Why we invested in datascalehr

The messiest, most intractable integration problem meets AI and experience

(announcement here)

We are not afraid to say we know a bit about payroll at Acadian Ventures. We have invested in 4 companies who build payrolls in different parts of the world. David, our partner, was the Workday CTO, and many of our LPs have built or run payroll companies. I was an NED on the board of Immedis. I ran the payroll product line at SAP and it pays over 100 million people when I last looked. I’ve felt the pain of a customer missing payroll. Heck, I’ve even read the project minutes of the very first payroll software project.

Many people in software find payroll boring. I think it is fascinating. It is where tax, employment law, hard work, social norms all intersect. How a society works is instantiated in its payroll code. It is, as Larry Lessig would put it, code as law. Changes in technology empower new forms of reward and tax.

Agile got started on a payroll project, btw.

At one level it seems like a solved problem, after all everyone gets paid somehow. But it isn’t.

I’ve been sceptical of well funded HR or EOR companies suddenly announcing they have native payrolls in 90 countries. I ask them about the German ITSG, and suddenly it’s only 89 countries. Payroll doesn’t do MVP, there are no edge cases you can cut.

I’m not going to write about the complexities of building payroll engines, and gross to net calculation routines, or Belgium’s fiendishly complex Bilan Social. I’ll spare you that today.

One of the most challenging elements of payroll is getting the right data into payroll in the first place. Payrolls need perfect data to work.

Unlike most interfaces, the data timing and formats change over time, every country does it differently, and the systems producing the input data don’t have the same level of data quality as the payroll needs. The next biggest challenge is getting those payroll results back in a way in which they can be understood and reconciled - did we actually pay the right people the right amounts?

Global HR vendors don’t know what the Latvian payroll needs to work, so isn’t surprising the interface ends up doing a lot of heavy lifting. For the Latvian payroll provider, interfacing to SAP, Oracle, Workday or yet another payroll aggregator is the last thing they are wanting to build, as most of their business is supporting local companies. Mapping, building, testing and maintaining interfaces between HR, payrolls, time and attendance, finance has to date involved weeks or even months of tedious and arduous work. Reconciling pay data between HR and multiple country payrolls eats up a large part of payroll outsourcers time and money. Same too for Employer of Record providers.

To complicate things further, every HR system is implemented slightly differently. Every customer has its own mix of data sources, formats and requirements. Same for every payroll engine and every implementation of every engine.

This was a challenge the very first payroll project identified (Lyons tea room in UK in the 1950s).

One must not, however, give the impression that to put a payroll onto a computer is an easy matter: it is not. The amount of detail to be considered is very large and the number of exceptional circumstances to be dealt with is beyond belief except by someone who has been concerned with a payroll application.

The methods of keeping the records for time worked and other pay entitlements differ very much from one organisation to another, often for no obvious reason. Ideas differ also on what should appear on an employee's pay slip, and the way the information should appear on the payslip is very much a matter of taste. (Thompson, 1958)

More on that here.

Fixing this problem would change the economics of the payroll industry.

Sometimes we invest in wicked smart young people. They bring energy and blast through doors attitude, and often they apply new technologies and ideas to old problems and delight us.

But for solving deep payroll integrations, we invested in a first time founder with nearly 30 years of deep, enterprise, global payroll experience, and who has spent the last few years learning about AI in depth. I’ve known Jerome Gouvernel, the CEO and co-founder, since the early days of the SAP / ADP partnership. If he were on the BBC TV show Mastermind, his special subject would be payroll engines. But since leaving the world of ADP, Jerome has, together with his co-founders, immersed himself in the world of deterministic and probabilistic AI, and datascalehr has completely turned how to build these integrations inside out.

The team has built a mighty API- first product. It can connect any source system to any backend payroll. Interfaces that took weeks or months to build, can now be built in an afternoon, data migrations that took months, now take days. You could describe it as an agent in today’s parlance. When Jerome walked us through his idea, we were intrigued. Once they were able to prove that it worked for early customers, we knew we just had to invest.

3,2 billion people get paid every day. Taking massive inefficiencies out of that process is a huge market, and datascalehr is just getting started. We think they will build the industry’s connectivity standard and be the central repository of payroll rules. At some point they will know how every data input gives rise to every payroll output, then a whole lot of new opportunities open up.

If you have a payroll business, or you run an EOR you really ought to talk to datascalehr. We can’t wait to help them grow this into a massive endeavour.

As I usually do, I’ll end with a tune. Although this song is about postal delivery, about 25 years ago I was testing an SAP payroll interface on a project (I was a consultant, but not a very good one) and it kept coming back with the message no such code. Every time I hear this song, I think fondly of interfaces.