About
Fred Puech, PhD. Economist, data scientist, pricing strategist.
I help companies with complex pricing, from restaurant chains and consumer brands to manufacturers and distributors, grow revenue and margin through better pricing, promotion and revenue growth decisions. The work combines two things that rarely sit in the same person: the econometrics to find out what your prices and promotions actually do, and the business judgment to turn that into a decision a leadership team will act on.

Atlanta, USA · clients worldwide
Background
How I got here
Economics consulting
Measuring policy effects
World Bank, European Commission and national governments.
20 years ago
The first pricing model
Price elasticity by part for an online spare-parts marketplace.
Pricing Solutions
Led the Analytics practice
How decisions really get made, under time pressure.
Since 2021
Keenalytix
Independent. Clients work directly with me.
I trained as an economist and started as a consultant advising major international organizations (World Bank, European Commission) and national governments, where the job was to measure the effect of policies on people’s behavior with imperfect data and to explain the result to people who had to act on it.
I came to pricing by accident, I stayed by conviction. One day, my boss’ best friend, who was the founder and CEO of an online marketplace for spare parts, asked for our help with his pricing. I gave him a price optimization model based on price elasticity, for each part. He tested it and saw the impact right away…and so did I. That was 20 years ago, and that’s when I decided to dedicate myself to pricing full-time.
I then led the Analytics practice inside Pricing Solutions, a boutique consultancy, where I learned how decisions really get made: under time pressure, with partial information, by executives who need a clear answer and the confidence behind it.
Since 2021 I have run Keenalytix as an independent practice, so clients work directly with me from the first call to the final read.
Experience
What 150+ engagements taught me
LESSON 1
The first number is usually wrong, in a predictable direction.
Promotion lift is overstated because pull-forward and cannibalization go uncounted. The damage from a price increase is overstated because traffic, weather and mix moved at the same time. And its benefit is often overstated too: at one industrial manufacturer, a list price increase never showed up on the invoices of several top-selling items, because discounts had quietly taken it back. Read the case
LESSON 2
The money is in the spread, not the average.
Average margins by segment almost always look reasonable. The opportunity sits in the dispersion underneath: one building products manufacturer sold the same item at 15 to 20 different prices, and an online parts distributor had deals losing money after freight and handling that its gross margin reports never showed. Read the case
LESSON 3
Customers react less to price than teams fear, and more to badly targeted price.
A quick-service chain lost over 12% of its guests after increases applied in the wrong places. With better-targeted decisions, it then raised prices for 12 consecutive quarters without losing traffic. Read the case
LESSON 4
Most pricing problems are several decisions working against each other.
At one DTC brand, list price increases, deep discounts, a loyalty program and waived shipping fees each made sense alone; together they gave margin away. At a manufacturer, price lists, discount codes, contracts and manual overrides had grown into more than 220 pricing combinations. In both cases the fix came from treating pricing as one system. Read the case
LESSON 5
Policies don't hold without a reference point.
Written discount rules rarely survive contract-by-contract negotiation. At a logistics provider, a model predicting the right discount from each customer's profile gave sales a defensible starting point, exposed the 15% of contracts out of line, and is now used to price every new customer. Read the case
LESSON 6
Sometimes the data can't answer the question, and saying so is part of the value.
No price variation, too few comparable stores, a product line priced almost entirely by manual override: a model will still return an estimate. Knowing when not to trust it, and what to do instead, prevents expensive decisions made with false confidence.
LESSON 7
A recommendation is only as good as its reception in the room.
The best model is worthless if the CFO can't follow it or the board doesn't trust it. I build the argument along with the analysis.
Principles
How I work
Profitable growth is the only test.
Any analysis or model that won't change a decision is dropped before we start. Any decision that won't grow profit gets a clear no.
The question comes before the data.
Half the work of a sound decision is framing it correctly: what is being decided, against which alternative, and what would change our mind.
Numbers don't speak; they're read.
Every result comes with the trade-offs, the risk and the range around it, in your business's terms.
What a decision did is the input to the next one.
I read the outcome after implementation, so each decision improves the next.
Your team keeps the logic.
Models, assumptions and reasoning stay with you. The goal is that you need me less over time.
Method
On AI
Machine learning is part of my toolkit, used where it outperforms simpler models and validated the same way. Commercial AI tools help me with desk research and finding public data. Recommendations are never AI-generated: each one is mine, built on your data and explained so you can check it.
Summary
At a glance
| Name | Fred Puech, PhD (Economics) |
|---|---|
| Firm | Keenalytix, independent pricing practice, founded 2021 |
| Based in | Atlanta, USA, with a global client base |
| Focus | Pricing, promotion and revenue growth strategy |
| Industries | 20+ industries, B2B and B2C: restaurants and QSR, CPG, alcohol beverages, retail, e-commerce and DTC, industrial manufacturing and distribution, food manufacturing, chemicals, B2B software, associations |
| Clients | Mid-market and enterprise companies with large product, customer and transaction data; PE firms and portfolio companies |
| Experience | 20 years, 150+ engagements |
| Methods | Econometrics, causal inference, Bayesian hierarchical models, machine learning, price and promotion testing |
Free 60-minute working session
Bring one decision.
Bring one pricing or promotion decision to a free 60-minute working session.
