Tracy Graham - Investing in Overlooked Businesses – [Invest Like the Best, EP.206]
What this covers
Patrick talks with Tracy Graham, founder of Graham Allen Partners, starting with his path to Notre Dame football and what he took from coach Lou Holtz. Holtz declined to invest in Graham's first company but introduced him to Jay Jordan, who bought 60 percent. That company made about 23 acquisitions and exited in 1999, which became the template for Graham Allen. Along the way Graham discusses growing up in Englewood and why doing business the right way shows most when decisions get tough.
The middle of the conversation is concrete. Graham offers a test for a worthwhile dataset: exclusive access, proprietary, and perishable. His example is a regional oncology database covering about 80 percent of local chemotherapy, which lets the firm forecast drug usage roughly eight weeks ahead, something Pfizer and Eli Lilly want. He also warns that owning data is worthless without a business model, and that AI for clinical decisions has been fool's gold because 90 percent accuracy fails clinically but beats the 60 percent competitors manage on back office cost tasks. Other threads include field sensors in manufacturing, maintenance companies, and sorting customers into profit drivers, drains, and drags, since gross margin hides cost-to-serve.
The contested turn is delivery: whether secondary-market companies want another software tool or a fully serviced solution, and how managed service provider middlemen gatekeep distribution. Patrick frames the larger arc as a coming deployment era for traditional businesses.
Graham Allen Partners argues that lower-middle-market companies in secondary and tertiary US markets are undervalued because of their geography, and that applying data, AI, and transactional-level analytics to these businesses produces the largest margin and growth gains.
- Secondary-market owners lack data talent and technology adoption, so modernization yields the most margin lift there
- Proprietary, exclusive datasets (such as a regional oncology dataset covering about 80% of chemotherapy) are underpriced by sellers who do not use their data
- Entry at roughly 8x LTM EBITDA with cash-producing businesses gives downside protection while data-driven improvements add upside
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
Small companies typically evaluate customers at gross margin or gross profit, but cost-to-serve varies drastically by customer, so only transaction-level data reveals true net income by customer and shows which customers drain margin.
“Most people operate at a gross when they're evaluating. These companies operate at a gross margin level or gross profit level and not a net income level by customer.”
When Lou Holtz refused to invest in Graham's early company, he introduced him to Jay Jordan, who later invested and bought 60% of the business, which was the most important thing Holtz did.
“he introduced me to Jay Jordan. Jay and I had a meeting at 6:00 a.m. in the Morris Inn”
Owning a lot of data is not valuable on its own; without the creativity and a business model to wrap around it, and without useful findings, the data does not produce value, since real insights are hard to come by.
“Very different between having data and insights. As you know, real insights are hard to come by.”
The back office of healthcare providers is still done the way it was 30 years ago, with more regulation and compliance added but no change to the underlying model, making it highly labor-intensive and inefficient.
“It's still being done the way it was done 30 years ago, literally, except there's more regulation, more compliance, more of everything.”
Doing business the right way is most revealed when decisions get tough, since it is easy to act well when things go your way, and durable customer relationships depend on consistent integrity.
“the business lesson for me is you really know what people are all about when decisions get tough”
Because secondary-market companies lack data scientists and engineers, it is an open question whether they want another software tool or a fully serviced solution, and full-service delivery through an intermediary may raise costs; the winning model depends on which customers prefer.
“is tooling what customers really want, or is it the service that they want?”
Helping secondary-market companies adopt technology and data yields the most lift because those businesses lack talent and are run by individuals who have done the same job for 30 years, making the transition hard but the payoff large.
“you get the most margin expansion, the ability to grow those companies well, for a lot of reasons... they don't tend to have the talent and most are run by individuals who have been doing it for 30 years”
Using AI for clinical care decisions looked valuable but has proven fool's gold for now, because clinical models must be far more accurate than administrative ones: 90 percent accuracy is worthless clinically, whereas 90 percent beats the 60 percent competitors achieve on back-office cost tasks.
“if we're at 90%, we're competing with 60%. The clinical side of it, if you're at 90%, it's worthless.”
Maintenance companies that service manufacturers' machines are attractive because connected machines would let them predict servicing needs, such as oil changes, and use machines more efficiently, which current labor-intensive servicing cannot do; the financial profile looks promising.
“as those machines are outfitted with connectivity to Wi-Fi... we can be far more efficient in servicing those machines”
Labor-intensive businesses that sell intelligence and ideas, such as insurance brokerages, are ripe for better technology in the same way hedge funds use technology to augment very smart people.
“What they're selling is a lot of intelligence and ideas... not unlike a hedge fund. I mean you use technology to augment really, really smart people.”
Growing up in Englewood developed skills like intuition, creativity, figuring things out, and constant awareness, which became innate and valuable in evaluating companies and structuring win-win deals.
“Some of those things that people would not think about as assets or skills that you develop in those types of environments, intuition is an example of that, creativity, figuring things out. Always being aware is important.”
A dataset is worth pursuing if it can be obtained with exclusive access, is proprietary in some way, and is perishable, and if models run on it can yield insights valuable and proprietary to others.
“Can we get exclusive access to it? Is it proprietary in any way? And is it perishable?”
Over the next couple of decades, many very traditional businesses will catch up in technology and data, applying best practices to modernize their operations and significantly improve their economic models, an era Patrick calls the deployment era of technology.
“what my friend, Jerry Newman, calls the deployment era of technology. Meaning that we're going to see tons of what I'll call very traditional businesses catch up in the technology and data spheres”
Software distribution into secondary markets is controlled by managed service providers acting as middlemen, since large vendors such as Amazon or Salesforce rarely sell directly in these regions, making MSPs the real distribution gatekeepers.
“we have never seen anyone from Amazon come to sell technology, Salesforce, or any of those companies. There's always a middle person”
Chris Murphy's offer to fly Tracy to Naples and Jay Jordan's trip to the Hamptons gave a young person exposure to a different world, which Tracy believes changed his family's trajectory across generations.
“I think it's a metaphor for just giving a kid access and letting him see the world differently than they've ever seen the world and the impact that it could make on a kid”
Lower-middle-market sellers place little value on their own data because they do not use it, and they often lack data scientists, which creates a mispricing opportunity for buyers who can exploit the data.
“in the lower-middle market space, there's very little value by the seller that's put on data because they don't use it.”
Through ownership of or client ties to all the regional oncology groups, Graham Allen's database captures about 80 percent of the region's chemotherapy, including patient stage, drug use, outcomes, cost, insurer, and physician schedules, giving it an exclusive proprietary dataset.
“So we have probably about I would say 80% of the region's chemotherapy is running through our database. That's a proprietary dataset.”
Because the oncology dataset links physician schedules to patients, the firm can forecast drug usage about eight weeks ahead, and pharma companies such as Pfizer and Eli Lilly want those regional sales forecasts and real-time prescribing-choice data.
“We know every patient that's going to receive chemotherapy for the next let's call it 8 weeks or so.”
Lower-middle-market companies in secondary and tertiary markets are generally undervalued in many cases simply because of their location or geography, since people give less value to those companies.
“we believe that lower-middle market companies and specifically those in secondary and tertiary markets are generally undervalued in many cases simply because of their location or their geography”
The most useful manufacturing data is collected on how products perform in the field through sensors, which lets engineers improve products, but lower-middle-market manufacturers do not capture this well, making data acquisition a heavy lift requiring sensors and supplier integrations.
“collecting data on the use of a product in the field and having that data, analyzing that data in a way that engineers would have the ability to improve that product.”
Helping lower-middle-market companies in secondary markets adopt technology and now data gives the most lift initially, producing the most margin expansion and the ability to grow well.
“you can get the most lift by helping those companies really start to use technology and now data.”
Tracy's first company, backed by Jay Jordan, which sold 60 percent to Jordan, did 23 or so acquisitions and exited in 1999, teaching him how to find, integrate, and build midsize and lower-middle-market companies, which became the foundation for Graham Allen.
“We went on to do 23 or so acquisitions, build up the company, exited in '99.”
Companies typically use only 10 to 20 percent of the capabilities of the software they buy yet pay for 100 percent, making software one of the most underutilized assets on the balance sheet.
“companies use 10%, 20% of the capabilities of the software, but they pay for 100% of it.”
Companies should sort customers into profit drivers to spend the most time with, profit drains, and profit drags, since a large group of customers is often quietly dragging net margin down while being treated the same as everyone else.
“You have your profit drivers, which are the ones you want to spend the most of your time with. You have your profit drains and then your profit drags.”
Smaller manufacturers have very little data and highly manual operations, sometimes with a receptionist handling billing, invoicing, and pricing negotiation, so the first step is building the capability to bring the data together.
“very little data, highly manual, a lot of people in the backroom.”
Traditional Silicon Valley software companies use direct sales for their largest customers and channel partner programs for everyone else, selling through managed service providers who handle implementation and customization.
“for every other customer, we're going to build a channel partner program and we're going to go out and effectively we're going to sell software through managed service providers.”
Over the next 10 to 20 years, small secondary-market companies face one of two outcomes: they modernize with data and AI to improve gross margins, fund sales, and grow faster, or larger competitors use AI to reach into secondary markets and outcompete the slow adopters.
“The opportunity is they modernize. They leverage data and AI to improve gross margins and then spur additional investment in sales”
Graham Allen acquired the back office of an oncology group, a labor-intensive process still run as it was 30 years ago, and modernized it, significantly raising margins and growth.
“It was the back office of an oncology group. We have, in effect, modernized that entire process, margins up significantly, growth up significantly.”
Graham Allen believes its alpha generation engine is its ability to use data and AI to help lower-middle-market companies improve margins and grow faster.
“Through that, we believe that's the alpha generation engine, which is our ability to use data and AI to help companies do that.”
Pharma companies want to know which drugs physicians chose over alternatives in near real time, so they can educate doctors on efficacy, and the dataset can supply that externally.
“when they don't choose their drug, what drug they did choose. They want to know that and they want to get that information in almost real-time”
The core business principle is to do right, do your best, and treat other people the way you want to be treated, a principle drawn from Lou Holtz that Tracy says has kept customers and partners loyal for decades, including Notre Dame as a customer and investor.
“I would have to go back to the foundation of just doing right, doing the best you can, and treating other people the way you want to be treated.”
Lou Holtz's leadership worked because he simplified a meaningful message, delivered it with passion so players believed him, and never changed it, which kept diverse 17- to 22-year-olds focused on something bigger than themselves.
“the message never changed... his ability to deliver it with passion and feeling and you really believed that he believed what he was telling you.”
Tracy's first company burned through an 80,000 dollar loan from Jerome Bettis within six months because he underestimated how far that capital would go, a lesson in cash-planning humility.
“That was my first entry into entrepreneurship which quickly led to me running out of money because I thought $80,000 was the world at that time.”
You really know what people are about when decisions get tough, since it is easy to do the right thing when everything is going your way and profits are high.
“It's easy when it's going your way. So if you have a huge profit... But the business lesson for me is you really know what people are all about when decisions get tough.”
Principles must be simplified and digestible for people with diverse backgrounds so the whole team can rally around a set of core principles, because complicated or profound-sounding ideas do not get adopted.
“Lou Holtz was able to make it simple. Don't complicate it. It doesn't have to be something that sounds profound. It has to be digestible”
Tracy named Graham Allen Partners after childhood friend Kenneth Allen, whose family could not pay a $20 entrance exam fee for St. Rita High School, after which their lives diverged, and who was killed in April 1991.
“Kenneth Allen is a childhood friend of mine that I named the company after”
There is no secret sauce to valuing a proprietary dataset; the task is estimating how much margin expansion access to the data can produce, which requires complicated modeling.
“What you're trying to figure out is how much margin expansion you think you can get as a result of having access to the data.”
Buying businesses that already produce cash at a moderate entry multiple provides air cover, so the technology and data upside is an added return rather than the basis of the investment.
“with an entry price that just provides more air cover because you've got something already producing cash.”
Graham Allen typically pays roughly 8 times LTM EBITDA, pro forma to about 6.5 to 7 times forward, rarely exceeding 8 times, though each deal is evaluated independently.
“I would say generally we pay somewhere in the 8 times LTM and then it's going to be 6.5 times forward... we typically don't pay greater than 8 times.”
Notre Dame and Lou Holtz continued reaching out to players over decades, a practice of never leaving your team that Tracy sees as central to his life.
“never leaving your team. I think the University of Notre Dame and specifically Coach Holtz continues to reach out to players”
The typical Midwest manufacturer in this market makes one or two products with one or two plants, generates roughly 50 to 100 million dollars in revenue and 5 to 15 million dollars in EBITDA, and has not adopted a scale mindset.
“They are two plants maximum, one to two plants. They're going to be somewhere between $50 and $100 million in revenue, somewhere between probably $5 and $15 million in EBITDA.”
AI allows a firm to analyze each customer individually, which is virtually impossible when a human does the analysis, so building this into the data platform is more efficient and effective.
“that's virtually impossible if you're going to have a human doing all the analysis”
Chris Murphy, chairman of 1st Source Bank, took Tracy on his plane to Naples, Florida for seven days with his family when Tracy was 22 or 23 and had waited a long time for the meeting, and the relationship has lasted nearly 25 years.
“We got on his plane and we flew to Naples, Florida. I spent 7 days with him and his family.”
Apple's own stores and Tesla's direct sales show companies choosing to control the customer experience end to end instead of relying on intermediaries, a model Tracy compares against intermediated software distribution.
“Apple decided, 'Look, we're going to build our own Apple stores. We're going to sell it direct.'”
Within the firm's ideation sessions, the team debates whether the real value is the model or the data, with the view that exclusive regional data gives a better chance to build a good model, while others argue competitors can recreate the data.
“What's the real value? Is it the model or the data?”
Through the technology companies Tracy acquired, which served lower-middle-market firms in secondary markets, he gained hundreds of conversations with owners and built an understanding of their culture and why they did or did not invest in technology, which motivated his later firm.
“I got to know the culture, the feeling, how they operated and how they thought about the world, why they did or didn't invest in technology.”
The most important lesson Lou Holtz taught is to do right, do the best you can, and treat others as you want to be treated, and to get people focused on something bigger than themselves, which Graham applies to leading his firm.
“Do right. Do the best you can, and treat other people the way you want to be treated.”
Lou Holtz told Tracy's mother he might never play football at Notre Dame but would earn a degree, calling it the best four-year investment for the next 40 years of his life.
“He may never play football at the University of Notre Dame, but he'll get a degree and it'll be the best 4-year investment that he makes for the next 40 years of his life.”
Customers paying identical prices for the same managed service can differ drastically in cost, for example one calling the call center 50 times versus two times, and standard pricing models typically ignore that difference.
“They're all paying the same prices, except one of those companies calls your call center 50 times and the other one calls 2 times.”
Tracy's first venture, CBD Networks, aimed to put the whole phone book online with a search engine in 1996, but few consumers had internet access and dial-up modems were too slow, so the company pivoted to providing internet access itself.
“people didn't have internet access yet so there was no way they were going to do any searching on the internet... we had pivoted already to providing internet access.”
A large part of the firm's portfolio is a managed IT services company, and when comparing customers, companies receiving the same service at the same price can differ dramatically in support costs.
“a fairly large part of our company is a managed services company that does IT for a number of different companies in the area.”
Lou Holtz declined to invest in Tracy's early company, saying he does not invest in businesses, but introduced him to Jay Jordan, who invested about a year later and was the first deal.
“he did the next best thing. In fact, I think one of the most important things he did was he introduced me to Jay Jordan.”
Most current AI applied to marketing simply helps target customers better, and that is now common across firms; the real gain comes from clean data and a deeper understanding of customers, not from targeting alone.
“basically everybody's doing the same thing. It's trying to help you target customers better.”
Over the next five years, people will start to understand the value of data in lower-middle-market companies, which will end the current mispricing.
“Now, Patrick, that'll change. Over the next 5 years, people will start to understand the value of data.”
No one has yet found the answer to how AI should be used in manufacturing, but the most promising near-term use cases are improving product quality, customer experience, and customer service, where many manufacturers already collect call-center data.
“I don't think anybody has the answer here, so I'll start with that. No one has the answer.”
A family-owned, single-plant manufacturer with about $35 million in annual revenue, a 25% EBITDA margin, and a product sold to U-Haul and others is a typical strong target that wants to grow but lacks scale mindset.
“They do about 25% EBITDA margins, one plant in this case about $35 million a year in revenue.”