What we learned about HR Analytics in 2014 – Part 2
Just as in the previous years we have been rushing from HR Analytics project to project in 2014. It was the year with the highest number of HR Analytics projects ever and the projects became much more sophisticated. More and more large organisations are building capability for HR Analytics, either internally or externally. Meanwhile our blog has reached an incredible traffic which means that HR people from around the world are doing their home work in preparing for their analytical future.
Organisations seem to be in even greater distress than last year to gain insights into the relationship between investment in human capital and its impact on the results. As I did last year, I’ve been reflecting, together with the fantastic iNostix team, on a the most important lessons from the past busy year. This is part 2 with learnings #8 till #14, you can find part 1 here…enjoy the reading!
8. Hiring a just-promoted PhD is not a good start
A number of organisations recently chose to hire a young, just-promoted academic (on a permanent or temporary contract). Their rationale: these people learned to work with data and to research correlations in a scientifically sound way. In itself, this is a legitimate reason to take such a step, but I personally don’t think it’s a good starting point. PhDers may have the skills and knowledge I just mentioned, but they won’t be able to put HR analytics sufficiently high on the organisation’s radar.
Doing analyses is one thing, but determining the analytical agenda, adding value to the organisation’s core processes, steering often complex cross-functional collaboration, breaking through the difficult internal silos, bureaucracy and politics, presenting the business case to the management, etc. is another. If you would like to move the needle, you should start with a heavier (consultant-type) profile with strong analytical acumen that can make an impact on management level. Then, only in a second phase, add an academic to this as second in command!
9. Forget ‘big data’ at the start
I’ve written about this quite a few times: forget all the ado about big data and focus on the qualitative development of your internal (HR and/or business) data first. Almost every week, I have to explain that HR itself has no big data, but that we can learn a lot from the big data phenomenon: evolving from reporting to predicting, focusing on disciplined qualitative (vs. administrative) data gathering, adding (internal or external) data experts to the team, thinking about data governance, enriching decision-making with long data.
Long data? If we have learned anything in recent years, it is that excellent added value can be delivered with long data. Long data comprises data sets that cover a long history, paint a changing and moving picture, map out processes and interactions that change through time. For example, if you’re going to measure how successful your recruitment is (quality of hire), you will know the results only in a year from now or even later than that. On condition that you gathered all the relevant data in a disciplined way. So: long live long data! And…start tomorrow!
10. Let politics and bureaucracy not stand in the way of delivering added value
The greatest challenge does not always lie in doing predictive analytics itself, but in getting the organisation to move, based on the outcomes of those analyses. We experience this in every project, almost ad nauseam. Hypotheses, data, analyses and answers are undoubtedly all very important, but the way in which these questions, analyses and answers are in agreement or in conflict with an organisation’s internal politics, behaviours and convictions is more important many times over.
Sometimes even the best analytical outcomes prompt very counterproductive or defensive reactions. Getting an organisation to move based on analytical outcomes is hard work (and not a job for a junior PhDer). HR often seriously underestimates potential political resistance: it’s got great analyses, knowledgeable answers to essential business problems and still there’s no resonance in the organisation. That’s how analytics can (also) go wrong.
11. You can start tomorrow!
It’s a quote from HR Analytics Manager Patrick Coolen (read his HR analytics journey at ABN-AMRO here): ‘You can start tomorrow!’ Why does Patrick give this tip? Together with us, Patrick is convinced that you shouldn’t wait until all your data is in order before starting with HR analytics – a stance (and sometimes an excuse) often heard from HR. It’s a parallel and even interactive process: analysing data and at the same time getting to know its real quality when you start working with it. You can’t begin analysing early enough, certainly not if you want to work on your long data (see #9 above). From now on you’ll probably need to keep track of a whole series of data for longer, so it’s best to start doing this right now. Tomorrow! Thank you, Patrick, for your wise council!
12. Make HR analytics into a continuous process and not a series of loose projects
Because in most HR departments analytical maturity is still at a beginner stage, we often see a series of loose projects with relatively little cohesion. That’s completely legitimate (and even very appropriate) for testing and learning, but in the long term such a patchwork approach doesn’t help enough to embed HR’s analytical leadership in the organisation.
In a second phase, after the initial experiments, we always recommend focusing on a (limited) number of key processes in the organisation and to analyse those and follow them up for a longer period. The idea must be to add value to these key processes and optimise them in the long term. That can’t be done with one-off analyses, but with (strategically important) measurements that have a long time horizon, with regular updates, continuous collaboration with process owners and by collecting long data (see #9 above).
13. Engagement data must be handled with great care
I see lots of organisations making desperate attempts to use their engagement data and apply it to other organisation or business data, as a kind of first HR analytics experiment. Engagement analysis features on almost all HR analytics agendas and arises from an assumption that impactful outcomes will necessarily result from this kind of engagement-based analyses. I don’t find this a good evolution, for several reasons:
- Not an aim in itself: using engagement data should not be an aim in itself. It’s a supporting and supplementary source of data to help analyse organisations’ core processes better. So, depart from your business or organisation problems and examine whether engagement, together with other predictors, has an impact on this. Organisations that depart from engagement data and want to study what could potentially be influenced by engagement, mostly find only marginal and meaningless correlations (after endlessly long analysis!) that are very difficult to substantiate in a business meeting. So be (very) careful!
- Aged data: I see organisations using engagement data that was gathered sometimes one year ago or even before that. Don’t do this! In today’s quick-changing organisations where one announcement from the management can change everything in the blink of an eye, engagement outcomes are only a snapshot. If you want to use this data at all, it’s best to do it immediately after (or even at the same time as) the survey. The shorter the period between the engagement meeting and the linking with data on core processes, the more reliable your analysis will be.
- For frontline employees: the impact of engagement can be made visible the most easily in frontline employees, those who are in direct contact with customers: baristas in the coffee shop, call agents at the bank, field technicians, sales reps, coach drivers, etc. If you compare business output in clusters of high or low engagement in these frontliners, you’ll probably find interesting differences. A few examples of output differences between high/low frontline engagement are productivity, sales, customer satisfaction, absenteeism, staff turnover, etc. If you apply these same high/low clusters to the (subjective) assessment scores of your (non-client facing) internal employees, the story becomes very flimsy though.
- Linking on an individual level: only very few companies have engagement data on an individual level, due to the anonymity of it. However, because of its statistical robustness, it’s a must to link it to individuals, in order to put everything together again later to ensure that same anonymity. If you can only link on a team or department level, don’t put too much energy into it, because it will generally provide very little usable output.
14. When starting out, avoid using an analytical umbrella ‘Centre of Excellence’
Some larger organisations (like banks, insurance companies, contact centres, etc.) already have important analytical activities going on in their marketing, risk management, actuary or commercial departments. Analytical umbrella ‘Centres of Excellence’ are often set up to be able to serve several departments at the same time, to work in a more productive way on shared technical platforms, to share (scarce) analytical resources, etc. HR professionals often ask me whether it wouldn’t be appropriate to start using the services of these COEs for advanced analyses if HR is lacking in the necessary skills. I’m not really in favour of that, for the following reasons:
- HR is not a priority: these COEs’ diaries are mostly filled to the limit with business-related projects. Experience has shown that, if HR joins the queue as well – without a strong analytical track record – it will unfortunately and inevitably end up at the bottom of the ladder of priorities and will have to wait for many months to be served… if at all.
- Too few learnings: if HR transfers its analyses to a COE, they will pass by all the learnings I described in part 1 under #5 and they’ll develop insufficient maturity, gain insufficient insights in the quality of the (long) data and barely be able to come up with structural measures to add value to the organisation’s core processes. This means the focus will shift exclusively onto the output of analyses (done by the COE) and not on the wider context.
- Problem with confidentiality: and last but not least: if you were to drop HR data with a COE, you will undoubtedly have an instant problem with the confidentiality of your employee data. Up till now I haven’t yet found any good solutions to this issue, unless it is building up analytical capacity within HR.
Maybe the COE concept can be worth considering in the longer term, but first HR needs to reach a higher maturity level to be able to manage the analysts in such a COE and to find a solution to the issue of employee data confidentiality.
About the author
Luk Smeyers is an experienced senior HR executive who has lead complex transition projects for compelling Fortune 500 companies, such as PepsiCo, Starbucks and Nielsen. In 2008, Luk started a cutting-edge predictive HR Analytics consultancy, together with academic partner Dr. Jeroen Delmotte. Luk is widely recognized as one of the few European top HR analytics experts. He is revered as a leading thinker, educator, influencer and is a well-known content contributor, blogger, columnist and author of many articles. He is an invited speaker at international conferences and academic programs and helps clients set a higher ambition for strategic HR intelligence, leading consultative projects in the Benelux countries with such organizations as Ahold, ING, KPN, ABN-AMRO, Philips, Rabobank, UWV, RealDolmen, Acerta, NS, BASF, Besix, Strukton, Bekaert, Randstad, Eandis, AG, Postnl, AON, Raet, etc.
Interested in using predictive HR analytics as a key component in your HR strategy? Get in touch with CEO and co-founder Luk Smeyers for more information (email@example.com or via Google+). Or follow iNostix on Twitter and/or Facebook for exciting international articles on HR analytics. And don’t forget to register for this blog!