Three signals your B2B SaaS marketing budget is funding the wrong half of the funnel

A B2B SaaS founder I talked to last month had a problem she couldn't square with her dashboard. Pipeline was up. Revenue was up. The CFO was happy. Her marketing dashboard was telling her that paid search was 70% of attributed pipeline, retargeting was another 15%, and a long tail of branded direct made up the rest. So she did what the dashboard told her to do — she pushed more budget into paid search and retargeting. Six weeks later, pipeline didn't move. The cost of acquiring each customer went up. The dashboard still said paid search was 70%.

She wasn't wrong about the dashboard. The dashboard was reading the bottom of a funnel that wasn't where the buyers actually came from. The work that had built her pipeline was a podcast appearance she'd done in March, a long-form piece a partner had run a month before that, and three months of dense LinkedIn presence in front of an ideal customer her competitors hadn't found yet. None of that showed up. The dashboard saw the last click. The pipeline saw everything before it. The harder question — the one this post is about — is what budget you're actually funding when the dashboard can only see the half of the buyer's journey that happened inside your funnel.

The two halves, named

The cleanest published frame for this gap splits acquisition activity into two structurally distinct classes. Chris Walker (founder of Refine Labs, now running Passetto) has been naming it for years. Most B2B marketing operations still conflate them.

Demand capture is everything that catches a buyer who already has intent. They're already searching the category. They're already comparing vendors. They're already in-market. The buyer arrives with the intent, and your funnel catches them at the bottom. Capture work looks like: paid search on category keywords, retargeting, gated comparison content, outbound sales teams working a list of target companies, branded search ads. Software attribution can measure capture cleanly because the buyer's last click happens inside the funnel you own.

Demand creation is the opposite. The buyer doesn't yet have intent. The work is the work of creating it. Creation work looks like: podcasts the founder does, original research the company publishes, partner content swaps, peer recommendations, founder presence on LinkedIn, communities the team genuinely lives in. The buyer's journey here happens off your funnel, on surfaces you don't own. His clearest published claim, from years of consulting B2B SaaS budgets: most teams spend roughly 80% of marketing budget on capture and 20% on creation. The ratio that produces durable demand is closer to 60% creation and 40% capture.

The flip isn't a small adjustment. It's an inversion. And most teams haven't done it because the work to do it lives in places the marketing dashboard can't see.

Three signals you're on the wrong side of the flip

These are the questions I'd ask a B2B SaaS founder before suggesting they rebudget anything. None of them are about AI directly. AI didn't cause the gap; AI changes who can afford to act on it. The signals first.

Signal 1. The pipeline keeps hitting on quarters when you slow paid down

The dashboard says paid search is the largest source of pipeline. Then paid gets cut for a quarter — a board reason, a budget freeze, a CFO reshuffle — and pipeline doesn't fall by the proportion the dashboard would have predicted. The buyers find you anyway. The paid search ads were catching demand that other work, upstream, had already created. The dashboard was reading the last click in a journey it never saw the rest of.

The test is mechanical. When paid spend gets cut, does pipeline drop in proportion, or less than the proportion? If less, you're paying for the catch on demand you've already paid to create. Both line items exist; only one of them shows up on the dashboard.

Signal 2. "How did you hear about us?" doesn't match the attribution dashboard

The cleanest first-party test is also the cheapest one to run. Every inbound form has a self-reported attribution field — usually optional, sometimes mandatory. Pull the last six months of those responses. Cluster them. Compare against what the attribution dashboard says about the same six months.

If the dashboard says "60% paid search, 15% retargeting" and the self-reported responses say "30% podcast, 20% recommendation from a colleague, 15% LinkedIn, 10% paid search," you have a measurement gap. Refine Labs published a study putting that gap at about 90% across multiple B2B SaaS companies — the difference between what software attribution claims and what first-party customer-led data reveals. Passetto, his current consultancy, went further and disclosed its own attribution: 97% of revenue self-reported as coming from dark social — the off-platform channels (podcasts, Slack groups, peer recommendations, LinkedIn DMs) that don't show up in any tracked attribution model — and 0% reported by software attribution. Same company. Two answers. Off by a factor of, mathematically, infinity.

If your team has never run this test, the test itself is the thing to install first. Before you flip a single dollar of budget. Before you rebrief any agency. Run the question on the next 90 days of inbound, cluster the responses weekly, and read what comes back. Self-reported attribution has its own biases — covered in the carveouts below.

Signal 3. Your team can name your paid keywords but can't name your top three off-platform demand surfaces

This is the management-side version of the same gap. The growth team can recite the top 20 paid keywords by spend, the cost per click, the conversion rates, the customer value per keyword. They have a dashboard. They have a meeting on it weekly. They have a tool they pay for.

Ask the same team: what are the three off-platform surfaces where our buyers are doing their research? Which podcasts are our ideal customers listening to right now? Which three operators in our category have audiences that overlap with ours? Which Slack groups or peer communities are our buyers in?

If the team can answer the first set in detail and can't answer the second set at all, you have a structural picture of where the work is going. The team is staffed for capture. The dashboard rewards capture. The tools the team has are capture tools. None of which is a problem if your category is actually capture-dominant — short-cycle, transactional, low-consideration. Almost no B2B SaaS category is. Most are six-figure annual contracts with multi-stakeholder buying committees and 6–18-month research windows. The buyer is doing the research somewhere. If your team can't name where, the team isn't equipped to influence the half of the funnel that actually matters.

How we'd rebuild this for AI-native delivery

Walker had the diagnosis right. The install he described pre-dates AI-native delivery — which is why the framework reads as harder to act on than it actually is in 2026. The resourcing math has changed, and that changes who can flip.

The old objection from a small team was always operational. Demand creation looked expensive: a podcast takes hours, original research takes weeks, founder LinkedIn presence takes daily discipline. Capture campaigns spin up in a morning. On hourly cost, creation lost.

Under AI-native delivery that math inverts. The discipline doesn't get easier — the founder still has to show up, the prose still has to be written by a human who knows the market — but the operational tax around the discipline falls by enough that a two-founder team can now run creation at a cadence a 15-person team couldn't have managed two years ago. The flow section below shows what that looks like step by step.

Which means the choice for a B2B SaaS team in 2026 isn't between capture and creation. It's between funding the visible work and funding the work that built the pipeline before any of it became visible.

The carveouts the honest version of this argument has to make

Three.

Before product-market fit, capture still wins — but start creation in parallel anyway. His own counterpoint, sharpened. Below roughly $1M ARR — before there's a thesis, an audience, and time for demand to compound — capture-heavy spend is the right main motion. But don't skip demand creation entirely. Start it in parallel at a low cadence: founder presence on LinkedIn, one or two podcast appearances a quarter, early research drops in front of the audience you're trying to build. Creation compounds slowly. The founder who waits until product-market fit lands to start it will spend the next year catching up to founders who didn't wait. Under AI-native delivery the parallel cost is small enough that there's no good reason to skip the warm-up. The full flip to 60/40 waits until the thesis stabilizes — and the marker for that isn't a revenue number, it's whether the market is starting to come back to a thesis you've put out.

Self-reported attribution is biased, not pure. Hybrid attribution is less wrong than software attribution alone. It is not a clean readout of the truth. Recency effects, branded-search recall, social desirability — all real. The pragmatic version: use self-reported as a directional triangulation against software attribution, not as a replacement, and read the gap as the size of the demand creation engine you're undermeasuring.

The flip isn't a cliff edge. A team going from 80/20 to 60/40 doesn't make the move in a quarter. It makes it over six to eight quarters, with the budget shifting in 5% increments and each shift measured against the self-reported attribution pulse. The teams that try to flip the full ratio in a quarter usually find a pipeline collapse waiting two months later when the capture half drops faster than the creation half compounds.

The flow — first 30 days of demand creation rollout

The shape below is the same one we ship inside a per-client install. It's not a playbook to download; it's a 30-day sequence to read and adapt. Where it breaks is in step 4 — read that part honestly before you commit.

The classic flow. A demand-gen manager builds a paid search campaign on a high-intent category keyword, gates a piece of content behind a form, and counts qualified leads every Friday. The form catches buyers already in-market — the ones the dashboard can see. The much larger group of buyers who are researching the category somewhere off-dashboard never touches the form, and the measurement-gap study cited above is the cleanest published number on how big "much larger" actually is.

The AI-native version.

1. Source the demand signal. Pull self-reported attribution from every inbound form for the last six months. An AI agent clusters the responses and surfaces the 3–5 channels that actually drive buyers. The output is a one-page brief naming which off-platform surfaces are doing the real work. We ship this as a growth/self-reported-attribution-clustering.md skill inside the per-client install.

2. Audit the founder's current presence. Map where the founder shows up — podcasts they've appeared on, LinkedIn presence (cadence and engagement), panels, partner content, Slack groups. Rate each by how concentrated your ideal customers are on that surface.

3. Move 20% from capture to creation. Not the full flip — just the first slice. For the first 30 days, three concrete shipments: one founder podcast appearance (booked or recorded), unbranded research drops on LinkedIn twice weekly, and partner content swaps with 2–3 named voices whose audiences overlap with your ideal customer. This is the on-ramp; the full ratio shift takes 6–8 quarters.

4. Build the daily input loop. A research agent monitors mentions, podcast invitations the founder gets, and LinkedIn comments from buyers who fit the target customer profile — surfaces them every morning as a 5-line "demand signal brief." Skill: growth/daily-demand-signal-brief.md. The founder reads the brief in five minutes, picks one item, and acts on it the same day. The discipline is reading the brief daily, not the brief itself.

5. Measure with the hybrid. Every inbound for the next 90 days asks the self-reported attribution question. The dashboard keeps tracking software attribution. The weekly review reads both, not either.

The closing edge. The channel mix in step 5 updates step 3's rebudget every quarter. If founder podcast appearances are driving 25% of inbound by month three of measurement, that's where next quarter's "ad spend" goes — not as ad spend, but as founder time and partner-content investment.

Where it breaks. Founders who hate being on camera or refuse to post. The flow assumes founder presence as input — if the founder isn't willing to be the channel, demand creation has to come from research artifacts or partner content alone, and the rebudget gets harder. The honest version of "where it breaks" also includes step 2: founders who think they have a presence and don't, or who post weekly to no engagement and treat that as creation work. The signal isn't the activity. The signal is the ideal customer showing up in the comments.

Install note. We ship the shape above as a growth/ skill set inside the per-client install file. The shape is replicable. The install isn't — figuring out which 3–5 channels matter for your brand is week-one work, and it's the part we charge for.

If you're a B2B SaaS founder reading this and you're not sure whether you're on the wrong side of the flip, the cheapest test is signal 2 — pull the self-reported attribution from the last six months and read it against the dashboard. The gap is the size of the demand creation engine you're undermeasuring. The dashboard isn't going to tell you it's there. The buyers will, if you ask them.

The hardest part of the flip is starting. Capture is what the dashboard rewards and what the team already knows how to do. Creation requires a founder willing to show up, a research discipline that doesn't exist yet, and a measurement gap the team has to learn to read. But once the loop starts compounding — the first podcast appearance that drives inbound, the first piece of research that gets cited back to you, the first "someone in my Slack mentioned you" — the question stops being whether to flip and becomes how fast. We rarely see a team get past the on-ramp and retreat to capture-only.