Most of what a company is worth sits in its customers: how many of the right ones you win, the margin each brings, how long they stay, and what it costs to win them. I grow all four with modern data science and AI, and measure every move in orders and margin.
Digital business P&L owned, with full P&L responsibility
A year in media, through a performance function I built
Ad spend managed over my career, B2B and B2C
A data science and analytics function from the ground up
Digital transformation led across the business
Most of a B2B supplier's value comes from what its customers earn it over their lifetime, minus what it costs to win them. I grow that number through four drivers, built on three foundations. The formula below is the one I use with finance. Select a driver to see how I work it.
Customer equity = Customers × (Annual margin per customer × Years retained − Cost to win)
Put in your own numbers. Modest gains on each driver add up, and a valuation multiple turns each added dollar of yearly EBITDA into enterprise value.
Three analyses I run early in an engagement. Each answers one question a CFO would ask, and each ends in a decision about where the money goes.
Matched regions, one with the campaign and one without.
Orders by monthly spend, from a marketing mix model.
Months of contribution margin to recover the cost to win a customer.
Illustrative examples of the analysis. Numbers are not from a client.
Want these three analyses for your business? The free one-page read is where I start.
AI now shapes how buyers find suppliers and how marketing gets done. I use it where it moves one of the four drivers, and I measure it the same way as everything else: in orders and margin.
Engineers and buyers now ask ChatGPT, Gemini and Perplexity which suppliers to consider. I count how often you're named against competitors, then fix the product data, distributor listings and content those answers draw on.
Models on your order history flag accounts that are late to reorder or quietly buying less, so sales calls the right customer in the right week.
Models find the discounts that don't win more business and set guardrails by segment and account size.
My team pairs specialists with AI agents for content, campaign production, reporting and sales follow-up, so more gets shipped each month for the same budget.
How I keep it safe and honest. Your data stays in systems you control, and no model moves budget until it beats a holdout test.
See how AI describes your company →The method holds together because I've owned the P&L, spent the media and built the measurement. These are the rules that came out of doing all three.
Rank accounts and segments by lifetime margin, then grow the ones worth growing.
Move money on holdout tests and mix models, so budgets follow measured lift.
Channels lift each other, so fund the combinations that compound.
One customer view and a lean stack come before any new platform.
Reorders, referrals and product use that bring in the next order and the next customer.
Eight measures I score on day one and review with the board each quarter. Rate each honestly; nothing is sent anywhere.
Strategy and execution from one accountable person. I set the plan, and a team of specialists I've worked with builds and runs it alongside your people and your agency.
A first look at where your customer equity leaks, from public data.
The Revenue and Margin Review on your own data.
My team builds what the plan calls for.
Test, learn and move money on evidence.
Short pieces on questions CEOs and CFOs of B2B suppliers ask about growth and margin. Each comes with a one-page worksheet you can use with your team.
Group five years of orders by the year each customer first bought, and in most established suppliers the older layers carry the business. That changes where growth money should go.
The cohort view takes a day to build from ERP order lines: customer ID, first order date, invoice date and margin. Read three numbers from it. Net revenue retention tells you whether existing accounts grow or shrink. The share of revenue from customers older than three years tells you how much of the business is the installed base. The gap between reorder timing and actual orders tells you where accounts are slipping. Fixing reorder timing and account coverage is usually where I'd start, before adding new-logo spend.
B2B buying runs through distributors, reps and months of research, so the last tracked click before an order is a poor guide to what caused it.
Two tools fix most of this. A holdout test, where matched regions or accounts don't see a campaign, shows what the campaign added. A marketing mix model, fit on two or three years of weekly spend and orders, shows where each channel's returns flatten. Holdouts suit a single big decision; a mix model suits the annual budget. A small team can run its first holdout in a quarter, and a first mix model once order and spend data are joined.
Between list price and the cash you keep sit discounts, rebates, freight and payment terms, each usually approved by a different person.
A price waterfall puts every step from list to pocket price on one page, by segment and by account. It often shows a few points of margin going to accounts that would have bought anyway, and freight or terms given away without anyone deciding to. The fix is usually a set of guardrails on discount levels by segment, with exceptions that need a reason, reviewed monthly alongside the sales pipeline.
Worksheets come with your one-page read.
I've sat in the three chairs that decide whether marketing pays: the one that owns the P&L, the one that spends the media, and the one that measures what it returned.
MilliporeSigma. Owned a $1.5B digital business P&L, about $70M a year in media, and helped take a lab-software product past $5M in annual recurring revenue.
Allstate. Built the performance marketing function, about $350M a year in media, with its mix and attribution models.
Liberty Mutual and MilliporeSigma. Built Liberty Mutual's data science and analytics function from the ground up, and an econometric model at MilliporeSigma.
I write each one myself from public data and send it within 3 business days. Here is what a read looks like.
One-page read · Sample| Engine | Acme | Top competitor |
|---|---|---|
| ChatGPT | 3 of 10 | 8 of 10 |
| Gemini | 2 of 10 | 7 of 10 |
| Perplexity | 4 of 10 | 9 of 10 |
Fix distributor product data for the top 50 SKUs, then re-count AI answers in 30 days.
Bring one question about growth or marketing spend. I'll give you a straight answer and tell you whether I can help.
Marketing that answers to the P&L.