They got part of it right. But on measurement, the part that actually matters, they name the problem and then skip the one method that solves it. That’s a controlled test.

Let’s go through it.

Subscribe now

What does BCG get right?

Three things, and they’re worth saying up front.

Marketing is a growth investment, and cutting it because the P&L looks ugly is a mistake. The real budget question is where the next dollar works hardest, not whether to spend more or less. And measurement belongs inside the company, not with an agency or a tech vendor.

I agree with all three. The trouble starts when BCG explains how to get there.

Does the new math add up?

BCG gives two numbers. A big CPG company that used to make about 500 marketing assets a year now needs five to six million. And a personalized product shot for a beauty brand that cost $4 two years ago now costs four cents.

Do the math on that. Volume goes up about 10,000x. Cost per asset drops 100x. If those numbers describe the same kind of asset, total content spend goes up about 100x.

If they don’t, the article gives you no way to figure out the total at all. Either way, BCG calls it AI collapsing the cost of content and never shows the bill. And the bill is the first thing a CFO asks for.

Does more content mean more growth?

No. More volume means more noise.

Good content compounds. A strong piece keeps paying off long after it ships. A million variations on a weak idea never add up to one strong one. They’re just cheaper to ignore.

And look at what those five million assets really are. Personalized product shots. The same asset, resized and retargeted over and over. That’s production work. If a plan needs 10,000 times the output to stay visible, the output was never what made it work.

Is AI actually cheaper?

Cheaper per token, sure. Cheaper on the invoice, no.

BCG lists token spend as one of the costs going up. In the next breath, it says agentic AI can triple ROI, but it never subtracts that cost. You can’t count the return and leave out the newest line item.

Token prices keep falling, but usage climbs faster. An agent hits a model repeatedly to finish a single task. Gartner puts agentic work at 5 to 30 times the tokens of a normal chatbot exchange. In a McKinsey survey, 93% of companies said they’ve blown past their AI budgets. Uber burned through its whole 2026 AI coding budget in four months.

And today’s prices aren’t the floor. The big AI labs still spend more than they bring in, and prices propped up by investor money don’t stay put.

So build the business case on the full cost per task, at prices that could double. If the ROI only works at today’s rates, it doesn’t work.

Does AI make marketing ROI measurable?

Not the way BCG describes it.

Their method uses AI to pull together fragmented signals and identify which investments drive incremental growth. That’s the same correlation problem marketers have been stuck with for decades. Stitch more signals together, faster, and you still have signals. None of them tells you what would’ve happened if you hadn’t spent the money.

That’s what incrementality needs: a comparison point. You spend in some markets, hold back in matched ones, and measure the gap. No amount of data stitching gives you that gap. You have to create it.

BCG also wants measurement that guides real-time spending. That’s an AI-induced pipe dream. A real test takes weeks to read. You can have real time, or you can have cause and effect, and the article never picks one.

Where’s the proof that AI triples ROI?

The article offers none.

One unnamed entertainment company tripled ROI. One unnamed auto OEM got “100% effectiveness gains” and a CFO who feels more confident: no baseline, no definition of effectiveness, no method. A confident CFO is nice. It isn’t a result.

Then there’s a CPG insights agent that went from 50% to over 90% accuracy writing campaign briefs. Accurate compared to what? A brief isn’t a true-or-false answer, so comparing it to a coin toss means nothing.

The threefold ROI claim shows up twice. Neither instance includes anything you can check.

Do David Protein and Poppi prove it drives growth?

No. They prove that a great product, a strong founder, and the right distribution can grow fast.

Poppi rode Shark Tank and TikTok. David launched with the RXBAR co-founder behind it. BCG calls AI the rocket fuel but gives you no way to separate AI from the product, the founder or the timing. And it doesn’t mention a single AI-native brand that flamed out. That’s survivorship bias.

The breakfast aisle example has the same hole. Insurgent cereal and granola brands took share from the big names, and BCG backs that up by linking its own article on CPG’s lost decade of R&D. That’s a product story. BCG is using it to sell a marketing fix.

What does “human judgment” actually mean?

BCG never says.

The article calls human creativity and judgment the new currency and says companies have to decide which calls stay with people. Then it stops. There’s no list of decisions and no rule for where the line goes. The only evidence that companies are doing this is a survey where about a third of CMOs say they are. That’s what they reported. Nobody observed it.

Judgment isn’t a skill you put on a slide. It’s a set of calls that someone owns, like which market to enter, which channel to cut, or what a test result means. Strategy is what you choose not to do, and someone owns that choice.

Write those decisions down and put an owner on each one. If nobody can do that, “human judgment” is just a phrase covering for a plan nobody made.

Is AI shopping a new threat?

Not really. It’s the same shift with a new interface.

BCG says nearly a third of shoppers now use AI somewhere in their purchase journey, and 13% buy what the AI recommends. It doesn’t say whether that 13% is all shoppers or only those using AI, and that’s a big difference.

We’ve been here before. Search changed how people found products. Then answer boxes cut the clicks, and everyone declared SEO dead. Now AI answers the question, and everyone’s declaring it dead again. SEO isn’t dead. It’s undead. The job never changed. Be the right answer; it’s that simple.

What did change is the buyer. Every buyer now has a research assistant that reads every review. BCG sees this. Its first answer is a stronger narrative, with fixing the product as the follow-up. That’s backwards. If the product doesn’t deliver, a better story just gets caught faster.

AI doesn’t bury bad products. It sends them to the trash compactor faster. Fix the product and the story takes care of itself.

How should upper-funnel spend be measured?

BCG says brand spend should be measured against leading indicators like consideration and mindshare. Those are survey numbers. They tell you people remember the ad. They don’t tell you the ad sold anything.

You can measure brand spend against sales. Run it in test markets, hold it back in matched ones, and look at the difference in revenue.

What actually answers the CFO?

Matched-market testing. Here’s how it works:

1. Group markets that behave alike, based on sales, demographics and seasonality.

2. Pick test and control markets from inside each group.

3. Change one thing in the test markets: spend, channel or message.

4. Measure each test group against its own control.

5. Combine the results to get the real effect, and how confident you can be in it.

This isn’t new. At Wharton’s 2014 customer analytics conference, a Santa Clara University study used this approach to prove paid search’s incremental value for big box retail. Retailers have been doing it for over a decade.

This is where AI earns its keep. It speeds up the grouping, flags bad matches and reads results faster. It makes testing cheaper. It doesn’t replace it.

The Cold Take

BCG makes an argument: AI makes marketing knowable. The evidence is unnamed clients and undefined metrics.

Don’t argue. Test it. Put the plan in front of real markets and let it prove itself or fail. The result is the only number a CFO can take to the board.

Until next time, keep your market tests honest.

Bill
Marketing Systems Guild.
Build it. Prove it. Scale it.

P.S.
Yes, I used AI to help write this. I also used a calculator to check my work and the internet to research my topics. Go figure.

This short video shows another way to plan. It borrows two ideas from aerospace and financial risk management to run a budget through thousands of possible quarters. You walk into the CFO’s office with the odds of breaking even instead of one number and a lot of hope.

That kind of analysis used to take a data science team. AI tools now put it within reach of any marketer.

The figures in the video are an illustrative example, not client results.

Subscribe now

Here’s the problem: every ad platform claims credit for the same sale. When finance asks what marketing actually delivers, no one has a solid answer.

The value: you get a clear Grow, Hold, Trim, or Cut decision for each channel, a direct read on how confident the model is, and a ranked list of channels to test before shifting your budget.

Subscribe now

How does it work?

The model uses ridge regression on your weekly data. It breaks down each week’s sales into base sales, trends, seasons, holidays, events, outside factors like price and the economy, and the impact from each channel.

Two things make media different from everything else in the file:

• Carry-over: Some of this week’s spending keeps having an effect next week. TV’s impact fades over several weeks, while search mostly works right away. The model figures out the rate for each channel.

• Saturation: The first dollars you spend reach the easiest buyers. Each extra dollar brings in fewer new customers. The model finds the point where each channel’s results start to slow down.

Next, the model refits itself 40 times using resampled versions of its own misses, including carry-over and saturation effects. Claude then creates a readout tailored to your role and builds an interactive tool you can open in your browser.

Sample data is synthetic and published with the repository.

Isn’t marketing mix modeling dead?

The old version is outdated. It was built once, gave just one number per channel, didn’t show where it was guessing, and came with a big bill.

This one does three things the old version does not:

• It tells you which answers hold across 40 refits and which ones move.

• It holds a channel’s budget still when the model cannot settle it.

• It takes your lift test results and anchors those channels to them.

Use the model to choose what to test, and use your test results to guide the model.

What do Grow, Hold, Trim, and Cut mean?

Every decision starts by looking at the next dollar spent, based on your gross margin.

Each decision also comes with a confidence level.

The plan never increases spending on a Cut or Trim channel, and never reduces spending on a Grow channel.

What does the readout look like?

Here’s the sample file: 208 weeks, six channels, and a 50% gross margin. Break-even is $2.00 in sales for every $1 spent.

Here are three lines a CEO can read in 30 seconds:

• Media drove $23.4M of $115.8M in sales on $11.6M of spend. Each $1 returned $1.01 in gross profit. Marketing as a whole barely breaks even.

• Same budget, better split: add $1.72M to Paid search, take $1.04M from Paid social and $900K from Online video. Worth $2.35M more in sales a year.

• Gross profit after media costs peaks at about $9.25M, which is 20% below today’s budget.

Why does the next dollar matter more than the average?

The average shows what a channel has done so far. The next dollar shows what extra spending will do.

Radio averages $3.38 per $1, the highest in the file. But its next dollar only returns $1.14, which is below break-even. Radio worked in the past, but spending more won’t help now.

Paid social is the opposite. It averages $0.71 per $1, but its next dollar returns $1.31. It’s at the slow start of an S-curve. Still, the model calls it Cut, because no spending level from a tenth to double today’s amount covers its cost.

Sample data: The curve shows what each spending level delivers. If the curve is above the dashed line, the channel’s gross profit covers its spend.

Why does it ask for my role first?

The math stays the same for everyone, but the answer is tailored to your specific question.

The CEO sees the payback line, budget changes, and the first test. The CFO gets gross profit per dollar, a list of channels below break-even with their ranges, and the held-back miss. Marketing gets the channels to defend, those at risk, and the new mix. The analyst gets the fit, data checks, and a ranked list of fixes.

How does it know when not to trust itself?

• Held-back weeks: The last 13 weeks are never used in the fit. In the sample, the model missed them by less than 2%.

• 40 refits: If a verdict doesn’t hold in at least 80% of them, it never gets marked as Clear.

• Overlap: If two channels’ spending moves together, the model can’t separate them. Both are sent to Test first.

• Flat spend: If a channel’s spending stayed the same every week, it’s left out and named. The model can’t analyze what never changed.

• Results posing as causes: Orders, leads, clicks, and impressions columns are left out. If you use a result as a driver, it absorbs the media effect.

• If the model doesn’t fit—explaining less than 70% of weekly changes or missing more than 15% on held-back weeks—the readout stops. No verdicts, no plan.

How close does it get to the truth?

We built the sample with known answers so the tool can check the model against the truth.

• Five out of six verdicts match the planted answer. The one miss: Paid social is Trim in reality, but Cut in the model. Both recommend spending less.

• The true return falls within the model’s range for four out of six channels.

• Radio is the biggest miss: $3.38 per $1 compared to a true $1.61. Radio is on the test list. If you add a simulated lift test result of $1.61, the model updates radio to $1.68.

That’s the point: a mix model finds patterns in past data, while a lift test proves what actually causes results.

What won’t it do?

• Prove cause: If a lift test and the model disagree, trust the test.

• Move budget: The model won’t shift budget on a channel it can’t settle. Channels marked Test first stay as they are.

• Quote a result outside half to double today’s spend. The model has not seen it.

• Measure a channel: The model can’t measure a channel if its spending never changed.

• Read leads or orders: Convert these to dollars before using them in the model.

What do I need to run it?

• A Claude account on any plan, with code execution turned on.

• You’ll need one CSV file, with one row per week: include the week’s date, sales in dollars, and one spend column for each channel.

• You need at least 52 weeks of data, but 104 or more helps separate seasonal effects from trends.

• Make sure your spending changed over time. Each channel should have at least 8 weeks with spending.

• You’ll also need your gross margin.

• Optional: You can also include price, promotions, events, economic data, and any lift test results.

No connectors or MCP servers are needed. Your files stay within your Claude session.

How was it tested?

There are five test files: one clean file with almost no noise, and a messy export with a dozen planted problems like text numbers, duplicate weeks, blank or negative spend, and bad dates. There’s also one file that’s too short, one monthly, and one daily—each must stop or roll up correctly.

On the clean file, the median error for each channel’s planted return is under 3%, and it gets all six verdicts right. Every count on the messy file matched an answer key based on how the file was made: 39 checks, all passed.

How do I get it?

1. Download the skill from GitHub.

2. In Claude, go to Customize, then Skills, click the plus sign, choose Create skill, and upload your skill.

3. Upload your weekly file and say: “What does each channel actually drive? I’m the CFO. Our gross margin is 45%.” Replace with your own role and margin as needed.

No data handy? Just say, “Run the marketing mix model demo.” Claude will build a sample brand and show the model compared to the planted truth.

The code, sample data, and tests are all public on GitHub: github.com/bbecht/demand-gen-claude-skills. Read every line before you run it.

Until next month, keep your CRM honest,

Bill
Marketing Systems Guild. Turning strangers into clients.

Subscribe now

This is a common scene at budget meetings. Marketing explains which channels bring in leads. Sales says those leads aren’t good enough. The CEO finds a compromise and moves on. Still, no one can say for sure which channel led to closed deals, even though the CRM has the answer.

This month’s skill solves that problem. It’s free, works inside Claude, and keeps your data private in your session.

What is the problem?

This is a common problem: no one can say which channel brings in closed revenue. People might know which channels get leads, clicks, or meetings, but those are different questions. As a result, budgets are often based on the wrong data.

The gap can be seen in three places.

The scoreboard often stops too soon. Channels are judged by top-of-funnel results like form fills, MQLs, and booked calls. A channel might look good on paper but still fail to deliver closed revenue. It’s easy to buy more leads, but much harder to get actual deals.

The data source often has problems, and no one checks it. Some deals don’t have a contact linked. Imports might just be labeled as “Offline.” One person types “linkedin” as the source, another writes “LinkedIn Ads.” Renewals keep the original deal’s source, making old channels look bigger than they are. Lost deals are sometimes deleted instead of marked as lost, so there’s no win rate. Each issue might seem small, but together they let the report show whatever the preparer wants.

Channels aren’t always the right unit to measure. For example, “paid search” isn’t just one thing. Branded search is for people already looking for you. Conquesting search tries to win over competitors’ customers. Non-brand search targets people who don’t know you yet. Each type works differently and needs its own budget. If you lump them together, the bigger one hides the smaller ones.

What does that cost?

It costs money in three ways.

First, you might spend money on channels that seem active but don’t close any deals. If a channel brings in 25% of your deals but only 5% of your revenue, it’s not a real lead engine. It just wastes your reps’ time.

Second, you might not invest enough in the channel that actually closes deals, just because it doesn’t show big numbers on a lead report.

Third, you lose time. Every quarter, the same argument happens because no one has a number everyone trusts. The decision gets pushed off, and putting it off is the most expensive option.

What does the skill do?

You upload a deal export from HubSpot, Salesforce, or any CRM. Claude shows you revenue, win rate, and sales cycle by source, checks if your data is reliable, and creates an interactive report with a budget optimizer.

The math is fixed. A script handles every calculation, so you get the same answer from the same file every time. Claude reports exactly what the numbers say and never makes up or rounds its own results.

Here is the first screen a CEO sees, run on the sample file that ships with the skill.

Who is it for?

The tool is designed for four roles, since each one looks at the data differently. You tell Claude your role. The analysis stays the same, but the answer is tailored to you.

The CEO gets a three-line answer and one suggested budget move. The CRO gets win rate and sales cycle by source, plus advice on which leads should get rep time first. RevOps gets a trust check and a ranked list of CRM fields to fix. Marketing gets the same comparison the CFO will use: each marketing channel versus referrals.

What does it check before it answers?

It checks if your data is good enough for an answer. Before making any recommendations, the skill counts every problem in your file: duplicate deals, the same deal under two IDs, open and closed deals mixed together, blank sources, spelling differences, amounts saved as text, mixed date formats, and deals that closed before they were created.

If more than 20% of your revenue has no source, the report changes for everyone. It starts by highlighting the data problem, because making budget decisions with that file would just be guessing.

It doesn’t try to fill in missing data. If a source is blank, it stays as “(No source).” If a channel has fewer than 10 deals, it just shows the count, like “3 of 7,” instead of a win rate. Three wins out of seven isn’t 43%. It’s just three wins out of seven.

This is the trust check on the messy sample file. Every problem gets counted before any answer.

How deep does it go?

If you add a campaign column to your export, the tool breaks down each channel by what the campaigns were designed to achieve.

It sorts campaigns by keywords in their names. If it can’t read a name, it puts it in “Unclassified.” The number of unclassified campaigns shows you which ones need better names.

Here is what the split looks like on the sample file that ships with the skill. The sample is built to behave like a small B2B firm. It is not client data.

• At the channel level, paid search looks fine: 103 deals, a 28% win rate, right on the company average.

• Inside it, branded search wins 52% of 27 deals. Non-brand wins 17% of 35.

• On LinkedIn, lead gen campaigns produced 42 deals and won 12%. Content promotion produced 19 and won 26%.

Same channel, but very different results. A channel-level report might tell you to keep spending on paid search and lead gen campaigns. By breaking it down, the report lets you see the real differences right away.

Why doesn’t it tell me to spend more on branded search?

That’s because branded search only captures demand that’s already there. People clicking already know your name. Spending more just means paying more for the same buyers. A strong branded search result shows that something else created the demand: referrals, content, events, or word of mouth.

So, the skill never suggests increasing branded search, and the optimizer keeps it fixed by default. It does the same for referrals, offline and unsourced revenue, since you can’t buy those, and for organic search, which takes 6 to 12 months to show results.

What does the budget optimizer do?

It finds the best mix to get the most closed revenue for your budget, or the lowest spend needed to reach your revenue goal.

You enter your monthly spend for each channel or group. The export doesn’t include spend data, so all cost numbers come from what you type in, and they stay in your browser. Each channel or group gets its own row, and the fixed lines are already turned off.

Behind the scenes, it uses a linear program that accounts for diminishing returns. As you spend more on a channel, its results level off. The tenth thousand dollars on LinkedIn won’t perform as well as the first. This prevents the optimizer from putting your whole budget into one channel. Each channel also stays within limits you set, usually 50% to 150% of your current spend, so a small data set can’t eliminate a channel completely.

How honest is the forecast?

It gives you a range, and the range depends on your data. The base case uses each channel’s real win rate. The worst and best cases use the low and high ends of a 90% confidence range. If a channel has 45 deals, the range is narrow. If it has only 9, the range is much wider.

In the sample file, with a sample spend of $26,000 a month, the solver kept total spend the same but shifted $5,650 a month away from LinkedIn lead gen, non-brand search, and brand awareness. That money went into trade shows, conquesting, content promotion, and webinars. The expected result: about $65,000 more closed revenue over 12 months, with a realistic range of $27,000 to $127,000.

It also pointed out the limit. After about $21,500 a month, each extra dollar brought in less than a dollar in closed revenue. That means you’d lose money, no matter your margin. Here’s the full result.

What does it tell me in plain language?

Each run ends with a written recommendation. It explains what to move, what it’s worth, why, how confident to be, when extra budget stops helping, and what to do next. The next step is always the same: move half the budget now, move the rest after 30 days if deal flow is steady, then rerun the report after a full sales cycle. Here’s the exact recommendation from the run above.

If you decide to cut the budget, it tells you how much you’ll save, what you’ll lose in closed revenue, and whether the cut will actually increase your profit at your margin.

What do I need before I run it?

The skill uses whatever your CRM recorded. If your CRM didn’t track where a deal came from, no analysis can fill in that gap. Here’s what you need, from the basics to the full answer.

To run it at all

• Any Claude plan, Free included, with code execution turned on

• A CSV of won and lost deals from the last 12 to 24 months

• Five columns: deal stage, amount, create date, close date, source

• At least 20 closed new-business deals. 50 or more is where the answer gets useful

To trust the answer

• Your CRM’s tracking code on every page of your site

• Every form on the site feeding the CRM

• A contact attached to every deal in HubSpot, because HubSpot sets a deal’s source from its earliest associated contact

• Lead source carried over on conversion in Salesforce

• Offline leads labeled when they are entered

• Deal type set on every deal, new or existing

• Lost deals marked Closed Lost, not deleted

For the channel detail

• UTM tags on every paid link

• One campaign naming pattern, used every time, with the objective as a plain word. The pattern I use: platform-objective-theme-audience-YYYYMM, for example linkedin-leadgen-diagnostic-cmo-202606

• The campaign fields in your export. In HubSpot, that means both Original Traffic Source drill-down columns, because HubSpot stores the campaign name in drill-down 1 for paid search and drill-down 2 for paid social

For the optimizer

• Monthly spend per channel or bucket

• Your gross margin, to judge whether the extra revenue is worth it

No connectors, no integrations, no API keys. It works from the file you upload.

What if my data is a mess?

Run it anyway. That’s the first thing you’ll learn.

The RevOps readout turns every problem into a ranked list of fixes: which field to fix first, so next quarter’s export is clean. In the messy sample file, 37% of revenue had no source, making “(No source)” the biggest revenue line. The right move isn’t to shift the budget. It’s to fix the source field before debating channels again.

How do I get it?

1. Download the skill from GitHub.

2. In Claude, go to Customize > Skills > + > Create skill > Upload a skill.

3. Turn on code execution.

4. Export your deals, won and lost, and upload the CSV.

5. Tell Claude your seat: “Where did our revenue come from? I’m the CEO.”

No export handy? The repo includes three sample files. Try the campaign sample first and watch the channel detail work.

Download Skill

What does this unlock?

A number that both sides of the budget meeting can agree on.

For the CEO, this means the next dollar goes to what actually closes deals, and the quarterly argument is over. For sales, reps spend time on sources that win. For RevOps, there’s a fix list with field names. For marketing, it’s a defense based on closed revenue, the only kind a CFO will respect.

It also turns a vague problem into something specific. Saying “Our attribution is bad” is just a complaint. Saying “37% of revenue has no source because deals are created without a contact attached” is a task you can assign on Monday.

Why give this away?

The hardest part of demand generation isn’t the analysis. It’s building the system that makes the analysis possible: tracking, naming, handoffs, and the discipline to mark a lost deal as lost. That’s what Marketing Systems Guild does.

This skill shows you exactly where you stand. If it reveals you need to build a better system, you know where to find me.

Every month, The Cold Take releases a free skill like this one. The next one is coming soon.

Until next month, keep your CRM honest,

Bill

Marketing Systems Guild. Turning strangers into clients.

Read the job description closely and you can call the outcome before the role is even filled.

It asks for someone to build scalable growth infrastructure in HubSpot: automation, lead scoring, routing, reporting. Two sentences later it asks that same person to launch campaigns, generate qualified pipeline, and own a number. Build the engine and drive the car, at the same time, as the company’s first dedicated growth hire, with nothing waiting on day one but a login and a quota.

I have watched this play out enough times to know the shape of month three. The hire is still inside HubSpot, wiring lifecycle stages and arguing with Sales about what counts as a qualified lead. There is no pipeline yet, because there is no system yet. Leadership starts to wonder, quietly, whether they hired the wrong person.

They didn’t. They sequenced wrong. And that mistake has a name: time to pipeline.

Subscribe now

The real cost is time to pipeline

When one person has to build the system and produce pipeline from it, those two jobs don’t run in parallel. They run in sequence. A demand gen hire takes three to six months to reach full productivity even on a clean runway. Hand them an empty one and the clock starts later still, because before they can run anything they have to build the thing that makes running possible: an ICP that Sales and marketing actually share, a CRM with working lifecycle stages and routing that fires, an agreed definition of a qualified lead, and enough attribution to tell which channels work.

None of that is optional and none of it is fast. So your real time to first meaningful pipeline is not the ninety days the board is picturing. It is two quarters, sometimes three, with a full senior salary running the whole time. That is the problem. Not the person, not the JD. The calendar.

And to be fair to the JD: a great first hire should be a builder. Anyone who hires a pure runner expecting a finished machine deserves the result. The mistake isn’t asking someone to build. It’s asking them to build the foundation and hit a pipeline number on the same clock, then starting that clock the day they walk in to bare ground.

How the BOT model fixes it

The fix is both cheaper and faster than the path almost everyone defaults to: build the system in the background, in parallel with the search, before the hire ever signs.

That is what build, operate, transfer does. The foundation gets stood up while you recruit, on a parallel track, so the two clocks run at once instead of back to back. The system is operated until it is genuinely producing, not until a deck says it should be. Real campaigns, real routing, real pipeline, proven against your numbers. By the time your hire signs, the growth machine is already running.

That collapses time to pipeline from two or three quarters down to roughly day one. And it costs less, because you are not paying a senior full-time salary to do months of setup, and you are not financing a ramp before anything produces.

The leverage is in the overlap. A transfer is not a template dropped on your doorstep. The build runs alongside your new hire at handoff, tuned to your business: your ICP, your sales motion, the way your buyers actually move. They do not inherit a generic engine to reverse-engineer. They inherit a working system already optimized for you, with the reasoning behind every decision handed over alongside the system itself. The artifact transfers, and so does the thinking.

Then it transfers out, on a fixed date agreed up front, and that is the part that protects the leader buying it. A function builder who stays indefinitely is structurally motivated to keep you dependent, because the longer the engagement runs the more it bills. A transfer date inverts that incentive completely. When the deal is build it, prove it, and hand it over on a date, the build only succeeds if the team can run it without help. The exit is not the catch. The exit is the proof.

That is the model end to end, and it is the bridge between bare-bones, founder-led marketing and an actual growth machine: built while you hire the people who will own it, optimized for your business at the handoff, and transferred the moment it works.

The tell, one more time

The companies that win at demand gen did not hire better people. They started their hire on running ground instead of bare ground.

So before you approve that growth marketing req, read it the way your candidate will read it in month three. If it says build and run, you are asking someone to architect a house while they sell it. People survive that. It just costs you two quarters, and often the person.

One question to sit with. If your demand gen hire started Monday, would they spend their first two weeks building the system or running it? If the honest answer is building, there is a faster way to the same place, and it starts before the offer ever goes out.

Subscribe now

The incentive structure of fractional work creates real short-term pressure. A fractional CMO is paid monthly. They are hired to solve a specific problem. They are measured by visible deliverables, not by whether the strategy compounds over three years. When the engagement ends, they leave before the marketing flywheel has time to prove anything.

That mechanism is real. It does produce behavior optimized for quick wins. Jon Miller — co-founder of Marketo, one of the architects of modern B2B demand gen — is right about the mechanism.

The problem he is describing is not a fractional CMO problem. It is a thinking problem. And the thinking problem is just as common in the full-time version.

The average full-time CMO tenure at Fortune 500 companies is 4.3 years, per Spencer Stuart. The average time to meaningful brand compound is longer than that. The full-time CMO who runs borrowed playbooks for four years and then moves on has produced the same short-termism, just with a different org chart.

Short-termism does not come from the staffing model. It comes from what happens inside the model. From treating last quarter’s campaign as this quarter’s strategy. From running the same lead scoring approach because everyone else runs it. From optimizing the funnel for the buyer your analytics described instead of the buyer who actually exists.

Jon Miller built Marketo on a real insight: that B2B buyers do not make decisions the way the industry assumed. That was first principles thinking. The problem is not that he was wrong. The problem is that when Marketo became a best practice, ten thousand marketers copied the form of what he built without copying the thinking underneath it. The form has a shelf life. The thinking compounds.

That is where short-termism actually comes from. Not from the org chart.

If you pulled out your current marketing strategy and asked “how much of this did we reason from first principles, and how much did we adopt because someone credible was doing it” — what would the ratio look like?

Most honest answers land somewhere between 20/80 and 40/60 in favor of borrowed.

Best practices fill that vacuum because they are already packaged, already credible, already ready to present. When a quarter needs defending, the first thing that gets cut is the time to ask whether the underlying assumptions are still true.

The fractional CMO is an easy target because the model is visible and the tenure is short. But the full-time leader who runs borrowed playbooks for four years before moving on has produced the same outcome at greater expense.

The issue was never fractional vs. full-time. The issue is whether whoever is running marketing is reasoning from first principles or running the play they got from a conference in 2022.

Those are different problems. Only one of them gets fixed by changing the staffing model.

The Cold Take is a newsletter about what is actually underneath the assumptions that run B2B marketing. Not best practices. The first principles the tactics are supposed to be built on.

Forward this to whoever just hired a fractional CMO to fix their pipeline

.

We came from completely different worlds. I grew up in Clifton Heights, Pennsylvania, a blue collar town outside Philadelphia where people worked with their hands and didn’t spend a lot of time in the kinds of rooms Mr. Berman moved through. He was the founder of Palisade Capital. We met by chance, which is either the whole story or no part of it, depending on how you think about luck.

He took a liking to me. I couldn’t tell you exactly why. At some point he handed me the keys to his Ferrari and let me drive it.

I always called him Mr. Berman. He called me Billy, the only person outside my family who ever did. I never asked him to stop.

He had a thing he’d say: “Luck is at the intersection of opportunity and preparedness.”

He didn’t say it as inspiration. He said it the way you say something you’ve simply worked out, matter-of-fact, a little impatient with anyone who hadn’t gotten there yet. That’s how most of what he said landed.

I kept that relationship private for most of my career. I was never a name-dropper and I’m not starting now. I’m writing this because he’s gone, and because the thing he told me about luck is one of the most useful frameworks I’ve carried, and it belongs to him.

The standard model of luck treats it as a fixed variable. You have it or you don’t. You were born into the right situation or you weren’t. When someone succeeds, the easy explanation is that they got lucky, which is another way of saying you’ve stopped thinking about it.

Mr. Berman’s formulation does something different. Break luck into its actual parts and you find two things: opportunity and preparedness.

Opportunity isn’t entirely yours to control, but you’re not passive either. You choose which rooms you walk into. You decide where your attention goes. You can move closer to where things are happening or further away; that part is yours.

Preparedness is entirely yours. Nobody builds that for you. Nobody takes it from you.

What most people call luck is what happens when those two things arrive at the same time. Stop treating it as something that visits you. Build the half you own, and stay close to where the other half moves.

The Ferrari story matters for what it actually was. Two people from completely different places, meeting by accident, one of them choosing to extend genuine trust to the other for no obvious reason. There was nothing in it for him.

That’s what I come back to most. Not the car. The fact that he saw something and acted on it without needing a reason to.

Mr. Berman, I think about that intersection more than you know. You were already at it. You helped me find my way there.

Nobody calls me Billy anymore.

Thanks for reading. If this one resonated, pass it along.

“Directional progress.” “Leading indicators.” “Great learnings.” “Strong signal.” “Right trajectory.”

When sales says they missed the number but blames pipeline volatility, the CFO asks what closed. When marketing says it generated significant top-of-funnel signal, the room nods and moves on.

That’s the asymmetry. Marketing’s soft language survives scrutiny in a way other functions’ don’t.

I’ve spent more than 20 years watching this play out across retail, business technology, and financial services. It’s not new. It’s not anyone’s specific fault. And once you start hearing the pattern, you can’t unhear it. Let me try to explain why we got here.

Marketing metrics really are harder to nail down than sales quotas or engineering ship dates. Lead quality is subjective. Attribution is contested. Brand impact takes years to register. When the underlying measurement is this slippery, soft language is what naturally emerges. Over time, the language becomes habit. The habit becomes culture.

That’s the first piece. The second is cohort.

Most marketers shaping today’s vocabulary came up professionally in an era when “optimize the funnel” was a real discipline and “crush Q4” was a real ask. The tactics worked. Then somewhere in the 2010s, the tactics got copied widely enough that the edge faded. The playbooks that worked for the early HubSpot or Marketo era stopped paying off when everyone was running them.

Marketers in that gap had a few paths. Reinvent. Abandon. Hybridize. Or quietly develop a vocabulary that let the old playbook keep running while sounding like it was producing results. The system rewarded that last path more than the others.

Which brings me to the third piece. From what I’ve watched, marketing leaders who insisted on hard metrics got fired faster than the ones who used soft ones. The average CMO tenure is 4.1 years now, the shortest in the C-suite. There’s a survival logic to soft vocabulary. Smart people in any system optimize for the system, and this one rewarded the soft talk.

None of this is anyone’s fault. All of it is everyone’s problem.

You can see the asymmetry most clearly in the QBR itself.

Picture the same meeting with sales and marketing both presenting.

Sales VP: “We hit 87% of plan. The miss was new logos in mid-market. We need to fix outbound or accept a smaller forecast.”

That’s a person taking accountability, even with the macro caveats sales always includes.

Marketing VP, immediately after: “We had a strong quarter on top-of-funnel. MQLs are up 22%, but conversion is a leading indicator we’re still optimizing against. Significant directional progress.”

That’s the same underlying business reality, described in a way that doesn’t commit to anything.

The point isn’t that marketers are dishonest. They aren’t. They’re talking the way the system trained them to talk.

There’s another way to do this. It doesn’t start with a new playbook.

It starts with marketing using the same vocabulary as the rest of the business.

The teams I’ve watched compound over years had this in common, though I’d be cautious about turning observation into rule. When they missed, they said they missed. When they hit, they explained why with specificity. When something was working, they could trace it to revenue. When something wasn’t, they killed it.

Sounds simple. It isn’t. Most marketing departments have spent 20 years building measurement systems that make accountability optional. Changing the vocabulary is the easy part. Changing what you measure, and how, is the actual work.

There’s a framework that cuts through this. Five principles, briefly:

1. Understand how customers actually decide. Not how you assume they do.

2. See your market as a system. Feedback loops shape outcomes more than competitor lists do.

3. Connect marketing to value. If it doesn’t tie to cash flow, it isn’t yet creating value.

4. Strategy is what you choose not to do. Focus beats breadth every time.

5. Ask better questions. The quality of marketing starts with the quality of thinking.

None of these are radical. What’s interesting is how rarely they show up in a quarterly review.

A QBR that used first-principles language would sound like this:

“We invested X last quarter. Roughly 60% we can trace to revenue we expect to close in the next two quarters. The other 40% we can’t yet trace, and three specific tests are running to figure out whether it’s working. Here’s what we’re going to do about the untraced 40%.”

That’s a different conversation than “we generated significant signal across the funnel.” The first is a finance discussion. The second is a vocabulary problem.

One thing to try this week.

Open your most recent results discussion. Could be a formal QBR, a Friday team update, a board deck, an internal Slack thread. Find what summarized how marketing performed.

Read each bullet point and ask: would a salesperson, an engineer, or a CFO use this language to describe their own performance?

If the language sounds plausible coming from those mouths, your measurement culture is healthier than most. If it doesn’t, if it’s full of “signals” and “directionals” and “learnings,” that’s worth knowing. The people writing the deck aren’t doing anything wrong. The system has trained them to write it that way.

What you do about that is a longer conversation. Noticing is half the work. The other half starts with one question you should be able to answer in under ten seconds: what is the one number marketing owns this quarter that the CFO is tracking? If it takes longer than that, the vocabulary problem is the least of it.

That’s the first cold take.

The pattern, for future issues: take one widely-held assumption about marketing and look at what’s actually underneath it.

This first one’s assumption was that marketing has a unique relationship with results, and that’s just the way things are. What’s underneath: a system that trained marketers to talk softly about results, because it’s what the system rewards. The vocabulary is the visible artifact of that system. But that makes it the easiest thing to start noticing.

If you want a more concrete diagnostic of where your marketing actually stands across these five principles and a few others, there’s a free workbook at marketingsystemsguild.com.

Get the free workbook →

See you in two weeks.

— Bill

Here’s what to expect:

Every other week, one long-form essay. Each one takes a widely-held assumption about B2B marketing and looks at what’s actually underneath it. Not contrarian for sport. Considered.

What you won’t get: listicles, sponsored content, “in case you missed it” recap emails, or anyone trying to sell you anything in your inbox.

What you will get: an hour-long read, written by an operator who’s spent 20+ years inside marketing organizations watching what works and what doesn’t.

First issue is on marketing’s strange relationship with underperformance.. It’s worth thinking about why.

See you soon.

— Bill

Founder, Marketing Systems Guild

Subscribe now