The Self-Serve B2B Buyer: How Decisions Form Before Sales

In late 2023, I watched a $400K SaaS deal close where the buyer’s first conversation with the sales team was the negotiation call. They had read the product documentation. Attended a recorded webinar. Watched three customer testimonial videos on the vendor’s YouTube channel. Built an internal business case in a Notion doc that they shared with three department heads. The sales rep touched the account twice in 90 days, both times in response to outbound emails the buyer initiated. By the time anyone at the vendor knew the deal was real, it was a procurement conversation.

That deal was the moment the Growleads thesis crystallized for me. The buying behavior the data had been describing for years had become operational: the B2B buyer makes the decision, then announces it. Sales is invited to the closing meeting, not the discovery one.

Most B2B revenue leaders treat this as a marketing problem. Produce more content. Run more ads. Build a better website. Those responses are not wrong, but they miss what the data actually says. The shift is not from “informed buyer” to “more informed buyer.” It is from “buyer reachable through a sales motion” to “buyer who has already decided when sales finds out.” That requires a different architecture, not just more content.

This piece is a slow read. It is for the revenue leader who has noticed that close rates are flat or declining despite better marketing, and is wondering whether the playbook is still the right one.

What “self-serve” actually means in B2B

Self-serve is a misleading term. It suggests the buyer is doing things alone. The buyer is not alone. They are working with a buying committee, often six to ten internal stakeholders (Gartner, 2024), and the committee is consulting peers, industry communities, and AI tools to assemble a decision.

What “self-serve” actually means is that the seller is not in the room. The buying motion has internalized within the buying organization. The vendor’s job is to be findable, credible, and stick-to-able when the buyer surfaces, not to run a discovery call.

This is a structural shift, not a cyclical one. Three forces drive it.

The first is the cost of the seller’s time. SDR costs have risen 30 percent or more since 2020, while connect rates have fallen by similar amounts. A seller-led journey is more expensive to run and produces less signal per dollar than it did five years ago. Buyers who can self-serve know this and increasingly prefer to.

The second is the abundance of credible content. Product documentation, customer reviews on G2 and TrustRadius, YouTube case studies, peer communities like Pavilion and RevGenius, and increasingly AI-mediated comparison content. A buyer can build a confident point of view on a $200K to $500K B2B purchase without ever speaking to the vendor’s sales team. The information asymmetry that used to require a sales discovery call has collapsed.

The third is the rise of AI-mediated buying. Tools like Perplexity, ChatGPT search, and category-specific comparison platforms now sit between the buyer and the vendor’s marketing site. A buyer asking ChatGPT “what should I look for in a B2B intent data provider” gets a synthesized answer from twenty sources. The vendor’s sales narrative is one of the twenty, weighted by citations the vendor cannot directly control.

These forces compound. A more expensive sales motion meeting a more capable self-serve infrastructure produces a buyer who decides before sales contact almost by default.

The 73% number and what it really says

Gartner reported in 2024 that B2B buyers spend only 17 percent of their total purchase journey time meeting with potential suppliers, and when buyers are choosing among multiple vendors, that 17 percent is split across all of them (Gartner, 2024). The rest of the time is split between independent research (27 percent), buying group meetings (22 percent), and other activities. The widely-cited “73 percent of the decision happens before sales contact” number is a derivation from this data.

The number is right. The interpretation usually is not.

The standard interpretation is “the buyer has already decided 73 percent of what they want before they call us.” That is the marketing-content reading: produce more content to influence the 73 percent. It treats the dark funnel as a measurement problem.

A more useful interpretation is operational. The buyer has built 73 percent of the decision using inputs that the vendor can shape but cannot observe. The vendor sees the 17 percent of the journey that happens with them. They see almost none of the 83 percent that happens elsewhere. The architectural question is not how to instrument the dark funnel; it is how to operate effectively when the most important inputs are unobservable.

This shift is the entire reason demand intelligence exists as a category. The infrastructure has to be able to act on signal trails that buyers leave incidentally as they self-serve, not on the conversations they would have had with a sales team. The full operating model is in our demand intelligence framework cornerstone. The high-level summary: instead of trying to instrument the buyer, instrument the signals the buyer’s environment produces while they decide.

Three operational shifts the data implies

The 73 percent number, taken seriously, implies three changes to how a B2B revenue function operates. Most companies have made one of them. Few have made all three.

Shift one: from sales-discovers to sales-confirms

The traditional sales motion treats discovery as the point at which the seller learns about the buyer’s need. In a self-serve world, the buyer has already discovered their own need. Discovery becomes a confirmation step, not an information-gathering step.

This changes what an effective discovery call looks like. The seller is no longer asking “tell me about your business and your pain points.” The buyer wears that conversation as a tax. The seller is now confirming a hypothesis: “based on the signals we observed before this call, our read is that you are doing X, considering Y, and have constraint Z. Where are we wrong.” The conversation moves from information extraction to hypothesis testing. It also takes 30 minutes instead of 60.

Most sales teams have not made this shift. They were trained on a discovery model that no longer matches buyer behavior. The result is calls that buyers describe as “wasting their time” because the seller is asking what they could have learned from the buyer’s public LinkedIn or the buying committee’s earnings call.

Shift two: from outbound-as-prospecting to outbound-as-pattern-matching

Traditional outbound starts from a target list and prospects into it. Volume is the lever; persistence is the discipline. This works when buyers are reachable through a sales motion. It does not work when buyers are deciding without sales motion involvement.

The shift is to outbound that pattern-matches buying signals across the entire ICP universe and surfaces the accounts that are actively self-serving right now. Not “is this account on our target list” but “is this account exhibiting the signal pattern of a buying committee assembling a decision.” Our signal-based outbound guide walks the operating mechanics.

We tried to outbound the dark funnel ourselves in early 2024 using classic SDR methodology. Eleven weeks of work. The data we collected was not interpretable because the prospecting motion was the wrong instrument for the question. The accounts that were actively buying were not the ones reacting to outbound; they were the ones who had already decided and would surface on their own timeline. The outbound effort just produced noise.

Shift three: from credit attribution to signal attribution

In a self-serve world, marketing-sourced revenue percentage stops being a useful metric. The buyer touched twenty marketing assets across the dark funnel. Assigning credit to one of them produces a number, but the number says nothing about which marketing investment actually moved the decision.

The replacement is signal attribution: which marketing-surfaced signals correlated with closed-won revenue, regardless of which specific touch was first or last. The full math is in our marketing-sourced revenue breakdown. The shift, in one sentence, is from “did we get the credit” to “which signals predict pipeline.”

These three shifts together describe what an architecture optimized for self-serve buying actually looks like. None of them is about producing more content. All of them are about reading the buyer’s environment more accurately than competitors do.

Why the dark funnel resists instrumentation

There is a tempting response to all of this that goes “if the buyer is self-serving, we just need to instrument more carefully.” Better web analytics. More tracking pixels. Heavier intent data. AI-mediated journey reconstruction.

This response is mostly wrong, and the reason it is wrong is structural.

The dark funnel is dark by design, not by accident. Buyers prefer it. They learn faster, share information internally without sales-driven framing, and avoid the cost of a sales motion. The infrastructure they use (peer communities, AI tools, anonymous reviews) protects their privacy on purpose. Even when individual buyers are willing to be tracked, the buying committee they sit on includes people who are not.

The math is also against instrumentation. Suppose a vendor invests in a perfect-fidelity tracking infrastructure. They can now see which accounts visit which pages. They still cannot see which Slack messages the buying committee exchanges, which Notion docs they pass around, which peer calls they take, which AI prompts they run. The trackable surface is a small fraction of the decision-formation surface, and getting smaller.

The architectural response is not to track more. It is to act on the signals that emit naturally when buyers self-serve, even when the underlying journey is invisible. A new VP hire at a target account is observable on LinkedIn. A funding event is observable in news feeds. A topic surge is observable in third-party intent data. A stack change is observable in technographic data. None of these requires instrumenting the buyer’s actual decision process; all of them correlate with active buying behavior in our 200+ campaign dataset since 2023.

This is what signal-based outbound and the buying signal database playbook describe at the operating level. The signals are the proxy for the decision. They are imperfect, but they are also legal, observable, and predictive.

What this means for the next quarter

If you are a revenue leader reading this, the first question is not what to change in your marketing or sales tactics. It is whether your operating model has internalized that the buyer’s decision is forming somewhere you cannot see, and that your job is to be findable and credible when they surface, not to run a discovery process.

A practical test. Pull the last twenty closed-won deals over $100K. For each, ask the AE three questions. When did the buyer first decide they had a problem worth solving. When did the buyer first identify your company as a candidate solution. When did your team first know the deal was real. The gap between question 2 and question 3 is your visibility window. If the gap is more than 30 days for most deals, you are operating without architectural awareness of the dark funnel.

The corrective work is at the revenue architecture level, not the campaign level. Layer 1 (signal detection) needs to capture the signals that buyers emit while self-serving. Layer 2 (scoring) needs to weight those signals against closed-won history. Layer 3 (orchestration) needs to route the right play to the right account at the right moment, knowing the buyer is mostly invisible. The action layer (Layer 4) inherits its quality from those upstream layers.

If you want to see what an architecture optimized for self-serve buyers looks like for your specific revenue mix, our demand intelligence service starts with a buyer-journey audit grounded in signal data, not pipeline reports. The output tells you where in the dark funnel your accounts are and which signals correlate with deals that closed last quarter. Talk to us about that.

The buying behavior is not changing back. The architecture either matches it or does not. The companies that match it earliest will compound an advantage that the companies still running discovery-driven motions will spend the next decade trying to close.

Frequently asked questions

About the author: Anuj Agrawal is the Founder of Growleads, a demand intelligence agency for B2B revenue teams at $50M to $500M ARR. He started Growleads in 2024 after watching the B2B buying motion shift from seller-discovers to buyer-decides over the prior decade. Connect on LinkedIn.