The abandoned analyst

You finally consolidated all the important data sources in your company. No more outdated monthly report.

You deployed your ETL jobs and optimised SQL scripts. You built a beautiful dashboard and documented all the business rules. You presented your solution. The business loved it.

A few weeks past. Nobody logged into your dashboard …

Maybe just a communications issue. You sent another comms and attached a user guide.

One day the server crashed. Nobody noticed …

They were running things on spreadsheets again.

So you sat there in the company town hall, with an astonished look as the CEO emphasised how important data analytics is for the future of the company.

It didn’t feel that way to you, or to the data analysts and engineers who found this story painfully familiar.

I want to talk about why sometimes our expertly executed data projects are met with such cold apathy.

I will discuss:

• The often-mentioned, not-often-delivered actionable insights.
• Impacts vs metrics.
• Our obsession with accuracy.
• Being an ally to the business.

The dreadful actionable insights

“Actionable insights” must be the most uttered words in data professions.

What does it even mean?

Insight <> problem

A broken water pipe is a problem. It’s not an insight. Your water pipe broke because you fixed it yourself after 15 minutes learning on Youtube to save money is an insight.

The revenue has been trending down the last three quarters is not an insight. Anyone with access to the data and minimal excel skills could figure that out.

The revenue is trending down because the new checkout page sucks is an insight.

Too often, we come to meetings with only problems on those bar charts. Sales are bad, staff utilisation is low, tech support demand is going up …

Maybe we should keep in mind that the messenger of bad news does get shot.

We don’t have to be just the messenger, we can be enablers.

Actionable insights are insights that you can act on immediately. Our job as an analyst is to find the root cause and recommend actions.

One approach to ensure your data inform decisions is to work backwards - find out what action you are willing to take and can take within a reasonable time frame, then decide on what data you need to decide if you should take that action.

Measure impacts, not metrics

If I look at your dashboard right now, does it tell me if I should buy your company shares?

Any metrics that do not immediately tell you if your business is doing well or not is just that, a metric.

Is high traffic to our physical stores a good measure of business impact? Not if most people just want to see the products and go buy them online from our competitor.

Website visit figure is also just a good-to-know if it is not translated into profit. Even when it is, you want to measure how much money you’re making from clicks, not the site visit itself.

If you don’t measure the impact, you leave the interpretation to the audience, then you’re providing reporting, not analysis.

As a side note, we should also be careful when measuring success by dollar.

For example, you can show how the running cost of your department is going down, (Yay!). But wait. Is it a result of laying off 20 senior consultants who have been bringing in a healthy profit margin for you?

So while measuring concrete outcomes is valuable, context is, too. This is when a good relationship with people from operations and finance is indispensable. They can help you navigate the complex web of processes, cost landing, and teach you what you should capture to show real success (or failure).

The moment your audience stops thinking “so what?” when they look at your numbers and want to act, you’re on the right track.

Our obsession with accuracy

I will be the first to admit I spent way too many hours trying the get my number accurate. At the same time, I am aware that my data suffer from “garbage in, garbage out”.

The users did not put in the right numbers, the tool we use only records time in 30 minutes blocks, the vendor software only samples 10% of our data in their extracts. And many more reasons why the data was wrong in the first place.

This is not a call to abandon efforts to improve accuracy. It’s a call to accept the limitation of our data.

As long as the way you measure impacts is consistent, i.e., I can compare figures from this quarter to the last few, or between customer segments, then your data still brings values.

You may spend a behemoth effort on making the numbers accurate, only to find out that nobody cares.

Being liberated from the need to be painstakingly correct - which is impossible anyway - you have the time and resources to achieve something better.

Being an ally

The reports and summaries you produce are very important. But you are not.

If you’re gone, someone else can be trained to do your job. Or worse, your reports are automated (maybe by you!), and you are obsolete.

What you want is to position yourself so that people want to get you involved in their endeavours.

You want the product manager to ask you to set up a dashboard to track the performance of a new product offering. You want to be helping the marketer to set up an experiment to test different marketing email layouts.

You need to build a good relationship with the business and be a data evangelic because you are the expert here. Take advantage of your access to the plethora of data and help people. Show them what value you can bring.

Be a helpful ally, not a carrier of bad news.

The caveat

Most of the above ask you to be proactive and plugged in. However, this may be not possible due to the culture of your company.

It may be siloed, and knowledge is not shared between areas, or even within a team. In that case, you have no choice but to provide reports to enable leaders to do the analysis themselves and make decisions.

Endnotes

Being a data analyst is a rewarding career that keeps you in touch with the business and how it is performing.

Armed with technical skills and business knowledge, you can become a valuable asset to your company and get to do more interesting works than just summarising numbers.