AMA Highlights: How to implement machine learning into your business

ama highlights

 

Samir Sharma, CEO & Founder at Datazuum joined the Venturi Slack community on the 31st of May to host an ‘Ask Me Anything’ event. He is a data and analytics consultant and helps companies get the most out of their analytics tools and execute more effective data strategies. The AMA focused on three main areas:

 

  • What businesses can do to ensure “data-driven” doesn’t become a passing fad
  • Machine learning and its progression through the Gartner hype cycle
  • Keeping end goals in mind when building data strategies

 

Below are the highlights from the event.

You can read over the whole discussion in the Venturi’s Voice Slack Channel.


 

Question from @Sam Davie

Hey, Samir I had a read over your articles yesterday, a lot of food for thought!

You said in one of them that ‘50% percent of CDO’s fail in their journey’! That doesn’t make super comfortable reading! What would your advice be to a CDO to ensure they succeed?

Hi Sam and thanks for your question. No, it doesn’t sound good does it. I believe that many companies are still working through ownership issues with data and whether or not the CDO sits at the ExCo level or the board.  

There are many theories about this and we need to be a little cautious here.  One company we are working with right now are battling against the CDO as they believe that the CFO is the right person for this area to fall into.  This sounds like a throwback to the 80s where MI was with the CFO or FD.

Many CDOs are put in posts with little autonomy and with little budget mainly due to the fact that everyone is doing so – a data me too movement.  So the CDO becomes worthless in their position as they aren’t given the confidence or the access to the board to ensure they are at the forefront of data.

Follow up question from @Sam Davie

I also wanted to ask whether you think we’ll see the disappearance of the CDO after a few years. I’ve read that some people see the CDO as a transitional role. What’re your thoughts on that?

Really good question Sam. I honestly struggle with this one being a data practitioner.  Not every company needs a CDO. Let’s face it we can have a Head of Data or other such position.  

For now, I see the position staying, in the long-term I see the position fading and the ownership going back to the more traditional places like the CIO or COO.  We are seeing a lot of mandates for the CDO, but until the role itself, in my opinion, is either linked to revenue generation and essentially seen as part of the established presence at the C-Suite we will get a lot of mixed messages.

It’s a tough one as I have a lot of friends who are CDOs and they have struggled in posts to overcome the “oh we appointed you as a vanity project” to “oh we are hiring you as we heard it’s the silver bullet to sorting all of our data issues!”  So, stuck between a rock and a hard place, the CDO needs a lot of support at the beginning and some real short term wins to build credibility. Without this, it will fade into insignificance.


Question from @Tommy G

Hey Samir, thanks for all the reading. You end your ‘Renaissance of Data’ article with the sentence:

‘If we follow the principles of the Renaissance artists, then the “Renaissance of Data”   will be declared as a wonderful era. If not, then I fear that most organisations not following a sensible approach with data, won’t be around much longer.’

That’s quite an ominous sign-off, do you believe in the near future it will be impossible for companies to succeed without a data strategy?

Hi Tommy, I see it now where companies are shooting from the hip and not creating or aligning their data strategy to the overall business strategy.  The data strategy itself will support the business objectives over the next 3 – 5 years.

Where I’ve seen this fall down is orgs not investing in a data strategy and every department doing what they want with data platforms and becoming siloed.  This is still a massive issue and nor having something to steer the ship towards a consolidated approach to what the org wants to do with data I think is akin to the wild west.

I believe that the data strategy enhances the understanding of what is required and the pursuit of these goals are then delivered in a consistent and achievable manner.

Follow up question from Tommy G

The data wild west sounds like a rough place to live. That’s interesting what you said about no data strategy leading to siloed teams all using data individually. I would have though implementing data would lead to a break down of silos.

Do you see a CDO as the captain of a companies data ship then? (Very tenuous analogies here)

LOL!  Maybe I’ve overdone it with the term “data wild west” but to be honest, sometimes it does feel like that when you see what departments are doing in many organisations.

I have seen one department use data without thinking about taking data from another department to gain more insights, they all live in their silos and the pattern isn’t going to change until a data strategy is in place to bring together the holistic view of what the org wants to achieve as a “whole”.  

Culture is the biggest impediment to most projects whether they are data or not. One company we worked with recently we asked two divisions to sit in a room and collaborate around their data as they were asking similar questions but doing their own things in dashbaords etc., which more often than not led to confusion about numbers at a higher level.  

Holding a workshop to bring together the teams, data sets, business questions and reviewing these, adding more into the mix around innovation gave them an output which they had never considered. One thing that came out of it was the fact that they spoke different terminology about the data they were looking at, we aligned that, they had metrics that were the same with differing results so we came up with a consolidated view of those and more powerful ones based on two sets of data and external data.

They began to speak in the same language and saw the benefits of working together to make decision making and support the overall business objectives and the data strategy alignment to it.

I believe there is still a long way to go due to tech being a factor and organisations are struggling with the “so-what” factor as they have invested in so much tech and now standing back saying “we have to go back to basics and relay the foundations”.


Question @James Deeney

Have you got any advice when it comes to “democratising” data and analytics in an organisation beyond just those with technical skills?

Awesome question James I believe this can be achieved with embedding principles of data literacy into the organisation. Having data that is more accessible to people or what they still term as “self-service” which hasn’t worked at all isn’t enough.

If we look at the work of Jordan Morrow and the Data Literacy foundation we will see a HUGE need for data literacy training across the entire business if we are to have more, better and effective citizen analysts.

The DL project did research on about 8000 companies globally and found that we are in a state of data illiteracy.  Here are some of the results.

 

 

Therefore, if those are the stats we have a long way to go and the fundamental area that needs to be looked at is education. Orgs need to invest in this in order to ensure the gap is reduced between those that have and those that don’t. There is an easy way of doing this. A survey which is an industry standard will classify people into 4 areas.  here they are:

Data Aristocrat

The most data literate employees have advanced skillsets and experience in data analytics—some may even be data scientists. Along with supporting continued learning in storytelling, algorithms and the latest methodologies for data analytics, your business should help Data Aristocrats develop skills in leadership and mentoring. They can serve as evangelists and mentor others in helping to drive data-driven insights throughout your organization.

Data Knight

Driven to become more data literate, Data Knights are eager to further their skills in data science, algorithms, and statistical analysis. With an eye to progressing to Data Aristocrat, Data Knights are also looking to further their leadership, mentoring and overall business skills, including enhancing storytelling skills to demonstrate the power of data literacy.

Data Dreamer

Data Dreamers are still in the beginning stages of data literacy but have recognized the benefits of working with data in their current roles. They first need foundational learning in data and analysis as well as critical and analytical thinking. They can then build on this foundation with skills in advanced analytical concepts, visualization and storytelling.

Data Doubter

Often skeptical of the value of data-driven decisions and processes in their roles as non-data scientists, Data Doubters need to see the benefits of using data to validate intuition and tribal knowledge on which they typically rely. Awareness training is pre-requisite to overcoming barriers to change. Doubters need to understand they can leverage their existing strengths as they begin foundational work in data literacy—that it is part of their role, not a burdensome add-on. Attention to this persona is critical to preventing roadblocks that can derail a successful roll-out of a data literacy program.

Tech skills don’t cut it any and the orgs that will thrive in the future will be those that add the data literacy dimension to their data strategy.


Question from @Kirtikumar

How do you see the data security when we performing so many actions on data?

Really good question Kirtikumar. This is where I see the CDO & CISO come together.  If we go down one notch we are looking at data governance and information security.

DG puts in the standards, policies, procedures, etc. So that we know the lineage of data i.e. where it came in and the value chain that is goes through the organisation – whether that ends up as an output on a dashboard or algorithm or externally to provide others with greater insights.  

If we have sound DG in place then the Infosec teams can go on the offensive and defend the data from coming under attack.  Many orgs even with the advent of GDPR don’t have an information asset list that gives them an inventory of where stuff is.

DG is the glue that holds everything together and as part of that if we have good governance with master and reference data we will have better outcomes in machine learning and AI.  

A lot of companies made the mistake of not getting serious about DG and have paid the price for this, some are ignoring it as they think its too late to get their data in order.  The foundations need to be laid so that orgs can be protected and less siloed.