The old advice was to be everywhere: run every channel, let attribution sort it out later. It predates Google Search getting this good at its job, and it’s cost B2B SaaS teams more than any bad campaign I’ve audited.
Here’s the short version. Every paid channel does one of three jobs: it generates demand, captures demand, or retains it. Pick the channel by the job you need, then check you can actually fund that job — because generating demand costs multiples of what capturing it costs. Google and YouTube can do all three. Microsoft captures. LinkedIn buys precision, not volume. Meta and Reddit buy volume, not precision. OpenAI Ads can generate demand but can’t retarget at all. Run one channel properly before you run two, and know that two is the long-term target — not six.
Why is “add another channel” usually the wrong instinct?
The request almost never starts in the ad account. It starts in a board meeting, or with a founder who saw a competitor’s LinkedIn ad. By the time it reaches us it’s already a conclusion: we should be running LinkedIn too.
What’s striking is that the same teams hold two beliefs at once. In our discovery calls, marketers tell us coverage is the strategy — run all the channels, all the match types, don’t miss anything — and in the same conversation tell us their budget is too small to compete. Both can’t drive the same decision. One of them has to give, and it should be the first one.
None of which means one channel is the answer forever. The argument for a mix is real, and we’ve made it at length in our case for running the mix rather than the channel. This post is about the part that comes first: deciding which channels, in what order, and whether you can afford the job you’re hiring them for.
What are the three jobs a paid channel can do?
Sort every channel by which of these it can do for you, at your budget:
- Capture demand. Someone is already looking for what you sell. You take the click.
- Generate demand. Your buyer isn’t looking yet. You make them aware there’s a problem worth solving, then that you solve it.
- Retain demand. Retargeting and nurture — you keep the audience something else already touched.
The load-bearing difference is cost. Capture scales down; generation doesn’t. Search works on a flexible budget because you’re paying for intent that already exists — a small account can buy a small slice of it and still see a result. Demand generation is a volume-and-quality purchase: you need enough impressions against enough of the right people, often enough, before anything moves. There’s no small version that works. Retention is the cheapest of the three and the least independent — a retargeting pool only exists because something upstream filled it.
Here’s how we’d score the platforms we run. This is our operating view from managing these accounts, not a benchmark table:
| Channel | Generate | Capture | Retain | The catch |
|---|---|---|---|---|
| Google Search | ◐ | ● | ● | Capture is native. Generating demand here needs a real budget |
| YouTube / Demand Gen | ● | ○ | ● | Audiences are vague for B2B — expensive to test into |
| Microsoft | ○ | ● | ◐ | Pairs with Google, works on flexible budgets |
| ● | ◐ | ● | Firmographic precision, expensive clicks, low volume | |
| Meta | ● | ◐ | ● | Creative-led and automated. More noise, worse conversions |
| ● | ○ | ● | Entirely dependent on the right communities existing | |
| OpenAI Ads | ◐ | ◐ | ✕ | Early, and not a retargeting channel |
(● strong · ◐ works with budget or effort · ○ weak · ✕ can’t)
Can the channel reach your buyer at all?
Before any of that matters, ask how common your audience is. If you sell to fraud directors at banks with $500 million or more in revenue, Meta and Reddit are not going to find them cold — those platforms may be worth a retargeting layer and nothing else. If your ICP is e-commerce operators, Meta is a genuinely reasonable place to try. Same channels, opposite answers, and the deciding variable is your buyer, not the platform’s reputation.
OpenAI Ads makes this concrete in a way I didn’t expect. We ran it for a company chasing enterprise customers and it went nowhere. The reason turned out to be structural: since ads arrived in ChatGPT in early 2026, they serve on the Free and Go tiers, and Plus, Pro, Business, Enterprise and Education subscribers don’t see ads at all. If your buyer works somewhere that pays for seats, the channel excludes them by design. That’s not a campaign you can optimize your way out of.
What does a channel actually cost to do that job?
This is where most “channel mix” advice hands you a number with nothing underneath it. The floor isn’t a price list — it’s whichever of three tests is hardest for your situation.
1. The reach test. For firmographic, list-like targeting (LinkedIn, ABM-style buys), you need to cover a known audience at enough frequency to be remembered:
monthly floor = (people you must reach × impressions each per month × CPM) ÷ 1,000
Pull the CPM from the platform’s own forecaster, not a benchmark post. Say you’ve got 1,500 target accounts with four relevant titles each — 6,000 people. Hold 6–8 impressions on each per month and you’re buying about 42,000 impressions. That’s one layer. A three-tier funnel is three of them, plus enough budget in each to optimize against.
And here’s the thing that surprises people: a smaller audience doesn’t cost less, it costs more per impression. You’re holding frequency against a narrow pool, in an auction, against everyone else who wants those same people. That’s why “our audience is too small for LinkedIn” is usually backwards — the audience isn’t the problem, the frequency math is.
2. The learning test. For algorithmic, creative-led channels (Meta, Reddit, Google’s Demand Gen), the floor is set by what the platform needs to optimize at all. Meta’s learning phase documentation puts it at roughly 50 optimization events per week, per ad set. Multiply that out: 50 events a week is about 215 a month, so at a $25 cost per conversion you need around $5,400 a month, and at $50 you need closer to $10,800. That’s not a coincidence — that’s where our $5–10k Meta demand-gen floor comes from. Google is more forgiving but plays the same game; its own guidance is to judge a Target CPA strategy over windows holding at least 30 conversions.
3. The sales test. If a human is closing the loop — and under the floor, a human always is — does the spend produce enough engaged companies per week to be worth a rep’s time? Fewer than about five a week and nobody works the list, the channel produces nothing, and everyone concludes the channel doesn’t work for B2B.
Rough floors, once you run those tests:
| Channel and job | Monthly floor |
|---|---|
| Meta — one demand-gen campaign | $5,000–10,000 |
| Reddit — demand gen | Similar, if the communities exist |
| Meta or Reddit — retargeting only | $1,000–2,000 |
| LinkedIn — demand gen and lead capture | $10,000+ |
| LinkedIn — full three-tier funnel | $15,000–20,000 |
The LinkedIn numbers have a wrinkle worth understanding: $15–20k is the full three-tier funnel, but $10k works if you have either larger audiences or more sales involvement. A bigger audience buys you cheaper reach; a sales team buys you a layer you don’t have to pay for. We break the LinkedIn tiers down properly in our post on matching LinkedIn play to budget — if LinkedIn is the channel you’re weighing, start there.
Which non-search channel should you add first?
For most B2B SaaS selling to mid-market and enterprise, with specific titles to hit: LinkedIn first, then Meta, then Reddit. Two things flip that order. A perfect subreddit leapfrogs everything — if there’s a community where your buyers already argue about the problem you solve, that’s worth more than firmographic filters. And if you sell to businesses with one to ten people, Meta and Reddit win outright on cost.
What you’re really choosing between is precision and volume, and your ICP has already made the choice for you. LinkedIn sells you precision at brutal click prices. Meta and Reddit sell you volume with much weaker firmographic control. Pick the one that matches your buyer’s shape.
Get that mismatch wrong and the channel will still “work” — it just works on the wrong people. We ran Reddit for an e-commerce SaaS that any size business could use, but whose growth targets were squarely medium-to-enterprise. Reddit delivered. We saw a direct influx of free trials and demos, and essentially none of it from the larger companies they wanted. The channel generated demand exactly as advertised. It was pointed at the wrong half of the market, and no amount of bid management was going to fix that.
That’s the failure mode worth internalizing: a channel that’s a bad fit rarely fails loudly. It brings you cheap conversions from people who’ll never close, which looks like success right up until you check what the pipeline did. That’s the same trap as a CPL that looks fine while pipeline stalls.
What if you only have $6,000 a month?
Run search, and nothing else — as far as your category terms will take you. Then put whatever’s left into one channel, chosen on fit. Not three.
What’s left over decides what’s possible. With $5,000 or more, you can fund one demand channel at its floor — pick it on ICP shape and commit to it. With $1,000 to $2,000, retargeting is the only thing that works, and only if something upstream is filling the audience; retargeting a trickle is paying premium prices to nag the same forty people. And below the floor entirely there’s exactly one play: a small audience, as many layers as the money allows, and a weekly list of the companies engaging most, handed to reps who reach out to the titles you’re already targeting. The human closes the loop the budget can’t.
That last one is the only honest way to run a demand channel under its floor. It’s also why “what’s the minimum for LinkedIn” has no single answer — it depends entirely on whether a person is going to work the output.
How do you know a channel is capped?
Mostly, it isn’t. I’m not sure a demand channel is ever truly capped — there’s always another audience, another offer, another creative angle.
Search is the real exception. You can’t force someone searching for something else to convert. Once you’ve worked through category terms, competitor terms and mid-funnel terms, you’re out of demand to capture. Two signals tell you you’re there: search impression share shows how much of the available auction you’re already taking, and cost per opportunity starts climbing as you buy volume. Not cost per demo — cost per opportunity. The distinction matters more the further down the funnel you measure.
Everywhere else, four things cap a channel, and here’s the part worth sitting with: only one of them belongs to the channel.
- Audience size and availability — the channel’s. On demand channels you watch audience penetration and frequency per ad and ad group; when frequency climbs and performance sags, you’ve saturated.
- Your offer — yours. If you have no mid-funnel or top-funnel offer, mid-funnel and top-funnel campaigns cannot work. On any channel. The cap isn’t the platform, it’s that you’ve got nothing to say to someone who isn’t ready to buy.
- Your budget — yours. See the floors above.
- Your patience — yours. LinkedIn can take six months to show its effects, and plenty of demand campaigns are the same. If leadership will pull the plug in eight weeks, you don’t have a channel problem, you have a mandate problem. Solve that first.
There’s a fifth, and it hides inside audience size: your creative production capacity. Larger audiences exhaust more slowly, and a wider variety of ads exhausts them more slowly still. Which means the real ceiling on a demand channel is often how much new material your team can actually make. That’s a staffing answer, not a media-buying one, and nobody wants to hear it.
If not demos, what’s the goal on a cold audience?
Not demos. Cold audiences that sales has never touched aren’t going to commit to a call, and asking them to is how cost per demo goes vertical. I’ve watched clients insist LinkedIn be a demo-capture channel and run cold demo campaigns on it — cost per demo ran up to ten times what other PPC channels delivered. LinkedIn isn’t good at that job. Clicks are expensive, volume is thin, and the algorithm has almost nothing to learn from. You don’t want to be learning on LinkedIn. You want to know what’s working for your sales team, and apply it there.
(The exception: demo capture on LinkedIn works fine when you’ve got a real retargeting pool to work against. Warm is a different job than cold.)
So what goes at the top instead? Offers and assets that solve the problem without requiring a purchase. You give away actual advice about the thing they’re dealing with, earn some trust, and move them down from there. Concretely: how-to docs, skill guides, data sets, FAQs, templates — gated or not. Newsletters are a great pipeline entry point, with the honest caveat that newsletter leads are lower-value on average. And anyone who visits your pricing or case-study pages should be in a retargeting audience by the end of the week.
Structurally, run at least two prongs on any demand channel — a top-funnel campaign and a bottom-funnel one. Three is better: awareness of the problem, one way to solve it, then how your software solves it. With a small audience you can run all three at the same people in different formats, which is a feature, not a compromise. With a bigger audience or budget, add retargeting layers off engagement.
How do you prove a demand channel is working?
The problem with demand channels is that they work by impression and convert off-site. The click-through attribution will always undersell them, and your Google brand campaign will quietly take the credit.
The way we solve it is a geo holdout, and it’s cheap enough that there’s no excuse not to run one:
- Baseline by state. Use server-side tracking to segment conversions by region and build a two-week baseline across all 50 states. (That plumbing is the same plumbing offline conversion tracking needs — if you don’t have it, that’s the first project, not this one.)
- Turn the channel on in six states. Which six barely matters. We used the core Midwest.
- Compare those six to their own prior two weeks — not to the rest of the country.
- Subtract what the other 44 states did over the same window. If the untargeted states rose 20% on their own, you strip at least 20% off your test states’ gain. That’s your external-noise control, and it’s the part that makes the test worth running.
- Read brand search volume, brand conversions and impression share. Optimize the campaign itself on click-through rate and micro-conversions; judge the experiment on what happened to brand.
- Roll out nationally if it holds.
On that e-commerce SaaS — the same account Reddit had misfired on — we ran YouTube and display with a mixed keyword and audience build, watched it over two months, and saw roughly 20% uplift in both brand-lift volume and demo volume in the test regions. Then we took it national.
Be straight about what this is: a directional quasi-experiment, not a clean randomized incrementality study. The states aren’t randomized, the Midwest isn’t the country, and there’s no significance test in that subtraction. It tells you whether to roll out. It doesn’t give you an incrementality coefficient, and anyone selling you one from a setup like this is overselling. What it does do is kill the seasonality objection — that’s what the 44-state control buys you — so you don’t have to wait for a quiet month to learn something.
The takeaway
Run fewer channels than the old playbook told you to. One, done properly, until it genuinely caps. Two for long-term growth. That’s the mix most B2B SaaS companies need, and it’s the opposite of the coverage instinct.
But don’t let anyone tell you a channel is disqualified by the size of its audience. I’ll argue this with other agencies: “LinkedIn audiences can be too small” isn’t true, and neither is “Reddit doesn’t work for B2B” or “Meta doesn’t work for B2B” or “demand gen doesn’t work for B2B.” Those are statements about how much technical work — list building, layering, frequency control, tracking — someone was willing to do before they gave up. The channel was never the decision. The job, the budget and the offer were.
If you’re weighing a second channel right now, our free PPC audit will tell you whether the one you’ve got is actually capped, or just badly fed.