The "frequency/quality-limit" in data gathering and it's impact on human capital management

Hurricane eyes put human lives into perspective. Those isolated in its wake have the immense privilege of understanding in seconds what it means to be alive and the immense value of the fragile gift they have received.

Last Friday, the collapse of SBV affected the entire Bay Area tech ecosystem, and while this is a first-world "problem" affecting only the most privileged ones, like myself, it's still a good moment to put things in perspective before the market opens in a few hours and the noise of our generation begins again.

Three years ago, Erudit AI was created with a singular goal: to make of workers' understanding and management the cornerstone of every business strategy worldwide, without any fluff. The hardest thing to be solved? the frequency/quality limit on current people-data gathering systems.

For business and HR professionals, two main sources of people data were available:

• Administrative data contained in the HRSIS
• Data collected from surveys.

The first source's data creation frequency is 100% manual in most companies and is subject to administrative and legal procedures, which may hinder its use. Although it can be a powerful cohorts creator, on its own, it will only serve as a source of lagged reports.

The second source is a powerful industry in itself, with the survey as its main asset. While it is valuable in providing insight into how workers may feel, interrupting "only" their work cycles and not their busy people managers ones, it has inherent problems such as huge biases, survey fatigue, frequency limits, and lack of sample representativeness.

The frequency limit

Every cutting-edge industry with a considerable economic impact on public or private markets, is facing the challenge of having better real-time predictive models, every day. Therefore, if we want to encourage investment in tools for improving the understanding, management, and well-being of workers, we have to increase the level of sophistication in the people analytics industry.

Any prediction will be only as good or bad as the size, quality, and updating frequency of its dataset.

Can you imagine the stock market, marketing professionals, or any kind of asset managers operating with data based on sporadic surveys, collected from biased samples weeks, months, or years ago?

The people analytics industry should understand that, on these days, important budgets will never be released in exchange of simplistic improvements on data visualization and data management, those are taken for granted. The real outcome should be a substantial improvement in the quality and frequency of the data gathered, allowing exponentially better predictive models for the human capital industry.

"Survey Fatigue" Stephen R. Porter, Michael E. Whitcomb, William H. Weitzer

Of course, there are ways to increase the frequency of data collection in the people management industry, and of course there are other variables that affect survey fatigue besides frequency, such as surveys content, length, communication, and actionability. But let's not kid ourselves, that glass ceiling was reached a long time ago, industrial revolution has arrived, and we are still stubbornly trying to haul goods with oxen.

Quantifying survey fatigue

"Exhaustive or exhausting? Evidence on respondent fatigue in long surveys"

Author: Dahyeon Jeong, Shilpa Aggarwal, Jonathan Robinson, Naresh Kumar, Alan Spearot, David Sungho Park

Publication: Journal of Development Economics

It's thoroughly demonstrated Surveys have a quantitative and qualitative limit, and it has been reached decades ago, today, there is no a single major business decisions taken at any industry based on this kind of data source.

For attracting heavy investments, showing solid patterns and tendencies with enough levels of confidence and accuracy is crucial. This is the first step towards building a solid future for the People Analytics industry success.

Breaking current limits of Human Capital management industry

Workplace survey systems aim to receive unbiased free-text responses as frequently as possible. However, this feedback is often hindered by the mandatory interruption of work and productivity. A healthy company culture can facilitate this type of feedback, but increasing survey frequency may result in more micro-interruptions and lead to a revision of employees' critical priorities and responsibilities. As a consequence, this may decrease response rates non-linearly and increase biases in the responses obtained.

We are in the right path, high volume of high quality and frequent free text, is the key for breaking HR tech limits, but how can we get this avoiding survey fatigue and inherent biases?

The raise of organizational psychology + AI

• Stage 01: Creation of specific NLP models for measuring key organizational-psychology related metrics from open text. - The metrics that matter and the ones driving Human capital management and it's economic impact are well defined. To create a definition for each of them for every macro scenario and a scoring scale, allows to create a simple system for labelling open text just including organizational psychologists in the data labelling process. From here any single open text message is transformed in multiple metrics' scorings automatically on real time by de-berta or other kind of AI basic models

• Stage 02: Give people the chance, to no answer surveys, removing workflows interruptions. - As worker I generate text at work in many channels, surveys, internal communications tools, docs, etc I want to decide which one I allow you to be used in and anonymized and aggregate way as data source. Some days I will be ok with answering surveys, but most of them I will have a better outcome to be focused on and I want my employer having my mood and pulse in consideration anyways.

• Stage 03: AI assisted actionability. - Curated suggested actions based on goals, real time metrics and scenarios are easy to create and offer today in substitution to old school low actionable catalogs formed by PDFs, playbooks etc

There is a lot of work to do, but once the motto crystallizes, to get trough the second half of the storm is always easier.

Greetings from the Bay Area ???