Continuous worker voice - data

Mar 30
5 min read

Updated: Apr 1

Imagine workers could tell you every day how their labour rights are being respected - perhaps providing up answers to 80+ questions that cover a full spectrum of critical, major and minor potential issues.

Wow!

BUT that is possible right now with continuous worker voice - low-cost, anonymous, authentic feedback direct from workers.

This leads quickly to another problem - how do we make data at this scale useful?

Data at scale -making it work for you

Continuous worker voice brings a new challenge - with data available every day and potentially from all the workers in a given location - we can expect up to 50,000 or more data points every month on a workplace.

Continuous data is very useful - in fact, it is essential.

As every responsible purchasing manager knows, workplace conditions can change from week-to-week reflecting everything from the weather (eg: food and fishery), to order books (eg: factories and processing plants), to the worker population (eg: migrant and agency workers), to the seasons (seasonal products).

Continuous data is not too much data - but the data needs to be presented properly.

Making sense of the data needs powerful dashboard supported by efficient and secure engineering "under the bonnet".

The bigger picture - the smaller picture

It is quite easy to think of the "big picture".

Ask The Workers is real-time. Workers use an app on their phones to provide their feedback every day.
Whilst each worker only answers 10 questions individually, these are pulled randomly from a larger panel ensuring all the questions are answered every day multiple times.
Efficient, buIlt-in financial incentives encourage daily participation.

With Ask the Workers, feedback is provided by workers numerically - by providing a score between 1 and 10 as their response to each question or answering yes or no.

The big picture can be a simple number, the average of all the scores provided by the all the workers for a given time period.

Here is an example of a set of averages over time - showing the average of the scores given by workers to questions that are mapped to the ETI base codes:

Continuous worker voice data and dashboards

These simple "average scores" can be useful to track long term trends, but they are not very useful at revealing what is really going on. Totals and averages can hide a lot of problems.

The smaller picture is very important.

We need to take that overall average and break it down into at least three dimensions for each workplace:

Time
Demographic
Base code and even question

Ask The Workers - "bottom slicing"

You can have your head in the fridge and your feet in the fire. Your average temperature will be fine, but you will not be comfortable!

This the point. We don't want to be misled by averages.

We need to be able to see quickly and easily into the corners of the data and uncover disadvantaged or vulnerable demograhpic groups within the worker population - and make sure they are not reporting concerns.

The Ask the Worker's platform uses several statistical techniques to help users quickly dig into the data as it flows in real-time.

One of the most powerful is "bottom-slicing."

What we do is remove the top 60% or even 80% of the responses by value so we can see the feedback from the bottom part of the response distribution. We can then further segment that lowest set of scores if needed by different demographic factors individually or in combination.

Here's an example from our live data

This is taken from the very same dataset and time period as above but now showing our bottom-slicing technique at work for one of the ETI base codes:

Continuous worker voice data and spider charts

Each spoke on the chart shows the responses of workers to a question that has been mapped to this ETI base code. It is the same data for the same time period as the line chart above, but now presented three ways:

the average of all the responses, and then
the average calculated but with the top 60% of responses removed; and
the average calcualated but with the top 80% of responses removed.

As you would expect, removing a large part of the better scores means the orange and the dark blue dots move closer to the centre (indicating a lower average for their part of the population).

It immediately reveals what is hidden in the data.

There is a minority cohort in the workforce that has issues with access to water and a larger group that reports that pregnant workers are not being given less onerous tasks.

You can see this clearly in the chart. For most questions, the relationship between the inner dots and the outer dots is similar. But for two questions, the light blue zone is much bigger and the dark blue and orange dots are much closer to the centre. There is a hidden cohort in the average data that has reported some issues. And, let's also note, this is not a small number of workers.

Digging deeper is easy - the user just clicks on a dot and a drilldown becomes available to zero-in on the specific question.

Here is part of the drilldown relating to water access based on this same dataset:

Continuous worker voice data and histograms

This shows the distribution of responses for each workplace, but with five segments for each (all, 80%, 60%. 40% and 20% (ie: it is a histogram)).

We can now see that there are a couple of workplaces where access to water is an issue that workers on those locations are reporting - although most workplaces are fine on this point. In fact, this dataset coincided with a heatwave in the area where these workplaces were located.

The average looked okay - but it was hiding an important health and safety issue.

Minor issue or critical issues - they do all matter

This is just an example of what continuous worker voice data can show taken from real-world data. It shows how well-designed dashboards can enable users to zero-in quickly and efficiently on where issues may reside - identifying vulnerable cohorts that may be reporting concerns.

This was a quick and easy fix for these workplaces to address - and because workers are providing data continuously, it was then possible to verify that workers accepted the remedy and that the remedy was then sustained.