How to Create a Demand Generation Plan from Scratch

Someone asks you for a demand generation plan. You open a blank doc. You write the company name. You stop.

The trap is that "plan" gets read three different ways inside the same conversation. To one person it means a channel list. To another it means a 40-slide strategy deck. To a third it means a campaign calendar. None of those is what the CEO actually wants when she asks for "our demand gen plan." She wants a short, opinionated document that explains who you are going after, what you are going to build, and what success will look like in numbers. It is the first artifact we write in any demand generation engagement.

This article is that document, walked through section by section. Eight sections. Each fits on a page. The whole thing fits on twelve. By the end you will have a draft you can argue with on Monday. The advice cuts hard against the default "do everything, run every channel, cover every persona" instinct that most planning docs accumulate over a quarter. That is intentional; the reader who needs this article is usually staring at a doc that already has too much in it, not too little.

What is a demand generation plan, and what is it not?

A demand generation plan is the artifact you circulate to leadership before any campaign goes live. It is shorter than a strategy, narrower than a roadmap, and less granular than a calendar. It locks in four decisions: who the program serves, what it builds, how it will be measured, and what gets left out.

It is not the strategy. The strategy is the stance behind the plan; we have written about the B2B demand generation strategy elsewhere. It is also not the engine. The engine is the operating system that runs once the plan is approved, and we cover how to build a demand generation engine in detail there. The plan is the bridge between the two; it is what gets you the budget and the headcount and the alignment that the engine then turns into pipeline.

Three failure modes show up over and over. The first is channel-first planning, where the doc opens with "we will do SEO, LinkedIn, paid, and email" before anyone has named who the program is for. The second is the tactic-dump, where every named activity in the martech stack lands on the page because cutting any of them feels like an admission of weakness. The third is no decision rules, so the plan never has a defined moment when a campaign gets killed, promoted, or scaled.

If your current plan fits any of those three patterns, the rest of this article gives you a way to rebuild it.

Which eight sections does every plan need?

Eight sections, in this order.

• One ICP and the painful problem you will solve. Name the persona, the trigger, the commercial reality.
• Why us, why now. One paragraph of differentiation, not a tagline.
• The funnel table. Aware, Interested, Engaged, MQL, Won, with numbers attached.
• The content product. One hub, one cornerstone asset, three to five supporting assets.
• The distribution mix. Three to five channels, each with a defined role.
• Capture, scoring, and handover. The matrix, the threshold, the ticket sales actually receives.
• Nurture by segment. Four to six segments, what each gets, how often.
• Operating rhythm, budget split, and decision rules. Cadence of review, percentage by activity, explicit triggers for kill, promote, and scale.

That is the spine. Everything below this is one of those sections, with a short example and a paragraph of "what to leave out."

Section 1. Who is the one ICP, and what painful problem do they have?

Write one ICP, one trigger, and one painful problem in plain language. Half a page. Done.

The mechanic is straightforward. Name the persona by job title and the commercial reality they sit inside. Name the trigger that makes them search for a solution, what Forrester and Gartner have started calling a category entry point. Name the painful problem in the words they would use, not in your category's vocabulary. If you have a customer interview transcript open while you write this section, it will read better.

An example you can paste over. "VP of Marketing at a B2B SaaS company with $5 million to $30 million ARR, in the 90 days after she has inherited a content engine that produces traffic but no pipeline. She has been told by her board to fix attribution, and she has six months before her first quarterly review. The painful problem is that her predecessor sold a CMO on volume metrics, and she now has to defend a smaller, sharper program to the same CFO who wrote the original cheque."

One ICP, not three. The temptation to add a second is real and almost always wrong. Companies that name a buyer sharply on the page are rare; in our analysis of fintech websites for the State of Content Marketing for fintech 2026, only about one in ten sites showed high ICP specificity in their content. That is the position you are starting from. Naming a buyer well is already differentiation.

The commercial case for narrowing is concrete. Demandbase's 2024 ABX Benchmark Report on 600 B2B GTM teams found that ICP-fit accounts win 68% more often than non-ICP accounts, with the lift reaching 2.4x when intent signals are layered in. DealHub's 2024 State of Sales Cycle Report found that ICP-fit deals close 1.7x faster, with the median enterprise cycle running 84 days for ICP-fit versus 143 days for everything else. Those are not marginal differences; that is the cost of fuzziness.

What to leave out. Aspirational ICPs you do not yet serve. Persona documents that list hobbies and reading habits but skip the buying trigger. Anything more granular than one paragraph. "B2B mid-market" is not an ICP; it is a way of avoiding the decision.

Section 2. Why us, why now, in one paragraph

Write a single paragraph that answers three questions. Why should they change? Why now? Why us? Two to three sentences total. If it runs longer, you have a tagline backlog, not a position.

The job of this paragraph is to be repeatable. The sales team should be able to say it on a call. The content team should be able to write to it. The CEO should be able to use it in a board meeting. If three people inside the company would phrase it differently, you do not have differentiation; you have an opinion.

What good looks like. "B2B companies in long, trust-heavy sales hire us when their content engine produces traffic but no qualified pipeline. The reason most fail is that content sits next to the product rather than inside the buying journey. We rebuild it as infrastructure, with a flagship hub, conversion paths, and a sales layer that primes buyers before they ever speak to a rep." Three sentences. Specific enough to be argued with.

The reason this section earns its own page is that most B2B companies do not have it. Wynter's 2025 survey of 100 B2B SaaS marketing leaders at $50 million plus found that only 6% rated their brand as "very distinctive"; 30% admitted to "only slightly distinctive," and 8% were honest enough to call themselves "not at all distinctive." Ninety-four percent of the sample blend in by their own admission.

The buyer notices. In 2024 B2B buying research covering 3,200 enterprise buyers across 15 industries, 67% of buyers said they cannot distinguish between competing brands in their consideration set; that figure is up from 55% in 2023. And the commercial cost of staying generic is measurable. Industry B2B buyer insights research of 1,165 global buyers found a 35% close-rate lift for sellers who articulate differentiated value clearly versus those who cannot.

What to leave out. "Best-in-class." "Platform." "Enterprise-grade." Anything that survives the swap test, where you replace your company name with a competitor's and the sentence still makes sense. If it still makes sense, it is not differentiation.

Section 3. How do you build the funnel table with real numbers?

Draw a seven-column table. Stage, Audience definition, Tactic, What you measure, Conversion to next stage, Cost, Expected monthly output. Fill in five rows: Aware, Interested, Engaged, MQL, Won. Add a sixth row if your business takes demo requests directly; that row is a fast-track exit and skips Engaged.

The numbers do not need to be right on the first pass. They need to exist on the page, so they can be argued with. A funnel table on paper is the cheapest test of whether the plan is grounded. If you cannot back-solve from your target won-deal count to a credible aware-stage volume, your plan is wishful, no matter what it says elsewhere.

A worked example, because abstraction is what makes most funnels useless. Say you sell B2B SaaS at a $20,000 ACV and your year-one target is 24 closed-won deals. Back-solve.

Use FirstPageSage's 2025 B2B SaaS funnel benchmarks as the starting point, since it is the most-cited public dataset and runs from 2017 to 2025: Lead-to-MQL 39%, MQL-to-SQL 38%, SQL-to-Opportunity 42%, Opportunity-to-Closed-Won 37%. To hit 24 wins, you need about 65 opportunities, which back-solves to about 155 SQLs, about 408 MQLs, and about 1,046 leads at top of funnel. At the SMB visitor-to-lead rate of 1.4%, that is roughly 75,000 unique visitors over the year, or about 6,200 per month.

That number is now arguable. If your current monthly visitor count is 1,500, the plan has to either ramp aware-stage volume sixfold, raise conversion rates somewhere mid-funnel, or revise the target. Without the table, that conversation never happens; with it, you can debate which lever to pull.

One thing to watch in the worked example. The MQL-to-SQL row is where most B2B funnels actually bleed. Salesforce's State of Sales 2025 study of 5,500 sales professionals places the B2B SaaS median at 13%, with the top quartile only reaching 18%. The 38% Lead-to-MQL benchmark looks healthy on the page; the 13% downstream conversion is the bottleneck the rest of the plan has to address. Section 6 covers how.

What to leave out. Rows that do not connect to revenue in two hops. If you cannot get from "blog impressions" to "closed deal" in two steps, the row belongs in your tracking dashboard, not your plan. Vanity metric rows are the most common offender; they make the table feel comprehensive and the math feel real, which is the worst possible combination.

The deeper version of this exercise is the one we ran with a fintech client; once we operationalized their hub as the spine of the funnel and tied every asset to a numbered conversion step, their sales cycle compressed by 30 to 35% and the program drove $3 to $4 million in content-influenced pipeline. The funnel table was the first artifact we wrote. Everything else inherited from it.

Section 4. Which content product anchors the plan?

Name one hub, one cornerstone asset, and three to five supporting assets. That is the whole section.

The hub is the destination, the cornerstone is what does the work, and the supporting assets feed and distribute the cornerstone. Most plans skip this section entirely and replace it with "we will publish four blog posts a month." A blog cadence is not a content product. A content product has a name, a thesis, a conversion path, and an owner.

What good looks like. "Renewal Readiness Hub at /renewal. Cornerstone: the Renewal Readiness Diagnostic, an interactive eight-question audit that produces a personalized readiness score and a tailored playbook. Supporting: four problem-led deep articles ranked for renewal-related search, one quarterly webinar with a CIO who has run a renewal, and a downloadable comparison template. Owner: senior content lead. Refresh cadence: cornerstone monthly, articles quarterly."

The case for naming one hub rather than running a content calendar is empirical. HubSpot's Topic Cluster Research found that pillar pages with 10 or more supporting cluster articles generate 3.4x more organic traffic than standalone long-form posts, measured 12 months post-publication on matched-pair samples. The shape of the architecture matters; the volume of words does not.

The scale you need is also concrete. TheStacc's 2026 Topical Authority Impact Study, which analyzed 400-plus SEO campaigns and 12 published case studies, found a 25-article inflection point. Sites that publish 25 to 30 tightly-linked articles in one cluster see a 40% to 70% keyword ranking lift within three to six months; below 25, the lift is inconsistent. Most plans dramatically under-build the cluster, name a hub with three articles attached, then conclude the architecture does not work.

What to leave out. "We will publish more blogs." A list of formats with no thesis attached. Anything you would describe as "thought leadership content" without naming the thesis. The cluster needs a job to do. When the architecture works at scale, the hub stops being a content asset and starts being a product, with a name, a refresh cadence, and an owner.

Section 5. Which three to five channels does the plan name?

Two columns. Channel. Role in the plan. Three to five rows. Each channel serves a specific funnel stage and has an owner. That is the whole table.

The default mix to argue with: SEO, LinkedIn organic, webinars, email, and one paid lever as amplifier rather than engine. Five rows. Each row gets a one-line role statement, the funnel stage it serves, and the name of the person who runs it. If a channel has no owner, it does not go in the plan; it goes in next year's planning doc.

The argument for naming a small number is that the alternative is everywhere at half-strength. The Norwest 2025 B2B Sales & Marketing Benchmark Report, fielded across hundreds of B2B portfolio companies, found that paid digital media at 39% and events and sponsorships at 38% together account for 77% of program budget. That is what concentration looks like in practice. The plans that try to be present on twelve channels are the plans that fund none of them properly.

There is a sharper reason to be selective in 2025 and 2026 than there was a year ago. Channel concentration risk has moved from theoretical to measured. Seer Interactive's 2025 study of 25.1 million organic impressions across 42 organizations found that organic CTR collapses from 1.76% to 0.61% when an AI Overview appears for an informational query, a 61% drop. The B2B impact is sharper still; 73% of B2B websites lost significant traffic between 2024 and 2025, averaging a 34% YoY decline, and AI Overview coverage of B2B-tech queries grew from 36% to 70% in a single year.

This is not a reason to drop organic from the plan; it is a reason to write the distribution row honestly. SEO is now a citation channel as much as a traffic channel. AI-referred visitors convert at 14.2% versus 2.8% for traditional organic, a 5x quality premium, and brands cited in AI Overviews earn 35% more organic clicks than they did before the AI Overview era. The job of the SEO row in your plan is to define what it is doing, ranking, being cited, or both.

What to leave out. Any channel you cannot staff. Any channel that does not have a defined funnel stage. Channels you added because a peer at a different company runs them. If LinkedIn organic does not have an owner who posts three or more times a week with a defined point of view, do not include LinkedIn organic. The line will read better with four channels well-run than five half-run.

This is also where you decide what the plan replaces. In a behavioral-data engagement we ran, the distribution row replaced expensive, episodic event-led growth with continuous inbound. Once events stopped being the engine, every other channel could be sized honestly.

Section 6. How does capture, scoring, and handover actually work?

Three things, on the same page. The lead-scoring matrix. The MQL threshold. The handover ticket. None of them is hard to write; all three are skipped in most plans.

The matrix is a simple two-axis grid. Fit on one axis, signal on the other. Fit is firmographic and ICP match; signal is behavioral. Assign point values. Decide what you give weight to. "VP-level title in a $5 million to $30 million ARR B2B SaaS company" might score +15; "viewed the pricing page twice in seven days" might score +20. Penalize bad fit, not just absence of fit; a free email domain or a company outside your supported geography scores negative.

The MQL threshold is one number. State it. "An MQL is a contact scoring above 60 on the combined matrix who has not received sales outreach in the past 30 days." Print that on the page. Do not write "marketing and sales will collaborate to define the MQL"; that sentence is a deferral disguised as a commitment, and it is the most common reason capture and handover fail.

The handover ticket has five fields. Persona fit, signal observed, enrichment data, timing, likely pain. Five sentences total. That is what sales should receive in the CRM, alongside the contact record, when an MQL crosses the threshold. Not a generic alert. Not a "marketing-sourced lead" tag. Five fields that tell the rep what to do in the next 24 hours.

The reason this section earns more room than the others is that it is where plans collapse. LeadSpot's 2026 B2B Pipeline Trust Report found that marketing-reported MQL-to-SQL rates run 31% on average; sales-reported acceptance rates at comparable companies run 8%. The 23-point gap is the largest single delusion in B2B planning. The same study found that teams that qualify before delivering a lead to sales achieve 28% MQL-to-SQL versus 9% for teams that qualify after; that is a 3.1x pipeline-per-dollar advantage, which is bigger than any channel optimization you could run in the same year.

The market has also moved away from heavyweight scoring models, which is worth noting. Norwest's 2025 benchmark shows that the share of B2B teams using scoring models to define MQLs dropped from 55% in 2023 to 25% in 2025, while reliance on high-intent signals (a demo or sales request) grew from 19% to 30%. The implication for your matrix: keep it simple. A matrix with eight to twelve scoring criteria is more defensible than one with forty.

The handover ticket is also where sales trust gets rebuilt. Only 49% of B2B sales reps trust marketing-sourced leads, per Salesforce's State of Sales 2025 study of 5,500 reps. Sales does not need more leads; sales needs context, and the five-field ticket is the cheapest way to provide it. Speed matters too. The original InsideSales and MIT five-minute rule, still replicated across 300-plus B2B SaaS companies in 2026 data, holds that leads contacted within five minutes are 21 times more likely to qualify than leads contacted at 30 minutes or later. Your plan should state the SLA in hours, not goals.

What to leave out. Eighty-seven-field scoring models nobody maintains. The phrase "we will define MQL with sales." Any handover that does not include the five fields. Routing rules that depend on tools you have not bought yet.

In a recent education-tech engagement, the routing layer was what made the rest of the engine usable. Scored leads streamed to the SDR in the right territory; demo requests routed to the right account executive automatically. The matrix and the threshold mattered, but the handover discipline was what sales could actually feel.

Section 7. How do you write nurture by segment?

Four to six segments. What each segment gets. The cadence. One page chart with a row per segment.

The default segments to start from are not exotic: resource-downloaders, webinar registrants, newsletter subscribers, engaged-but-stalled accounts, cold-but-in-ICP leads, and old CRM contacts. Six rows. For each, name the trigger that places someone in the segment, the content they receive, and the cadence.

The mechanic is segmentation-first, content-second. A piece of content does not nurture; a sequenced delivery of relevant content to a segmented audience nurtures. A weekly newsletter sent to every list you have ever captured is not nurture; it is a habit.

The most useful recent research on this is also the most surprising. Habel, Hartmann, Wiseman, Ahearne, and Vaid published a peer-reviewed study in the Journal of Marketing in 2025 on automated lead nurturing, and the finding cuts against vendor marketing. Automated nurture produces a 23 percentage-point increase in conversion rates, but only under specific conditions: new leads, short sales cycles, and lower-value deals. For high-ticket purchases or industries where buyers run extensive independent research, the benefit declines sharply. The implication for your plan is that nurture is not a single tactic; it is a segment-specific tool. The downloaders segment in a $200,000 ACV business gets a different nurture from the downloaders segment in a $5,000 ACV business, even if you sell the same product.

The practitioner data agrees. Pipeline360's 2025 State of B2B Pipeline Growth survey, fielded to 534 B2B marketers, found that 69% of high performers rate their nurture as "very good or exceptional" against 10% of low performers; that 59-point gap is bigger than the gap on almost any other capability. Seventy-four percent of marketers reported longer sales cycles, and the longer the cycle, the more pressure nurture has to absorb.

The cost case is straightforward. DemandGen Report's 2023 Lead Nurturing Benchmark Study found a 20% increase in sales opportunities and a 33% reduction in cost per opportunity for nurtured leads versus non-nurtured ones. The cadence guidance from the same publisher's 2024 survey, programs rated "excellent" run 11 or more touches across three to six months while "poor" programs run fewer touches and face six-to-nine-month cycles, anchors the cadence column of your segment table.

What to leave out. Thirteen-segment nurture models that look complete on paper and ship for three weeks. Two clean segments shipped and held to a cadence will beat six described and abandoned every time. A segment without a trigger and a cadence is a category, not a segment.

In a hardware-distribution engagement we ran, content-led nurture sequences kept cautious, price-sensitive buyers moving through a long cycle that otherwise stalled at the consideration stage. The cadence was modest. The segmentation was the lever.

Section 8. What goes into operating rhythm, budget split, and decision rules?

Three sub-sections under one heading. Operating rhythm, budget split, and decision rules. One page total.

Operating rhythm. Weekly review of what is in production and what is stuck. Monthly review of the funnel table, with a delta column. Quarterly review of the plan itself, including the ICP. Anything more frequent is theatre; anything less frequent loses the thread.

Budget split. A baseline percentage table across four buckets: create-demand, capture-demand, nurture, and operations. A defensible starting point is 45% / 25% / 20% / 10%. Adjust based on stage. Pre-product-market-fit companies skew create-demand higher. Mature post-fit programs skew capture-demand higher.

Decision rules. Three explicit triggers each for kill, pilot-to-ops promotion, and scale-up. Example. Kill: a campaign that does not produce a measurable contribution within 90 days, or that consistently produces leads that fail the handover ticket criteria, or whose owner has stopped reviewing it. Promote: a pilot that hits its target conversion rate for two consecutive months and has an identified ops owner. Scale: a campaign that exceeds its target by 50% or more for two consecutive months, with clear capacity to handle the next tier of input.

The reason the budget sub-section deserves its own paragraph is that the industry "average" almost no one runs is doing real damage. The Starr Conspiracy synthesized CMO Alliance's Q1 2025 survey and found a bimodal distribution in B2B SaaS create-vs-capture spend; 31% of teams run 70/30 or 60/40 toward creation, 44% run 20/80 or 10/90 toward capture, and only 25% sit between 40/60 and 60/40. The "industry average" describes a population that does not effectively exist; the middle is where companies hedge and get neither outcome. Pick a side and defend it.

Most growth-stage B2B SaaS above $50 million ARR pushes toward 60/40 in favor of creation as brand-search payoff materializes over six to 18 months. Most mid-market companies cannot afford that cash-conversion cycle and run closer to 40/60 with disciplined LinkedIn creation against a narrow ICP. Both work. The middle does not. The actual ratio in market is also more aggressive than the literature recommends; 6sense's 2025 Science of B2B Marketing Spend Report found a real-world brand-to-demand ratio of 30:70, against the Binet and Field 60:40 recommendation, and demand generation was the single category most likely to gain budget in 2025 at +11.7%.

What to leave out. Dashboards you cannot staff. Attribution methodologies that need a data engineer the team does not have. Quarterly business reviews that take three weeks to assemble. The operating rhythm needs to be light enough that it survives a hiring freeze.

What should the plan deliberately leave out?

A real plan is a defended set of choices, which means most of it is what is not on the page.

• More than one ICP. The temptation to add the second is real and always wrong in the first plan.
• More than five channels. If you cannot staff five owners, do not name five channels.
• "We will do ABM" with no named account list. Without the list, ABM is a sentiment, not a plan.
• Blog-post-count goals. Volume targets are how content quality dies.
• Attribution dreams that need tools you do not own. Plan against the tooling you have.
• "We will align with sales" with no shared definition of MQL. Section 6 closed this; this entry is the trapdoor it might fall back through.
• Every tactic on the martech stack. The fact that you bought the seat does not mean it goes in the plan.
• A 40-page strategy deck instead of a 12-page document. The plan is the artifact. The deck is the explanation of the artifact; it is not the artifact.

If the draft you are looking at contains any of those eight, cut them now, before the plan goes to leadership. The shorter, more defended version reads better and survives longer.

What does the one-page version of the plan look like?

The version your CEO actually reads is a single page. Eight sections compressed to two or three lines each, plus the funnel table at the bottom of page one.

A workable layout. Top half: ICP, problem, differentiation paragraph. Middle: the funnel table with five rows and the seven columns from Section 3. Bottom half: the content product (hub plus cornerstone), the channel mix (three to five named rows), the capture rules (MQL threshold plus handover SLA), the nurture segments (four to six named rows), the operating rhythm and budget split.

The argument for a one-page version is not that the 12-page version is unnecessary; both serve different audiences. The one-page version is what leadership scans before approving. The 12-page version is what the team works from. Both should exist by the time the plan ships. If only one of them does, it should be the one-page version, because the longer document is what most planners actually produce, and it is what most leadership teams do not read.

Where do you start this week?

Five days, five commitments. Do not plan past Friday.

• Day 1 and 2: Write Section 1. One paragraph. One ICP. One painful problem. Read it back to a customer if you can.
• Day 3 and 4: Draft the funnel table. Use the FirstPageSage 2025 benchmarks as your starting conversion rates. Back-solve to your year-one target.
• Day 5: Name three to five channels and the owner of each.

That is the spine of the plan. Sections 2, 4, 6, 7, and 8 can be drafted in the second week. The plan does not need to be done to be useful; it needs to exist in defensible form. A plan circulated at 70% completeness gets more useful feedback than a plan held back at 95%, because the gaps in a 70% plan are visible and the gaps in a 95% plan are buried.

The job of a demand generation plan is to make the next 12 months less random. That is achieved by writing fewer things down, with more specificity, and refusing to add the second ICP at the end of the planning cycle. The plan is the artifact. Defend it.