

ADP

One firm entering this space is ADP. At HR Tech Europe I got to sit down and talk to Tony Marzulli, VP Product Marketing about their data products. Following an interesting meeting I was given a demonstration of their current early beta version of the data product, currently with clients.

ADP have a large client base – 610,000 companies. This means that they have a very good overview of the employment market. For some time they’ve sponsored the ADP National Employment Report (see the methodology developed with Moody’s Analytics here) This covers more than 20 percent of all US private sector employees.

There are three distinct offerings that ADP are bringing to market in their data products:

  • Individual-level comparisons / benchmarks
  • Data-mashups (linking to other data sources for comparisons)
  • Predictive analytics.

Of the 3 the first is obviously the most developed. At present clients can do relatively standard reporting of their own data. The data visualisations are good and some of the best designed that I’ve seen in an HR product. Not once did I see visualisation horrors like a mutant pie chart that some of the HR visualisation world put out. They obviously have people who understand the psychology of data visualisation. There are a good numbers of filters available and the user interaction makes sense. Where it gets interesting is that you can then add context by bringing in aggregated data from the ADPs vast store. By default it will show other firms like you as a comparison – so if you’re a retailer in Texas you’ll see other retail companies in Texas. However if you are competing for talent with another sector you can change that.

 

an example screen. Copyright © 2014 ADP, LLC ALL RIGHTS RESERVED.

These comparisons aren’t just averages, they’re effectively kernel density plots showing the full distribution. This avoids many of the problems of the type that can be demonstrated by Francis Anscombe’s Quartet. Hey, they even have box plots!

There’s a hard balance between visualising for the statistically trained and the lay-person. I think ADP have walked that tightrope quite well.

We didn’t cover the other two aspects in as much detail – they’re obviously at early stages. With the mash-ups one can see that the explosion of datamarkets and opendata means that firms like ADP can easily add numerous public datasets into tools, again to add context. One can imagine it’s possible to add some standardised types of company’s own data this way.

At the demonstration I had a good conversation of the joys of prediction when you have so much data available. We also talked about the need for prediction in reporting as a way of encourage more resilient decision making. It’s obviously early days for ADP but what I’ve seen they’ve got a good team of data scientists and the opportunity, given their data resource, is there.

Where do I see this going?

Sierra-Cedar in their most recent HR Systems Survey talk about combinational solutions being the natural outcome of an increasing maturity of firms’ HR analytics activity. This is a view we share given our experience with clients.

If we think about tools like ADP’s providing the basis for providing insight on workforce data then I can see the addition of aggregated data making a big difference. Certainly it makes the reporting that is necessary as the basis of an HR analytics much more valuable.

What is clear is that to get the most out of these tools you’ll still need people who are data-literate to make sense of the data. If anything, even though the visualisation has been done well, these solutions probably increase the need for HR to learn how to be better consumers of data.

As an analyst, mostly working on problems that are beyond the scope of HR systems, what I would like most is an API to give access to the aggregate data.

Final thoughts

ADP are a quite conservative organization. For such a firm to produce such a tool is remarkable. We’ve worked with clients who have the opportunity to be Data Aggregators helping them think about developing products in this market. There are real challenges in doing this, even to get to the starting line, and I think for a first attempt they’ve produced a great product.

Most importantly, I think looking at ADP is a good example of how we see the HR analytics market going. I’d certainly recommend our clients consider it as part of a combinational approach.