eWeekTweetChat

A8 One of the biggest #data analytics myths is: The more data you have, the better decision you can make. The reality is the more data you have, the more challenges you face. What we seek is the “right” data. We get there by understanding the true meaning behind the

A8: The big myth is that collecting, transforming and cleansing the #data is the hard part. It’s not. Operationalizing the data, or even a subset of the data to lead to transformative business results is the hard part.

A8. 'Build it and they will come.'

A8 - Myth: Data and data analytics is rocket science… it’s not, and I was a rocket scientist.

A8: That all AI systems are black boxes. AI #explainability and interpretability are growing fields that allows us to peek inside these black boxes or fully explain models, behavior and results.

A8: That it will solve all your problems overnight — data collection, processing, and analytics is made easier every day, but it's a journey. Start small with tangible problems you can solve and as data, complexity, and skills grow so will ROI

"More data means better decisions." When it comes to data, quality is infinitely more important than quantity. The best insights are derived from data that is clean, trusted, timely and relevant. #DataAnalytics #eWEEKchat @CIOStraightTalk

A8. I can build my own platform.

@RKs2cents 100% agree, it's not more data, its data that leads to insights that move the needle for your business.

A8: Quantity more important than quality. While it's important to save & use all the data available, data quality is just as important
as data quantity .. Second Myth:Big data will solve all your problems:
Subject matter expertise on interpreting the data is important

(edited)

A8. Oh, never believe that 'Data is Truth', or machines are impartial. Data is manipulated – lies, damned lies, and statistics – and we build bias into the machines … ALL. THE. DAMN. TIME!!

A8 One myth is that all of your data has to be of the highest quality to be useful. This is the case for compliance reporting or when making critical business decisions, but it ultimately depends on individual use cases and data gov required

@BCorr_BMC That may be true for environments that have been around a long time (datacenter/cloud). What about edge? its the wild, wild, west out there.

Another myth is that a good analytics platform is sufficient for great analytic predictions and results. While the platform is a part of the entire puzzle, there are other key elements that drive predictions and decisions. #DataAnalytics #eWEEKchat @CIOStraightTalk

A8: The inverse is that analytics is too hard to do or that businesses have "lagged behind" — The truth is that technology, data, and how we do analytics changes very rapidly. You haven't missed the boat. #DataAnalytics #eWEEKchat @CIOStraightTalk

A8. Comes back to the idea that any science sufficiently advanced looks like magic. People look at MLAI and #analytics and think it is magic - and thus infallible. Nothing further from the truth.

@BCorr_BMC Operationalizing the #data is challenging, but the #DataQuality too often remains in question ... making it difficult to operationalize.

A8: That #data leads to binary decision making - deciding to or not to do something, whether something was a good or bad decision. In the majority of cases, good data is suggestive or predictive, but not prescriptive.

A8: We'll just buy some analytics software and BAM - we'll have our insights!

@brucekornfeld yes, the edge is the wild west right now, with many roll-your-own solutions dominating.

@kklive I like to say "decisions made on bad data is just bad decision that you don't know about yet".

@RKs2cents Yes, without data quality nothing else matters.

@kklive Yes KK, 100% agree. The platform is just another tool. Data quality, data volume, reliable algos, smart people all probably more important in fact

A8: A big myth is that there's no bias you're introducing by pulling a simple average. If all corners of your business are making their own definitions of KPIs and attributes, you're going to get into a number matching loop that wastes time and resources. 1/2

A8: For this reason, a shared data catalog/dictionary, well labelled and organized data, and orchestration to get the right data into the right platforms are absolutely vital. 2/2

A8: That the data is there and shows itself without strategy and extraction

@chimerasaurus Excellent! We over-rotate to believing data can somehow make complex decisions easy - and even binary. It cannot, and we create great risk if we believe it can.

A5 Artificial intelligence is changing the game for everyone. On the #data side, it’s helping us determine the best use cases for #DataOps and automating the management and processing of it. There’s so much more potential with #AI. But human oversight remains key.

A5: Leverage #DataOps technologies to automate the data capture, cleansing & transformation so the data science & engineering can work on manipulating the data to drive business results such as higher customer retention, reducing $$, or creating new digital offerings.

@RKs2cents 100% agree with #AI changing the game across every use case.

A5: Engage & interact w/ insights – tell #datastories that make an impact. Enhance engagement & understanding with 3D & immersive data visualizations to clearly illustrate complex findings. Improve collaboration & bring business and #datascience teams together in a virtual space

Q5 - Go from being IT central to business focused and federated. Data literacy is NOW just as important to traditional literacy in the workplace. Data sharing, federated data, and data literate organizations will be the most successful.

A5:The extra for best-in-class results is about ensuring reproducibility of data engineering and ML processes. This includes continuous model training, data and model drift analysis, feature engineering, CI/CD integrations. DataOps/MLOps is the Key

A5. Focus on the use cases that business can't easily or accurately get answers to today. Solving problems that are already solved, just using newer technology, doesn't move the needle. What can new tech enable you to do today that you couldn't before?

Make data privacy and compliance an integral part of your data analytics strategy. While the democratization of analytics is a current trend, we need to put an emphasis on data privacy, confidentiality and compliance. #DataAnalytics #eWEEKchat @CIOStraightTalk

A5. Retain great people. You will need them in ways you cannot predict.

@rqrivera #DataLiteracy and #DataOps go hand in hand. You need to be literate of the data to the agility to succeed in the market.

A5: With the third party cookie losing steam, we've learned that finding non-competitive or complementary brands that are suitable for data-sharing can complete the picture of your customers' behavior. 1/2

A5. Sounds obvious but...the data strategy needs to be aligned with the business strategy rather than managed separately. This means data leaders have to become more business literate & understand how the data strategy is going to contribute to business success

@kklive Completely agree, privacy, security and compliance are often afterthoughts in data strategy

A5: Automate everything you can. The faster you can move data from point of origin to a real-time dashboard, or ML model, or AI app, the more effective your business will become.

A5: By sharing anonymized, aggregated customer data, collaborating organizations can gain a clearer understanding of who their customers are and what motivates them. (2/2)

@jonnio20 Given the huge gap in supply of talent that won't be solved any time soon, this is the right answer.

A5: Choose tools and platforms that are additive over time. eg: In the #DataCloud we have #DataSharing (yay no more copying data!) but this feature will work with all the new features coming out too. Platform scales with data and tech changes seamlessly without effort.

(edited)

A5 Still on the 'trust' tip ... use open (incl. Open Source) ML algorithms. Black boxes are damaging for trust and effectiveness. Open algos drive up understanding, trust, and ultimately make data more actionable because decision-makers believe it.

@RKs2cents We should be focusing on Augmented Intelligence powered AI, not just powered by AI

A5: Prioritize and mandate data visualization and data democratization, so actionable data is not only, in fact, captured, but applied independently by teams for their goals.

@RKs2cents AI Ethics is long overdue, Already can see the black box ML products of the last AI hype period are losing favor in the democratization of AI - now we need to solve labeling problems (plug for @snorkelai)

A5: Working with massive data is still hard. Find approaches like universal data distribution with DataFlow and automated ETL for powering analytics and AI downstream. Look to unify your data lifecycle and move away from siloed and expensive solutions

@rqrivera Absolutely. Having an #AI powered data catalog is one way to solve it.

@namessanti You know, I normally "Automate all the things!!!" And #MLOps etc. can automate mundane tasks. But I would advise caution, coz #automation can do very bad things at massive scale and speed, but ML is not an exact science.

@fblissjr I would advise that you start with the simplest UC with the lease complexity that provides the most value in the shortest time. Then scale from there.

@hmschwenk 100%

A5 Data literacy needs to be a priority. People in the org need to understand the data value chain across the business and the part they individually play in it, for eg when a customer on boards