Demand Gen & Performance Marketing Playbook

Demand generation turns marketing spend into pipeline you can repeat: it creates demand among buyers who are not looking yet and captures it from those who are. The work is a cost ceiling, two or three channels, tests sized so the ad platforms can learn, and a budget moved by what each channel caused, not what it claimed.

The two terms in the title overlap. Demand generation is everything that makes buyers want what you sell and then find you: content, events, partnerships, communities, product-led loops and ads. Performance marketing is its paid, measurable part, such as search ads, paid social, retargeting and affiliates, bought against a cost per result. Early-stage startups need both. This playbook is for founders whose leads depend on one channel, whose marketing has many ideas and few repeatable winners, or whose customer acquisition cost is unclear or rising.

It is one of three paid-media playbooks, and the one to read first: it decides which channels to test and whether they caused the pipeline they claim. Splitting the budget across the channels that work is in the Paid Media Budget Allocation Playbook, and building the campaigns inside each ad platform is in the Digital Advertising Playbook.

It starts from a budget you already have; for the total, the 2026 benchmarks show how much to budget for acquisition by stage.

Agree a small set of goals for the next three to six months, in numbers: new customers, qualified opportunities or activated trials, the revenue they support, and the date they matter by, such as the next fundraise.

Then turn the goal into a cost ceiling. Customer acquisition cost (CAC) is the total sales and marketing cost of winning one paying customer, and payback is the number of months of gross profit it takes to earn that cost back. Multiply monthly revenue per customer by your gross margin and by the payback you can afford, and you have the most you can pay to win a customer. Divide that down the funnel and you have the most you can pay for each earlier event, which is the number you judge campaigns on long before a deal closes.

Without CAC history, the ceiling still works as a guardrail: use assumed conversion rates, and replace them with your own after one quarter. The Unit Economics Calculator does the arithmetic.

Last, write down your baseline: the channels that already bring customers (the founders’ network, referrals, communities, outbound or paid), past tests, even anecdotal ones, the assets a campaign can use, such as case studies, a demo or a guide, and whether your analytics and CRM can tell where a customer came from.

Capturing demand means being found by buyers who are already looking: search ads on problem and category terms, review sites, retargeting. Creating demand means reaching buyers who fit but are not looking yet, so that yours is the name they know when they start: content, LinkedIn or YouTube ads, events, podcasts. Capture is easy to measure and limited by how many people search; creation is hard to measure and is where next quarter’s searches come from. My article on demand creation vs. demand capture covers the shift; this step is about the split.

The case for funding creation is that most buyers are not in the market. John Dawes of the Ehrenberg-Bass Institute reasons that if companies change providers about once every five years, about 20% of business buyers are in the market in a year and about 5% in a quarter^[1]. He calls the 95% a heuristic, and your category’s cycle may be shorter or longer.

How much to put into creation has no precise answer. From fewer than 50 B2B cases in the IPA Databank, Les Binet and Peter Field found budgets worked hardest at about 46% brand building and 54% sales activation, a ratio they call a guiding principle; LinkedIn’s B2B Institute, which published the study, rounds it to 50/50^[2]. Brand building is not the same thing as demand creation, so I read it only as a warning: a startup that spends nothing on creation is far from the evidence.

My rule at seed: fund capture until you are buying the searches that exist, and put the rest into creation aimed at the accounts you most want. Judge capture by cost per opportunity. Judge creation by whether target accounts engage and whether demo requests that name you directly rise over the quarter.

Pick two or three primary channels and one or two supporting ones. In my view, two or three channels tested properly beat a little spend on eight. Choose by where your buyer can be reached, and by whether the channel can learn at your budget (Step 4). What each main channel does today:

• Google Search captures people searching for your problem or category, and is the first channel to try if they do. Google has taken AI Max for Search out of beta: it matches ads to searches beyond your keywords, adapts the ad text and can send people to the most relevant page of your site. Google is moving Dynamic Search Ads and campaign-level broad match into it, with the move of Dynamic Search Ads extended to February 2027^[3]. Whichever you use, read the search terms report every week.
• LinkedIn creates demand among buyers defined by job title and company, at auction prices that suit higher deal values. LinkedIn Ads for B2B Startups covers the setup.
• Meta (Facebook and Instagram) creates demand among buyers you can reach as people: consumers, prosumers and small-business owners. Meta has folded its Advantage+ shopping and app campaigns into one setup for sales, app and leads campaigns: a campaign is Advantage+ when Meta’s automation sets the budget across ad sets, the audience and the placements, and limiting any of the three turns it off^[4]. Meta Ads for Startups covers the setup.
• Google Demand Gen creates demand with image and video ads across YouTube, Discover, Gmail, Maps and the Display Network^[5]. It needs more budget than most seed companies can give it (Step 4).
• Performance Max runs one campaign across all of Google’s channels. In 2025 Google added reporting by channel and by search term, and campaign-level negative keywords^[6], but it chases whatever conversion you send it, so wait until opportunities flow back from your CRM (Step 5).
• Retargeting reaches people who have already visited; Retargeting for Startups covers what it really adds.

The supporting channels cost time rather than media: founder-led outbound, content and SEO, partnerships, communities and marketplaces, and product-led loops such as invites and referrals. They make the paid channels convert better.

Give each channel one job and one primary measure. For search, cost per demo request or trial; for paid social, cost per qualified lead, or engagement from target accounts when its job is creation; for outbound, meetings booked per 100 accounts contacted.

Treat the channel mix, the creative, the landing pages and the offers as experiments, not commitments, and keep a backlog of five to ten. For each, write down:

• Hypothesis: what you expect and why.
• Channel and segment: where, and to whom.
• Offer and call to action: what you ask people to do.
• Budget and read date: what you will spend, and the day you will judge it.
• Success threshold: taken from the ceiling in Step 1, such as “keep it if a demo request costs under $600”.

Then check that each test can learn. The platforms’ bidding learns from the conversion you tell it to optimize for, and it needs volume. Meta says an ad set needs about 50 optimization events within seven days of its last significant edit to leave the learning phase^[7]. Google recommends a Demand Gen budget of at least 20 times your target cost per conversion, and at least $100 a day per ad group to avoid a cold start^[8]; its Demand Gen A/B experiments show results only after 50 conversions in each arm^[9]. LinkedIn’s floor is low by comparison: $10 a day, and $100 for a new campaign^[10].

Divide a test’s monthly budget by its cost per conversion and you know how long it needs. A test gives a direction in two to four weeks only if its event is that frequent. If it would take months, you have three options: optimize for a cheaper event earlier in the funnel that still predicts revenue, such as a guide download by the right job titles; put the money into fewer tests; or judge the test on your own CRM data rather than the platform’s.

Three experiments most early-stage B2B companies can start with:

• Google Search, intent capture: problem keywords, two or three ad groups, a landing page for each.
• LinkedIn or Meta, message test: pain-led, outcome-led and story-led ads to the same audience.
• Founder-led outbound: a personal email and call sequence to 50 to 100 accounts that fit. My working rule, a judgment rather than a benchmark: four to six touches over two to three weeks, each with a new reason to reply; the sequence is in Cold Email for Founders.

Launch clean, simple campaigns rather than perfect ones: a few ad groups instead of a granular split, one naming convention for campaigns, ad sets and creatives, and one primary measure per campaign. The landing page, the offer and the follow-up are part of each test, not just the ad.

Tracking has to follow each lead from the click to the form, from the form to a qualified opportunity and from there to a closed deal, so you can see the cost per click, lead, opportunity and customer. That is harder than it was. Apple requires apps to ask before they track people across other companies’ apps and websites^[11] and strips tracking parameters from links shared in Messages and Mail and opened in Safari Private Browsing^[12]. Chrome, by contrast, will keep letting users choose whether to allow third-party cookies^[13]. And in the European Economic Area, Google uses a visitor’s data for conversion measurement only with the ad_user_data consent signal, and for personalized ads only with ad_personalization as well^[14].

So send your own outcomes back to each platform from the CRM. Google’s enhanced conversions for leads matches a hashed email from your form to the deal when it closes^[15]. LinkedIn’s Conversions API lets campaigns optimize for leads that meet your own MQL or SQL definition^[16]. Meta’s conversion leads optimization learns from your CRM stages, but it needs at least 200 leads a month and works only with Meta’s instant forms^[17]. My tracking setup post covers UTMs and events.

Hold a 30-minute review every week. Look at spend and reach per test, cost per lead and per opportunity against the ceiling, and lead quality once sales has seen enough leads. Then make one call per test: scale it where the cost is under the ceiling; fix it where there is a signal but the cost is high, by changing the creative, the offer, the targeting or the bid; and pause it where the cost is well above the ceiling or the leads are the wrong fit.

Between reviews, change little. On Meta a significant edit restarts the learning phase (Step 4), and on Google a new bid strategy, a changed setting or added campaigns start a learning period that usually lasts one or two conversion cycles^[18]. A test that is rebuilt every week never finishes.

The levers worth testing, in the order I would try them:

• Message: problem-, solution- or outcome-led, and which person it speaks to, such as a founder, a marketer or an operator.
• Offer: a trial, demo, audit, checklist, calculator or guide.
• Landing page: one value proposition, one call to action, and only the form fields sales needs.
• Format: a static image, a short video, a carousel or a document; a testimonial, a feature or a comparison.

Test one lever at a time with the platform’s own split test, which keeps the audiences apart: Meta’s divides people into groups that do not overlap and compares two to five creative cells^[19].

Every platform counts the conversions it touched, and several can claim the same one. What you need is the incremental effect: the conversions that would not have happened without the ads. Field experiments show how far apart the two can be. When eBay switched off its ads on searches for its own brand, shoppers clicked the free organic link instead, and the ads showed no measurable short-term benefit^[20]. Across 663 large Facebook experiments, even the better of two methods that work without a control group, using more than 5,000 features per user, could not reliably recover the effect the experiments measured^[21].

The platforms’ own lift tests are out of reach for most seed companies. Google’s Conversion Lift, which covers Search, Performance Max and Demand Gen, needs a campaign budget of at least $5,000 and at least 1,000 observed conversions, and is not available to every account^[22]; Meta’s lift studies go through a Meta representative^[23]. At low volume, three cheaper reads do most of the work: pause a campaign for a few weeks and count total demand, not just paid; ask every new lead how they heard of you, in their own words; and, once volume allows, hold a channel back in matched regions. My guide to incrementality testing for startups walks through the regional test, and the MMM-Lite Starter Kit is the step after it.

Report from the CRM, not the ad platforms, in three layers against the ceiling from Step 1: activity per campaign (impressions, clicks and click-through rate); cost per lead, trial or demo by channel; and the numbers that decide budgets: qualified opportunities, customers, CAC and payback, by channel and blended. Put each incremental read next to the platform’s own number. The Self-Reported Attribution Coder sorts the “How did you hear about us?” answers.

After eight to twelve weeks, put every test in one of three groups, using cost per qualified opportunity from the CRM and the incremental reads from Step 7:

• Scale: at or under the ceiling, with leads or trials sales wants.
• Fix: slightly above the ceiling (I use up to one and a half times it) with promising engagement, or weak creative with strong interest. Change one thing and run it again.
• Kill or park: well above the ceiling, or wrong-fit leads after more than one attempt.

Then reset the budget with a simple rule. Mine, a rule of thumb rather than a published benchmark: 60–70% to the one or two campaigns that proved themselves, 20–30% to new experiments, and 10–20% to evergreen demand generation such as content promotion and retargeting. Raise a winner’s budget in steps and watch its cost per opportunity, because the next dollar usually buys less than the last; the Paid Media Budget Allocation Playbook covers that arithmetic.

Last, write the next 90 days on one page: the channels to prioritize, the next experiments (new audiences, offers and landing pages) and the assets they need, such as a case study, a webinar or a calculator.

What changes by stage?

What mistakes should I avoid?

• Judging channels by the dashboard. Platform-reported conversions overlap and overclaim; check what a pause or a holdout shows (Step 7).
• Optimizing for an event you cannot afford to collect. If 50 conversions take months, choose a cheaper event that predicts revenue (Step 4).
• Spreading a small budget over many channels. Two or three, each with one job (Step 3).
• Sending back form fills, not deals. The bidding learns whatever you teach it (Step 5).
• Editing campaigns every few days. A significant change restarts the learning; change one thing per review (Step 6).
• Cutting creation because last click cannot see it. Most of next quarter’s buyers are not looking yet (Step 2).