Imagine your main channel is dead
Don't rush. You already know which platform it is.
You thought of it before you finished that sentence. You're thinking of the one that would take the quarter down with it, the one nobody says out loud in the planning meeting because everybody's budget depends on it staying alive. You might not even need an effectiveness audit. That flicker you just felt is the most useful diagnostic you'll get all year, and it cost you nothing.
Hold onto it. We're going to use it.
The dead channel test
Go to Google Docs, or god forbid - MS365, find the docx or a pptx with your marketing strategy. Find every place a platform is named — Meta, Google, LinkedIn, TikTok, Amazon, Pinterst (shoutout to people for whom pinterest is the main channel) whatever yours is. Delete the one with the most budget. Delete all of them.
Now check what's left.
If what remains is a coherent argument about who you're for, what you're claiming, and why anyone should believe it, then - Congratulations! you have a strategy and those platforms are how you deliver it. Good. Stop reading, go and do something useful, you'll get a sticker on your way out.
If what remains is a list of tactics with uncomfortably large holes in it, you don't have a strategy. You have a channel dependency with a deck addicted to them.
I've run this in enough rooms to know how it goes. Nobody fails it slightly.
Why do B2B, B2C and DTC all give the same answer?
Ask a B2B team where pipeline comes from and they say LinkedIn.
Ask a DTC brand and they say Meta.
Ask a B2C brand and they say Meta, or TikTok, and then they say "and retail" in a smaller voice. Brave pioneers of unexplored fronteirs would add "retail media!" Bless their souls.
Three disciplines that are absolutely certain they have nothing to learn from each other. B2B is sooooooo different. Different conferences. Different specialists, who will explain at length that B2B is a not just a different animal, it's a mechanism, a system, that DTC economics don't translate, that consumer brand-building doesn't apply when you sell gearboxes or laser-cutters. Different playbooks, different vocabulary, different thought leaders arguing that their segment is the one everybody else misunderstands.
And all three of them just named a single point of failure.
A gearbox manufacturer in Silesia and a sneaker brand in Stockholm have nothing in common. Different buyers, different cycles, different everything. Except that both would be in serious trouble by Q3 if Google forced them to move to invoicing instead of credit card billing, and neither has a plan for that beyond hoping it doesn't happen.
That's the thing about segmentation. It's real at the level of execution and fake at the level of law. Audiences differ — what you say to a procurement committee is not what you say to somebody buying trainers, and anybody who tells you otherwise is selling you a template. But the nice thing, a very pleasant though counter-intuitive thing, is that effectiveness doesn't segment. The mechanics that decide whether your money works are the same in both rooms, and one of them is whether your entire operation is renting its existence from a company that doesn't know you exist.
What happens when a channel actually dies?
The test sounds hypothetical. It isn't. It's just that when a channel dies, it rarely announces itself.
On 1 April this year, Meta moved high-spend accounts off credit cards. If you hadn't switched to monthly invoicing or direct debit in time, your ads paused. Zero impressions. Not a ban, not a policy strike, not anything you did — a billing change. Somebody else's finance department made a decision about payment rails and your quarter went with it. It cause minor heart attacks among our eCommerce clients all of whom where on Amex Diamond cards with fantastic cashback deals.
Take a moment to think about what that means. Not the inconvenience of losing the card points, though plenty of people felt that one. The structural fact: a company you have no relationship with, no account manager at and no recourse against can pause your revenue with an administrative update, and the first you hear about it is an email in a spam folder.
That's not a risk. Risk implies probability. This is a dependency, and dependencies don't have probabilities — they have owners, and the owner isn't you.
If you advertise in Europe you get a second version of the problem, and it's the one nobody plans for. In April 2025 the European Commission fined Meta €200 million for breaching the Digital Markets Act — its "consent or pay" model on Facebook and Instagram broke the rule requiring gatekeepers to obtain consent before combining personal data across services. Meta changed the model. The targeting and measurement available to every advertiser in the EU shifted with it.
Read that chain again, because it's the interesting part. Your campaign performance was altered by a negotiation between a regulator in Brussels and a company in California. You weren't consulted, you weren't notified, and there was no version of "better media buying" that would have protected you. If you build on somebody else's land, the zoning laws are not your department.
We also now have something close to a controlled experiment. TikTok went dark in the US for about fourteen hours in January 2025. Two economists, Donati and Fong, published the results in PNAS last September — a difference-in-differences study tracking roughly 30,000 advertisers through Meta's Ad Library, comparing the US against markets where TikTok stayed up.
Meta ad spend rose 22.4%. CPMs rose 12.1%. Impressions did not rise to match. Advertisers paid more and got the same.
Then the finding that should actually worry you. The demand shift was roughly three times greater for large advertisers than small ones. The small ones largely couldn't follow — many cut spend, and some left Meta entirely once TikTok came back. Fourteen hours was long enough to prove that when a channel dies, your ability to leave is a function of your size.
That's the doctrine of this whole series in one dataset. Attention is distributed unfairly, and the gap has a price.
There's a stranger failure mode as well, where the channel doesn't die and the preparation costs you anyway. The industry spent roughly six years rebuilding measurement around the death of the third-party cookie in Chrome. In July 2024 Google said it wouldn't deprecate them after all. In April 2025 it confirmed it was keeping things as they were. In October 2025 it retired most of the Privacy Sandbox APIs — the replacement technology — and the brand with them. Six years of roadmap, budget and headcount spent on somebody else's announcement.
Planning around one platform's roadmap is a dependency too. It just bills you in wasted quarters instead of paused campaigns.
And these are the polite failures — everybody got notice. In October 2021 Snap missed a quarter and told investors Apple's tracking changes were the reason; the stock fell about 22% in a day. One company's product decision, another company's market cap. Publishers have had it worse and faster: LittleThings had built to 58 million monthly uniques on Facebook, lost about 75% of its traffic in roughly a month after the January 2018 News Feed change, and shut down. The memo the founders sent staff put it plainly — no previous algorithm update had come close to that level of decimation.
The pattern doesn't care what you sell.
Why is your main channel load-bearing in the first place?
You're probably a strange sensation of comfort, agreeableness at this point. Most people stop when they feel it. They read this far, agree, and conclude they need a backup channel.
BZZZZ WRONG. It will cost you money if you just do a Plan B.
Two dependencies isn't a strategy. Youre replacing a petrol sedan with a diesel truck. Sure, a hedge, and an expensive one, because now you're mediocre in two auctions instead of one. Spreading a small budget across five platforms so that no single one can hurt you is how you become invisible everywhere at once, on purpose, with a slide explaining why it was prudent.
Aridor and colleagues published work in Management Science covering more than 4,200 e-commerce firms through Apple's tracking changes. Click-through on conversion-optimised Meta ads fell 37%. The most exposed firms lost somewhere between 8% and 40% of revenue. Stockbrokers jumped out of windows for less.
Those firms moved budget from Meta to Google — the textbook diversification response, the thing every consultant in the world would have told them to do — and they still lost the revenue.
Substitution isn't resilience. If you're only known inside one auction, you arrive at the next one as a stranger, and strangers pay full price.
But why is a single channel so load-bearing?
You didn't choose your main channel. Your best practices, your ai-generated media plans, and the platform-approved attribution model chose it for you, chose badly, and it has been quietly compounding that choice every planning cycle since.
Your main channel isn't distribution. Distribution is a solved problem — there are dozens of ways to put a message in front of a human being and most have existed for a century. Your main channel is a substitute for being known.
Here's the mechanic, literally. Not another exsquisite metaphor. It's in the platforms' own documentation.
Google doesn't sell ad positions to the highest bidder. Ad Rank combines your bid with auction-time ad quality — expected click-through rate, ad relevance, landing page experience — plus context and how competitive the auction is. Google's own help centre states the consequence in one sentence: "Higher quality ads can often lead to lower CPCs. That means you pay less per click when your ads are higher quality."
Meta's auction works the same way from a different direction. Its documentation lists three components of total value: your bid, its estimated action rates — how likely it thinks somebody is to do the thing you want — and ad quality. Two of the three have nothing to do with how much you're willing to pay.
Now read that with the acceptance of your own brand's obscurity in mind.
Two companies bid the same amount for the same impression. The one people recognise gets a better expected click rate, because recognition is precisely what makes a thumb stop. Better expected click rate, better quality signal, lower actual price paid. The brand nobody has heard of pays more for the identical placement — not as a penalty, not because the platform is punishing it, but because the auction is doing exactly what it was designed to do.
That gap has a price. It's sitting in your CAC, it has been for years, and nobody has ever put it on a line item.
So when your CFO asks why brand investment is worth it when you can't attribute it, the answer isn't a lecture about long-term equity. The answer is: we are already paying for our obscurity, every day, in cost per click, and it is the largest untracked expense in this department.
That's what makes the channel load-bearing. You're not addicted to the platform. You're renting recognition by the click — and rent is due every time, forever, and it never builds equity. Stop paying and you stop existing, which is a decent working definition of a dependency.
What is channel dependency actually costing you?
The obscurity tax has a sibling, and it's better documented.
System1 and Peter Field went through the IPA's effectiveness data and found that dull advertising needs 2.6 times the media spend to produce the same very large market share growth as interesting advertising. The same work, done with Adam Morgan's eatbigfish, puts it the other way round: fame-driving ads generate around 6.1 times more share growth than dull, rational ones.
Sit with the compounding. You're paying an obscurity premium in the auction and a dullness premium on the creative, and both are charged through the one channel you can't leave.
Meanwhile the measurement you use to justify all of it is wrong in a specific, quantified direction. Analytic Partners' ROI Genome work, built on mix modelling across hundreds of billions of dollars of spend, found that last-click attribution overstates paid search by roughly 336% and display by roughly 364%, and that about 35 cents of every dollar allocated on those models is wasted.
Notice what those two findings do together. Attribution overvalues the channels that harvest demand somebody else created. That makes those channels look like your best performers. That's why they became your main channel. That's why you're dependent on them.
The dependency isn't an accident of media planning. It's the output of a measurement system that can only see the last thing that happened before the sale.
You didn't choose your main channel. Your best practices, your ai-generated media plans, and the platform-approved attribution model chose it for you, chose badly, and it has been quietly compounding that choice every planning cycle since.
How the test plays out in B2B, B2C and DTC
Same sh-, urgh. Same law, different repair. This is what "no segmentations" actually means — not that B2B and DTC are the same job, but that they fail the same way and have to fix it differently.
B2B. Delete LinkedIn from the plan and what's usually left is a website nobody visits and a sales team cold-calling people who have never heard of the company. The specific B2B delusion is that a long, rational, committee-driven purchase is somehow immune to fame. It's the opposite. When four people have to agree, the recognisable name wins the room, because nobody gets fired for it and nobody has to spend their own credibility defending it.
Most of your market isn't buying this quarter. John Dawes at Ehrenberg-Bass framed this as a principle rather than a number — in categories with long purchase cycles, up to 95% of buyers aren't in the market at any given moment. The committee you want to reach is mostly people who won't need you for two years. Advertising only to those currently in-market, and calling that a strategy, means you've outsourced everything else to whoever they did hear of, back when they weren't looking. Peter Field and Les Binet's B2B work for the B2B Institute put the efficient split at roughly 46% brand building to 54% activation — tilted toward activation compared with the 60/40 they found in consumer categories, and still nothing like the split most B2B budgets actually run.
And if you're B2B and think this is a paid-media problem, look at what you've built on somebody else's API, app store or pricing page. When Reddit repriced its API in 2023, Apollo — a business with millions of users — was looking at roughly $20 million a year, around $2.50 per user per month against a Reddit ARPU a fraction of that. It closed in thirty days. Even HubSpot, the company that sold the world inbound marketing, watched its own blog traffic fall through 2024. It disputes the worst numbers that circulated, and fairly. But at its own conference last year its CMO described the preceding fifteen months as "the pit of despair." The company that wrote the playbook discovered the playbook was a channel bet.
DTC. Delete Meta and there is frequently nothing at all. This is the harshest version of the test, because DTC was born inside a channel — the whole category is a business model that one ad platform made temporarily cheap. The tracking changes of recent years didn't break DTC marketing. They revealed that a lot of DTC brands were arbitrage operations in brand costumes. The ones still standing had something the platform didn't own. Warby Parker cut online advertising and leaned on its shops instead; by 2024 its retail revenue was growing around 20% a year while ecommerce grew about one. The DTC brand that survived the DTC apocalypse did it by opening shops.
B2C. Usually the least dependent and the most complacent about it. Delete the main platform and something survives — retail, search, word of mouth. Which is exactly why the test gets waved through. The failure here isn't collapse, it's slow substitution: brand budget quietly becoming promotional budget over five years, until the only reason anybody buys is price, and now you're in a discount war you can't stop fighting because you trained your customers to wait for it.
Three rooms, three repairs, one law. Whatever survives the deletion is your actual strategy. Everything else was logistics with a nice deck.
Are you dependent, or just concentrated?
Concentration is not the sin. I want to be careful here, because the lazy version of this argument gets small companies killed.
If you're a six-person business selling one product to one kind of buyer, running seven channels is a way to do nothing seven times. Go where your buyers are, go hard, don't apologise. Focus is correct. A small budget spread thin is worse than a small budget spent decisively, and anybody handing a startup "diversify your channel mix" as general advice is giving them a way to fail more slowly.
The sin is concentration with nothing accruing.
Here's the distinction, and it's the whole argument: it matters enormously whether your spending builds an asset you keep or rents access you lose.
You can run one channel and be building — if the people it reaches come away knowing who you are, and some come back later through a door you own. You can run one channel and be renting — if switching it off means demand evaporates inside a fortnight.
Same channel. Same budget. Completely different businesses. And no dashboard on earth will tell you which one you're running, because both look identical in a weekly performance report.
Worth noting who has already diversified, by the way. McKinsey has run its B2B Pulse survey since 2016; buyers now use an average of ten channels across a purchase journey, double the five they used at the start. Your buyers spread out. Most sellers didn't.
The only way to find out which side you're on is to imagine it's dead.
What does channel independence actually look like?
Not a backup platform. Not a "diversified mix" that's the same dependency in a different suit.
It's demand that arrives with your name already attached to it.
People searching for you specifically rather than for your category. An audience you can reach without permission — email, community, physical presence, a sales team with actual relationships. Distribution nobody can switch off. Recognition that makes the auction cheaper for you than for the identical competitor bidding the same money.
The best evidence anyone has for this was an accident. In 2020 Airbnb cut marketing spend from $1.62 billion to $545 million, most of it performance, because the business was collapsing and there was nothing else to cut. Traffic came back to roughly 95% of its 2019 level anyway. So they made it permanent — cutting performance far harder than brand, and holding around 90% of traffic direct or unpaid, where it stayed.
Nobody would run that experiment deliberately. Airbnb ran it because it had no choice, published the result, and then had its CFO defend it to investors. That last detail is the one that matters. It wasn't the marketing department protecting its budget. It was the finance function looking at the numbers and concluding the brand had been doing work the bidding was taking credit for.
When you have that, the channel becomes what it should have been all along: delivery. And delivery is replaceable. If LinkedIn triples its prices, you move. If Meta pauses your account over a billing rail, you're irritated for a week rather than restructuring for a year. The platform stops being the source of your business and goes back to being a road to it.
That's the difference between marketing that survives a platform and marketing that is a platform.
One more thing
The reason this test works isn't that platform collapse is likely. Most of them will be fine. Meta will be here next year, LinkedIn will keep raising prices, Google will keep rewriting the rules of its own auction and calling it an improvement.
The test works because it's the only question that separates what you built from what you rented, and this industry has spent fifteen years making that distinction impossible to see from a dashboard.
So run it. Four minutes, one document, delete the platform names, read what's left.
If there's nothing there, that was never a strategy.
That was a subscription.