Before Mars tests a new Skittles flavor, it can ask consumers who don't exist.
Before Mars tests a new Skittles flavor, it can ask consumers who don’t exist.
It’s called a synthetic consumer panel: a model built to respond the way real consumer segments respond. How would a mid-30s urban professional receive this concept? A retiree? A kid? Run the concept through the model and you get a directionally accurate answer in a fraction of the time and cost of a full human panel.
Stanislav Rosenberg’s team oversees projects like that at Mars, where he has spent five and a half years as Global Director of Advanced Analytics. He owns the analytics lifecycle end to end: pitch the use case, oversee the build, then make sure the business actually uses it.
He is not a data scientist. That’s the point.
“It’s sort of like a business translator role,” he says. “My folks aren’t data scientists, data engineers by training. We are from business.” His team sits between the stakeholders who need answers and the technical teams who build the tools, and translates each side to the other.
The translator thread runs through his whole career. It started with a rejected plan.
He turned down diplomacy for the same life with more agency.
Stanislav wanted to be a diplomat. Born in Moscow to an American father, he moved to the US as a kid and grew up between countries. He “basically felt I’m not 100 percent at home anywhere,” so a career of postings abroad sounded right.
Then he looked closer. The State Department hands you a position and a posting, and your opinion doesn’t factor into either. He wanted the international life with a say in it. “More agency, essentially.”
So he chose finance, which travels. Frankfurt first. Then London. Then Moscow, working in an investment bank’s office of the president, directly for the CEO and chief strategy officer, when the global financial crisis ripped through the market. Russia got hit hard in 2009. “I thought that was a perfect time to do a two-year MBA.” Half his Wharton class had arrived off Wall Street with the same idea.
After that, the resume looks scattered until you see the pattern. Consulting: strategy and transformation work across industries. LexisNexis: analytics for a legal tech platform serving law firms. BCG: running an industry analytics center of excellence. Mars: AI transformation inside one of the world’s largest consumer companies.
Four industries, one repeated job: standing between the people with the questions and the people with the tools.
Five years ago he would have built it. Today he’d buy it.
Ask him what a project like that consumer panel would look like if he kicked it off now, and he doesn’t describe better technology. He describes a different default.
“Five years ago we had a bias to build everything,” he says. You wanted to own the IP, there weren’t many good options, and the ones that existed were expensive. Two of those three have since reversed. “SaaS solutions are cheaper and better, and thus the sort of ratio of buy versus build has flipped.”
Then the line worth taping to the wall of every enterprise architecture group:
“You can’t build everything ultimately, and you can’t build it well.”
Stanislav Rosenberg, Mars
Cheap agents running on off-the-shelf models push it further still. And what changes with it isn’t just the bill. It’s the shape of the decision. The old path was a twelve-week MVP at a couple hundred thousand dollars, then a call on whether to keep scaling. Now you can stand up a skeleton in days and kill it two weeks in for a fraction of that.
“This democratization of AI essentially is driving down the risk to fail.”
The overlooked part of AI transformation isn’t the technology. It’s the calendar.
Which is where he thinks most companies are quietly getting it wrong.
“One area that’s really overlooked is the process change element,” he says. Organizations adopt the AI transformation mindset and leave the legacy processes running untouched underneath it.
Take an innovation portfolio decision made once a year: what do we launch? If testing a concept now costs a fraction of what it did, once a year is the wrong cadence. You should be testing, and pivoting, far more often. But annual planning cycles don’t move at the speed dashboards do. “Those type of processes usually don’t evolve nearly as quickly as sort of the dashboards.”
The tooling gets faster. The decision rhythm doesn’t. And the rhythm is what actually sets how often a company can change its mind.
He expects jobs like his to come and go.
“AI is now starting to be essentially a tsunami, transforming all the different industries,” Stanislav says, and companies are staffing up business translators and transformation leads to meet it.
Here’s the part most of those leaders won’t say out loud: he expects roles like his to be temporary.
“The department spins up, does some broad transformation, and then essentially winds itself down.”
Stanislav Rosenberg, Mars
Once marketing, HR, sales, and pricing have absorbed AI into how they work, you don’t need a standing transformation department.
That should worry you only if you don’t know where you stand. Which is exactly his advice.
Lead with the “so what,” and don’t wait for perfect data.
The biggest pitfall he sees in technical people pitching the business: they lead with the methodology. It excites them. The business either can’t follow it or doesn’t have time to care. “Go in with the mindset in terms of the so what, or what value the tech piece brings to the business stakeholder.”
And when someone insists the team needs more data before deciding, he reaches for the Pareto principle. Do you really need 100 percent of the data, he asks, when 80 percent drives the same recommendation? Be explicit about where you are and what the delta is. Then decide whether closing it is worth the return, because “the ROI on pretty much everything you do needs to be fairly explicit.”
His closing advice sounds like a summary of his own career. “Figure out where to place yourself on the wave and ride as best you can.” Then the harder half: be able to say plainly what you’re for. “You got to be clear on what value your department, your team, your role is bringing.”
Otherwise, in his words, you’re “quite a bit more exposed” than you think.
This has been a DoGood Member Spotlight.