Why Josh Bersin is Wrong About Embedded Analytics
This is a "Gloves Off Friday" post.
"We walk the same path, but go on different shoes. Live in the same building, but we go different views." - Drake, "Right Above It"
Recently I heard Josh Bersin say that the future of our space (People Analytics) could be in embedded analytics.
He didn't say this, however - if you read between the lines - if you buy this, the hard work you put into hiring analysts over the last few years may be for nothing and if you are an analyst maybe so are you!
what Josh Bersin is saying to you is that operational systems are going to eat your analytical lunch, embed it in their right bicep, then flex it right in front of you.
This is a way of saying the work of People Analytics technology like One Model may be made irrelevant by HR’s operational HR systems (ERP’s, HRIS’s, Applicant Tracking Systems, etc.) as these systems offer analytics contextually inside what they do.
The thing about embedded analytics that unsettles me the most is that it actually could be a plausible threat. It seems plausible that the makers of systems designed to collect and push data around would eventually figure out that customers want their data out of those systems. It seems plausible that the customer would prefer to get one solution that does everything, as opposed to two or more solutions. It seems plausible that the customer would want to get the information they need to make a decision in the place that they act, as opposed to in an external context. It seems plausible that if a provider already has access to an existing budget line item and a known product they have an advantage against someone who doesn’t. It seems plausible that the people who don’t understand the details of what a People Analyst does* would think that an analyst is entirely unnecessary to produce good analytics (*unfortunately almost nobody does).
The designers of operational HR systems are figuring some of these things out already! Late last year I went to the big HR Tech convention in Las Vegas. If you spent any time walking around the expo floor you would have noticed that everyone else got the memo on analytics too. You couldn’t buy a box of popcorn that did not include bar charts, line graphs, pies or scatter plots somewhere on it. I don’t care what it was they were selling nearly every booth had something up about “big data”, “visualization” or “insights”. Each booth was lined by massive blow-ups of data visuals screenshots. I am a data type guy, but even to me, this place seemed even more like a circus than it ever did before: “Step right up, have a look at the world's only pink elephant, and while you are at it, check out her new reporting capability.” Walking the endless sea of vendors was bizarre - sort of like a house of mirrors, if not house of terror.
I will share a few observations I have had while working this problem over in my mind. If you disagree, please tell me – maybe I missed an important detail. I sometimes do. I welcome your observations.
My view: I believe the idea that embedded analytics will somehow replace the need for external analytical technology is attractive, however based on several flawed assumptions.
Here is the best example I have. I drive a Ford F150 – this truck is fully loaded, complete with climate controlled leather seats and a motorized step that comes down so you can climb in when a door is opened. It has a combination of the practical things you look for in a truck and some the comfort people hope for in any modern vehicle. Among other amenities this truck comes with an embedded navigation system. Convenient. Only, here is the problem - I don’t use it. I use Google Maps on an Apple iPhone sitting precariously on my dash.
Why don’t I use the navigation embedded in my truck?
1.) It is not nearly as good as Google Maps in any feature, starting with: accuracy, ease of use, traffic re-routing, search, updates or any of the useful interesting possibilities made accessible via the Google Maps API.
2.) The US Auto manufacturers have not figured out how to create an intuitive user interface in their vehicles.
If I look down to try to figure out how to make the car blow less cold air on me, it is an oversized difficult exercise for something so trivial. Figuring out the precise combination and sequence of buttons I need push to get the air to back off always seems put me at serious risk of killing myself or somebody else. How an automobile manufacturer managed to make turning the wind funnel coming off my dash down so difficult escapes the curvature of my forehead.
Embedded Navigation? I don’t even want to start. The navigation in my Ford F150 is a non-starter for me– and on this point yes, you do in fact have to get the navigation sequence complete before you start driving the truck. Once you have start moving, forget it. This is Ford’s way of admitting, “operating this interface in a moving vehicle is really unsafe! Just go ahead and forget about it. What if you want to change where you are going while resting at a stoplight – forget about that too. Fortunately, on “forgetting about it”, Ford and I totally agree.
Certainly I could learn to use navigation on my truck, but why would I take time to learn that when Google Maps is so much better?
What is the chance that the embedded navigation in my truck will overtake Google Maps?
The first reason it is unlikely the navigation unit in the car will ever be better than Google Maps is there is no chance that Ford is going to resolve relative inequalities in the time period I own my car. I am more likely to buy a new car to solve the navigation problem than Ford is to improve any of the things wrong with the unit in my car. This is true now but this would also be true the minute after I buy my next car. It takes Ford at least a few years to launch a new model – in this time period what do you think Google is going to do with Google Maps?
Another reason it is unlikely is that navigation unit in my car will not likely outperform Google Maps is not important enough in the stack rank of features desired in a truck relative to any other option to actually affect buying decisions. Therefore it is not important enough to Ford to make improvements to this feature over other feature. Specifically I bought the truck to be able to stop by Home Depot on my way home from work and throw 2000 pounds of concrete in the bed. I also want to be able reliably make my way out of mud on gravel roads when the sky’s open up and it pours like God thinks Noah is somewhere in Texas. We do everything big in Texas, including flooding.
It may be a little kludgy for me to set my mobile phone precariously on the dash while I drive at top speed off waterfalls however I do it gladly, I can always pick the phone off the floor. A much better alternative to all of this is that Ford just embeds my Apple iPhone in the dash. Then we are done with this problem and we can move on to the next one. Why they have not done this I have no idea. We are driving straight for a cliff – get on it Detroit.
A careful observer may also point out that I also actually don’t use the navigation embedded in my iPhone. Navigation on my iPhone is actually much better and more intuitive than the truck, but still not what I have learned to expect from a mapping application by Google. The first application I downloaded on my iphone was Google Maps. The second, Chrome. The third Gmail. Is it the same for anyone else – I have no idea really but probably not. That is the point! Does my mobile phone provider need to make my device and all the applications I want to be relevant to me? No. Do I want them to? No.
If you pay attention you can learn everything you need to know about technology from changes in mobile phones over the last 10 years. It used to be that everything that was intended to go with a phone was embedded in that phone. The implication of this was that we only got very small incremental improvements in features from one generation of phone to the next. iPhone (IOS) and Android changed all this. Now we buy a phone with essential features of a phone, but also with an operating system that provides access to a million other things created for the phone by independent developers who have nothing to do with the manufacturer of the phone or the phone company. I prefer it this way.
It is simply not possible to build a phone another way and keep pace with the innovation of the crowd on an Apple or Google smart phone. One company may achieve brilliance at T+1 (remember Blackberry), but by T+2 the crowd has moved well beyond whatever you could do (Blackberry who?). Wherever the phone manufacturer stops, a crowd of motivated entrepreneurs pick up to create anything I need. The result is not only a more useful phone but also a much faster pace of innovation. It would have taken my mobile phone manufacturer 2000 years to stumble upon the solutions to problems that a few kids and their buddies just cracked last weekend with a case of Mountain Dew.
If auto manufacturers can’t properly embed GPS navigation right in a vehicle designed to move people from one place to another, why would I expect any more remarkable success when HR system providers try to embed “analytics” in their HR system?
What is wrong with Embedded Analytics?
The main problem with embedded analytics, as currently conceived, is that the ability to access data stored in an operational system is a far cry from what is required to produce quality People Analytics.
Relevant Data Sources External to the Operational System
First, if I want to know something about most of the important people problems I might want to analyze in an organization it will necessarily incorporate 2 or more system sources.
When I talk to organizations about People Analytics I like to say , “your problem really does not care about the boundaries of your functional siloes”. I will extend this by saying “your problem really really doesn’t care about the systems designed to service the operational service needs of your functional silos”.
Example. A basic study of what attributes relate to quality of hire involves not just data from applicant tracking systems, but also historical data from HRIS, performance, surveys and data residing in other systems.
If I want to extend the analysis into actions, if I want to influence quality of hire by increasing the quality of the overall candidate pool, I might also want to know something about how to price that position relative to a market so that I can attract and actually retain a quality applicant pool. This is the heart of the problem for embedded analytics in this case : neither internal compensation structure, nor external market data reside in an applicant tracking system – so the things I might need to see how to increase the quality of candidates going through my pipeline halts my progress right there.
The reality is that most medium to large sized organizations on any ERP are running multiple best of breed HR operational systems and relying on the assistance of multiple outsource service providers alongside them. Each of these sources have data relevant to many people problems, not to the exclusion of any other.
Analytical Workflow. If I want to perform any analysis oriented data warehousing, data transformations, or statics beyond basic reporting from an operational data store then this is a high hill for an operational system to climb.
Interface Proliferation. If we intend to use reporting and analytics in each operational system this will require that users of this information be trained to interact with data in each system. If the organization has multiple systems it will require a substantial amount of orientation and training to these varied interfaces and where precisely the user would go for what.
Relativity – The Mysterious Time and Place of Decisions. Another consideration for embedded analytics is the time and place of decision-making. If analytics have been embedded in an operational system for the explicit intention of enabling better decisions, this may not be of great value if the user is not the decision-maker or if the user is not able to make a decision in that same time and place. Often the user of HR operational systems is not the actual decision-maker and/or the user may require input from other stakeholders.
Redress: What About HR System Consolidation?
You might point out the inevitable consequence of HR System proliferation over the last 15 years is a desire to implement a single system that consolidates multiple HR system silos (Applicant Tracking, Performance Management, Compensation Management, etc.) into a single source of truth? This is probably the number one reason organizations today are implementing WorkDay, however SAP and Oracle will point out that you can also implement their ERP systems for the exact same reason. In fact, if you go back 10 years this is why many organizations implemented the legacy systems previously! Over that time period, new HR system product ideas emerged and those “Best of Breed” systems made those features found in our HRIS look like a stone wheel so we implemented them alongside. What makes us think that the minute we implement any single source solution this will not begin to occur again?
The flawed assumption about WorkDay is that WorkDay can take all functions as well or better than the market. This is just is not the case. If anything, the reaction from Best of Breed solutions providers to WorkDay’s intrusion into their market is to double down efforts to be better at what they do. The displaced will increase the pace of innovation in their space in reaction to the existential threat of WorkDay. Do you think they are just going to lay down and die?
For example. I guarantee that somewhere in the crowd of innovators, someone is working on a new way of approaching performance and it will be completely different than how we did it in the past. When a new startup cracks performance wide open, WorkDay customer will need to decide if they want to stick with WorkDay’s ho-hum version of the Performance Mgt. past - over the future. The good companies are going with the future so we are back again to disparate sources. Maybe along come One Model?
The WorkDay Rub. Even if WorkDay could copy (or buy) each new idea as it occurs, it necessarily will need to move ALL of its customers to the new idea simultaneously because the WorkDay’s cloud platform means all customers are on a single software instance. As WorkDay takes on a growing base of customers, simultaneous conversion of all customers to a new point of view will simply not be possible and this will eventually/inevitably slow down the innovation cycle in WorkDay.
Can you imagine that your HRIS providers makes you change all your internal processes with each bi-annual update? WorkDay will not be able to do so – that being the case WorkDay will not be an appealing platform to those organizations who are the early adopters of new innovations.
If you are going onto WorkDay today, you may be currently riding up the curve of HR System innovation from the last 5 years (if you had missed it before) BUT in doing this you could be ceding much slower adoption of the most important changes over the next 5.
Boiling All This Down into Principles:
- Over time, a multi-focus provider cannot innovate (and execute) as well across multiple competitive fronts as a crowd of competitors can.
- Over time, a single focus provider should be able to innovate (and execute) better in its focus than any other without a focus.
- Operational and analytical intents are facilitated by different workflows and data structures, dictating totally different system designs.
- Under pressure, operational focused providers are likely to prioritize operational priorities over analytical.
- While operational problems are usually executed more efficiently by dividing responsibilities, analysis is executed more effectively by combining data from many sources.
- Organizations may cycle between “Best of Breed” and “Single Source” over some time schedule, however the weight of recent evidence is towards ongoing innovation resulting in a proliferation of “Best of Breed” systems.
What Do I Conclude?
- There will be an increasing amount of innovation, change and competition in HR Technology over time.
- To remain competitive HR operational technology providers will need to choose a focus and maintain a lead in that focus.
- We should think about our HR Technology needs in families of interconnected products supported by interconnected workflows.
- Providers should seek out data from other systems that would be relevant to create or enhance the embedded analytics in their systems.
- Networks of technology providers who are partner friendly should have advantages over isolated providers who are not partner friendly.
- In today’s technology world, the scope, operation and accessibility of each system's API is of central importance to its position and survival.
---------------------------------------------------------------------------------
Join the Linkedin People Analytics Community group (~700 strong)
Connect with Mike West on Linkedin
---------------------------------------------------------------------------------
Gloves Off Friday Series
How the Best Company Award is Wrong
Hillary, Trump, Human Resources & The Future of The World : The Most Epic Fight of Our Time
Why Josh Bersin is Wrong About Embedded Analytics?
The Most Dangerous Technology in HR Today
What Your HR Technology Sales Rep Doesn't Want You To Know
People Analytics Question and Answer Series
What is the history of People Analytics?
What are the key questions of People Analytics?
What is the actual work of People Analytics?
Lean People Analytics Series:
Introducing Lean People Analytics
Getting Results Faster with Lean People Analytics
The Ten Types of Waste in People Analytics
Making a Business Case for People Analytics (with the three A's of Lean People Analytics)
Making a Business Case for People Analytics at Your Company (Get Started Guide)
Fueling People Analytics with Feelings
The Five Models of People Analytics
Talent Acquisition Analytics Series
Creating Competitive Advantage with Talent Acquisition Analytics
Senior Technical Product Owner, IT at Intel Corporation
7ySpot on analysis, Mike. I've thought about this, and believe that 'pure-play' analytics will almost always be ahead of the stock offerings. A current leading HRMS vendor (rhymes with Wednesday) has this much ballyhooed 'embedded analytics' which is their way of saying 'we don't have the product management vision to do something substantial, so we'll just slap something in here, and hope you buy it'! Also, I saw what you did with 'Mountain' Dew!
Digital HR Strategist
7yInteresting points all round in this post and reply. The car example is instructive. Lane drifting is a known problem with a known solution so "prompting" by technology to return to normal is very powerful and valuable. But also extremely simple - as problems go. An individual who is in the bottom 10% for Compa ratio - is a know-able problem. However the solution is not known - there are multiple right answers that depend on the history, context and desired goal. Awarding pay to the level that brings the individual to 100% compa ratio is not the right answer. This highlights a core problem for embedded analytic technology that has to rely on just the transactional data it is working from. "Embedded" analytics can work when there is a known problem (for which it has enough data) and a known solution. This is the lane change example. If the problem is known - but the solution depends on judgement and context then the technology cannot prescribe a course of action. There is no "straight lines" to move between. Do we raise pay? Do we re-class the role? Was this actually just a data entry error? Is the contribution of this individual sufficient to warrant action? The more limited your data sets the more limited your ability to divine context from the data. Hence Josh is correct that HRIS systems will be able to automate prompts against simple, clearly defined problems with known solutions. There is low hanging fruit and operational value in this area. But it won't change the game and generate competitive advantage in the way that Mike envisions. Embedded analytics has faced this problem across a range of BI domains - when the problem requires a detailed understanding of the context behind the data, the relationship between interacting systems and the overall goal for the organization. There are either so many possible actions that the choice becomes over-whelming or it has to present the most basic answer - which is usually wrong, and therefore mistrusted. As is usual in a debate both authors have correct points. HRIS systems will bring some operational efficiencies by applying basic analytics to known problems with known solutions. This will only work if you are rigorous in your internal data processes. HRIS systems will never solve the more complex, high value and strategic aspects of the workforce intelligence opportunity. They will not combine the required data nor will they keep up from a tools,techniques and customization perspective. The choice is not one or the other. To deliver HR as it should be done in the 21st century you need both. NOTE: Modern analytic solutions allow for the "embedding" of analytics at the point of decision. The data does not have to live in the HRIS to make this feasible.
The underlying premise seems to be that integrated HR solutions cannot match up to specialized players. While there is some truth to this, the fact also is that customers would prefer an integrated solution if they are not making serious compromises on functionality. It depends on how big the gap is. In the case of legacy ERPs and best of breed solutions, the gap grew significantly over time. However, HR Analytics being an area of high customer interest, there is considerable focus by the integrated solution providers such as Workday, SAP & Oracle as well. That the user profile for analytics is different, and that data from several sources is needed is true – however, this is something the major players are providing for as well. This is an interesting discussion. I do not think there is a clear answer either way and both approaches are likely to find adoption at least in the near term.
Data Coach | Helping data teams achieve great performance | Independent LLM Advisory
7yWill goole add maps to their autonomous vehicles and if so who for? I think your comparison between google maps ad Ford's is more of a quality one than a category one. Truth is you want to know which way to turn the wheel when you are sitting behind it rather than before you jump in your car.
Human Resource Information System Manager
7yI personally connect with your conclusion #3 since I've been working in HR systems for some time and, quite honestly, we're all seemingly confounded by the same things we were talking about 15 years ago...how do we apply the data from our Frankensoft ecosystems..or optimize a consolidated system. What I'm waiting for is the HR systems agnositc middle-ware revolution that will eventually aggregate any data from any system and make it intuitively useful...a translator.