How to save — or tank — your data strategy
The least effective protocol is to attack this as a technical problem. Large organizations are slow to change, and a new SDK, API or library is not going to enable new behavior. A few motivated developers may pick them up and build apps that become useful and popular, but a bottoms-up approach doesn’t change people’s rewards, and therefore doesn’t change behavior at scale. The most effective protocol is to recruit the CEO as the change agent for becoming great at using data. I saw this at Autodesk in 2018 when I worked as part of Andrew Anagnost’s change strategy to move the company from desktop software to data services, and I see it in JPMorgan Chase as Jamie Dimon presents publicly on the business value of data and his mandate to all business managers to make AI part of their business plan for the next fiscal year. The CEO-led approach must be paired with effective, usable technology, or it will be an empty mandate that generates cynicism; and it must also carry a robust set of measurable goals and rewards. Promotions for technical and business leaders must include a “data change leadership” flag: This sends the message that good or great performance by the old standards is no longer enough, and that the bar is moving. This can create the necessary conditions for employees to self-organize into guilds or communities of practice, encourage each other, and shift the culture to continuously evolve towards higher competency with data.
The most effective ways to help organizations get better at using data is to focus on three key areas — relevancy, accessibility and a single source of truth. One of the problems with data is that often there is a lot of it to sort through. This is where relevancy comes in — it is critical to identify metrics that are relevant to the decision-making process. This requires picking the right metrics so organizations are looking at the right measurements when making a decision. This applies to both results and early indicators. For any data strategy to be effective, data needs to be accessible in an easy form across the organization. For example, if there is a requirement for an analyst to pull reports, then it is not operationally viable because we have inserted complexity into the process. Finally, there needs to be a single source of truth. It is paramount for organizations to ensure that everyone is working off a single, trusted version of the data. As far as ineffective strategies go, organizations need to accept that on some level data will never be perfect. Waiting for perfection will only impede progress, so there is a need to be comfortable making decisions with the data you have, rather than waiting for what’s missing. And while metrics are important as stated above, it is possible to have too many metrics — it is critical to take the time to define which metrics really matter rather than sifting through too many data points.
Too many data and AI projects fail because their creators rush into the technical implementation of solutions without first clearly defining what real-world problems they are trying to solve and what success looks like for their business. That’s a losing strategy. Using data properly means aligning the business strategy to the data and AI strategy to provide solutions that solve real human problems. A successful data strategy-setting process should: Set intent: Discover the intent by uncovering the targeted data-driven business opportunities and the unanswered questions that are critical to achieve the objectives. Identify: Define the use cases, selecting the types of data and solutions needed by the users. Evaluate: Take a critical look at the data sources needed to implement DataOps. Plan: Set concrete actions by using statements of intent as a guide for the technical implementation. Tweet this. Today, most companies aren’t getting the maximum value out of their data.
