Businesses Are Losing Faith In Machine Learning. Here's What We Can Do About It.

I talk with 10+ companies a month, and my network has thousands of C-suite-level people.  What I have noticed is growing doubt about data science. Not cost-effective, only profitable for big tech, speculative technology, inconsistent performing teams, and too complex. I've started writing down quotes, and those are the most common.

Big data bust has been said twice, and I wonder if someone is pushing a narrative around that term that hasn't hit the mainstream yet. How much of this will stick or fade away? History can teach us some lessons.

We can learn much about the current cycle from the 2000 dot com bust. There are several overlaps between 2000 and 2022. The Federal Reserve raised interest rates as aggressively as they are now. Tech valuations were in a bubble. Talent was in high demand, and businesses offered high salaries to attract the people they needed.

Dot com companies prioritized long-term revenue and required high initial investments. They were not profitable, but that's not a problem when money is cheap and easy to come by. Most startups and internet businesses were building for 3-5 years, so making a rapid transition to profitability was impossible.

The bubble burst. Several high-profile tech companies failed within the year. Hundreds more startups and small businesses followed. Early-stage tech companies landed at 10%-25% of their peak values in the stock market. Investors lost massive amounts of money. The pain spread from major funds to retirement funds and pensions. As losses mounted, people needed to know what went wrong.

Today's Cycle

Today, we are in the early days of a similar cycle. The dot com bust lasted for almost 3 years. Businesses backed away from their investments in software and the internet. Retail companies paid a heavy price for that mistake.

Many publicly questioned internet technologies and called business models based on the internet, 'Uninvestable.' CEOs blamed speculative bets on emerging technologies for their businesses' struggles. They cut R&D costs to improve the bottom line, and their stock prices were rewarded.

Prospective clients are now interested in cutting data costs and delivering returns faster. Projects are expected to break even in less than a year and generate positive returns over several years. I get asked, 'How do we pull in revenues from our current initiatives and data teams?'

Hiring freezes are in effect. Over the last 3 months, I have gotten a new question, 'How do we know who to let go and who to keep in the data team?' Senior leaders want to understand how to run their data teams leaner with smaller projects and immediate returns.

This is very similar to 2000. By 2001, senior leaders' sentiment on internet technologies was at an all-time low. Businesses that invested into the downturn and made intelligent investments in technology came out the other end of the dot com bust as industry leaders.

There is value to be gained by companies that shift investments to prioritize near-term gains and put smaller, targeted investments in high-value, long-term growth initiatives. Innovation is still required to survive, but senior leaders are unwilling to invest much into speculative projects.

How Do Data Science Teams Make The Transition?

Control the narrative about data initiatives. Start using phrases like value-centric, shortest path to break even, efficient investments, pulling in revenue, margin preserving initiative, competitive advantage initiative, short and long-term growth initiatives, etc.

Make a case for the team. Publishing metrics is critical to establishing the team as a revenue generator and cost saver vs. a cost center. Use AI strategy KPIs to justify and track initiatives. Publish quarterly cost savings and new revenue creation totals. Showcase the percentage of revenue growth and cost savings driven by data, analytics, and data science. Justify every initiative using expected return ranges and celebrate when they hit their targets.

Partner with senior leadership. Create opportunities to support the C-suite and solve their biggest challenges. Success metrics are a good start but getting C-suite buy-in is easier when they see the results. Deliver value to senior leaders. Give them a framework to manage value creation to combat their perception of complexity.

Solve high-pain and high-value problems for other organizations. Data teams need external advocates and promoters. The fastest way to get people on our side is to deliver solutions with obvious, significant value. Spend time with leaders and front-line workers. Find their most significant pain points, the things they hate the most or have been unable to make progress on. Deliver solutions, and they will remember. No one supports cutting back a group working to make their lives easier.

Work with product managers and start delivering revenue. Find features that will provide a competitive advantage or deliver significant customer value. Most businesses cannot deliver solutions to production that are reliable enough to integrate into products. Data teams can drive revenue growth, but the most challenging part is finding opportunities. The current process starts with the business asking, 'What can we do with this data?' Teach product managers and front-line workers to bring opportunities to the data team.

Most of these revolve around the business and people. Data teams make the mistake of spending more time asking for money than asking people what they need and delivering solutions. People, especially senior leaders, need to perceive the data team as problem solvers and value generators.

Pitching Innovation Projects

Intelligent, efficient innovation initiatives are critical to long-term success, even survival. Senior leaders must be able to justify longer-term investments before they sign off. It isn't easy, but possible to do.

Before putting an innovation initiative forward, get senior leaders to understand the need. Use the dot com bust as a case study for what happens to businesses that stop entirely innovating. Retail is a powerful story to tell because it clearly illustrates the point.

Create a track record of delivering before proposing an innovation initiative. The business must trust the team and technology to place more speculative bets.

Once senior leaders are bought in and asking for ideas, then bring them a well-supported case. Wait until they ask for ideas. The first goal is to establish the need for innovation. That part has succeeded once the business comes to the data team for growth. Pitch innovation too early, and senior leaders will tune out.

Select an initiative that can be delivered in phases. A 3 or 6-month delivery cycle is best for innovation initiatives. At the end of each cycle, deliver a part of the larger solution that will generate returns on its own. It will be too easy for senior leaders to end the project if returns don't start within the year.

Talk about the path to break even and profitability. Being close to breaking even can save an initiative from the chopping block. It's critical to remind senior leaders that point is just a few more months away.

Talk about the percentage of total growth driven by the initiative. Some businesses have no other growth driver except for the data team. Having a metric like 70% of growth over the next 3 years is tied to this initiative is powerful.

---

Last November, intelligent people started to call a top to the market and signal that the tech bubble would burst as soon as the Federal Reserve increased interest rates. I released Business Strategy For Data Scientists in December because I remembered the dot com bust.

The most valuable people bridged the gap between R&D and the C-suite. Teams positioned as value generators and selected high-value projects to work on survived the dot com bust intact. The partnership between R&D and business was critical to keeping the team from being seen as a con

We must get ahead of the big data bust and restore trust in data science. During the dot com bust, not all businesses and investors bought into the negative perceptions. Some maintained their investment in technology. Those businesses had teams that knew how to showcase their value.

Teams are used to growing, and cutting back R&D can seem unlikely. I went through the dot com bust. The sentiment in 1999 and early 2000 was very similar. Even as the first wave of cutbacks began, engineers still wouldn't accept that change had come.

For the next 5 years, R&D teams were usually part of even the first phase of cutbacks. Developers were asked to do more with less, which meant longer hours covering multiple roles.

Some data teams will go through a similar cycle. However, the teams who deliver value will sail through this downturn and prepare their businesses to thrive.