How To Become So Good At Analytics Your Boss Can’t Fire You

How To Become So Good At Analytics Your Boss Can’t Fire You

Analytics and big data are a potentially lucrative career path, but what should you do to ensure that you keep one of those roles? Here are some simple tips to make sure you’re indispensable.

Jim Sterne, chairman of the Digital Analytics Association, offered loads of advice for advancing analytics careers during a presentation at the 2015 Google Analytics User Conference in Sydney on Wednesday. His single biggest lesson? Make sure you’re slurping up data from every corner of the business, and not just concentrating on a handful of sources.

“You become dispensable with analytics by seeing things that other people can’t and using tools that other people don’t use,” he said. “All the data you can get your hands on is good stuff.”

That doesn’t mean you have to present what you do as “big data”, especially if you’re not dealing with multi-terabyte arrays. Sterne’s suggestion? Call it big data if you like, but what you do with it matters far more than the label applied to it.

Here are Sterne’s other key recommendations.

Don’t be afraid to ditch models

Analytics is based on generating models based on data, but you need to recognise their limitations. “All models are wrong, but some models are useful,” Sterne said. “All models are wrong because they are the map, not the territory” — and the challenge with territory is that it changes and evolves. “It has a limited time value.” So the approach needs to be ruthless: use the model for as long as it works, and dump it as soon as it doesn’t.

Your job is all about the odd

The key to succeeding in analytics is identifying anomalies, Sterne said. “Look for the thing that is weird. That’s the most important thing you can do. The question you should ask is: What’s wrong with this picture? The most exciting phrase is not ‘Eureka!’ but ‘That’s funny’.”

Make sure you’re constantly cleansing data

Most analysts have to deal with data that has been stored in multiple silos. Some will be in the company CRM system, some will be locked in financial data, some will have to extracted from website traffic logs. It often seems like the biggest technical challenge is integration: ensuring all that data is in one place.

While that’s important, you shouldn’t neglect an equally vital task: making sure that the data you’re accessing is up-to-date and consistent. “When data gets stale, it gets toxic, it gets dangerous.” keep data clean or you’re in trouble,” Sterne said.

If your data is going to be useful, it needs to match all of these criteria:


  • valid
  • credible
  • reliable
  • consistent
  • clean


  • unbiased
  • defined
  • relevant
  • correlate-able
  • understandable
  • timely


“You don’t need to know absolutely everything but you need to know you can trust the data that you’re working with,” Sterne said.

Know where your tools intersect

You’ll typically be using a range of tools to produce your results, from high-end analytics platforms like Hadoop through to basic spreadsheets stored online. You have to understand where all these things fit in the hierarchy,” Sterne suggests. Otherwise your silos of data will be clashing with silos of analyses, and you’ll spend your time fixing those issues rather than coming up with insights.

Don’t trust dashboards or averages

Dashboards are a popular choice for summarising analysis, because they’re cheerful and visual and easy to understand. They can be very useful, but you need to avoid the trap of just using pre-supplied dashboards. Chances are if you do that, you won’t be conducting analysis that’s relevant to your specific business problems.

“Dashboards are completely useless if you do not know the problem you are trying to solve,” Sterne said. Define the problem then build the dashboard — don’t work the other way

One thing your dashboard should not feature is any averages. Sterne was entertainingly blunt on this point: “Anybody who wants to know the average anything, just shoot them. I have one foot in a bucket of boiling water and one foot in a block of ice and on average, I feel great! On average Switzerland is flat. Averages are not your friend.”

Stop complaining, and don’t say too much

Having completed your analysis, you also need to communicate it effectively. Don’t fall into the trap of overloading other managers with numbers.

“Tell me the story, but don’t give me all the data,” Sterne said. “I hired you to play with the data. I don’t want to hear how difficult it was.

“Tell me what it means. You are the doctor and I want to know do I take these pills or do I need this surgery? I don’t want a course in advanced biochemistry.” To ensure the data is relevant, identify what your boss’ goals are, and tie the information you present into that.

Also recognise that it’s not ultimately about individual numbers. “Creativity is part of your job, don’t short change yourself. If you’re spending eight days a week doing the mechanics, you’re doing it wrong. Have an opinion. That is your value. That is your job.”

Angus Kidman is editor-in-chief for comparison site and, a former editor for Lifehacker Australia and a man who is always happy when he’s staring at a spreadsheet. Follow him on Twitter @gusworldau.


  • Great article – love the “Stop complaining, and don’t say too much” section. The culture of complaining about analytic and reporting in our work place is rife.

    Also, did Lifehacker’s editor post go unfilled when Angus departed?

    Angus’ bio at the bottom: […] who is always happy when he’s starting at a spreadsheet […]

    […] who is always happy when he’s **staring** (no T) at a spreadsheet […]

  • Used to do analytics in an old job, the team did such a good job we were absorbed into a bigger area and I was ultimately moved on to do something completely different. No matter, learned a lot doing the role (and it helps in the current job anyway) and was thankful. Anyhow, one of the things I always remember being told about analytics was that 99% of the work is to find the 1% thats unusual.

    Always seemed a good rule of thumb to me and a simple ratio, but what gets lost in reporting is that is IS about finding that needle, and the efforts getting there might not be understood.

    So while “Stop complaining, and don’t say too much” is good advice, so is “don’t short change yourself”. Dont overload people with numbers, but make sure they are aware of how big the needle is.

    The final analysis might be “we found X that could mean Y”, and ultimately be a one liner, but context is still important, so add that second line in to support the conclusions.

Log in to comment on this story!