How to build a sales pipeline from scratch, one account at a time

A sales pipeline is every account and deal you are working, sorted by stage, so you can see what will close, what is stuck and what is missing.

Most pipeline guides start at the first meeting and describe deal stages inside a CRM. When you are building a pipeline from scratch, the hard part comes before that: deciding which companies belong in it at all, and why now. So this guide treats the pipeline as a chain of accounts, with a piece of evidence that moves each account from one stage to the next. If you cannot name the evidence, the account has not moved.

A funnel shows the same stages as counts. A pipeline keeps the accounts and the evidence behind each count, which is what lets you fix it.

Build from accounts because B2B purchases start with a change at a company and are decided by a group, and a list of names gives you neither the reason nor the people.

Gartner states on its B2B buying journey page (archived Sep 25, 2026; no sample stated) that “99% of B2B purchases are driven by organizational changes”. Forrester’s State of Business Buying 2026 (Jan 21, 2026) reports that a typical buying decision now includes 13 internal stakeholders and nine external influencers. And buyers punish the alternative: in a Gartner survey of 632 B2B buyers published Jun 25, 2025, 73% said they actively avoid suppliers who send irrelevant outreach.

SalesOne’s own research data shows why fit alone does not fill a pipeline. Across four production research runs in October 2026, S1 agents decided on 1,051 companies that matched a profile on paper and kept 249, about one in four. Of the 802 they turned down, 67.1% (538) had no qualifying dated event inside its window. Size outside the profile accounted for 14.3%, and ownership, such as a subsidiary or a pending sale, for 9.7%. Most companies that look right are not ready now.

• Fit alone fills a pipeline with accounts that have no reason to buy yet.
• Keep the evidence for every stage, so a stalled account shows you which link broke.
• Plan for about four companies checked for every one you keep, then measure your own rate.

To build a sales pipeline from scratch, write an ideal customer profile, build a target account list, add a dated buying signal to each account, qualify it, map the buying committee, reach out with a point of view, and measure each stage. Do the seven steps in this order, and write down what each one decided before you start the next.

1. Write an ideal customer profile you can check

Turn your best customers into rules a stranger could apply: what the company sells, its size band, its ownership, its region, and what rules a company out. Add an exclusion list of customers, open deals and partners. The ideal customer profile guide explains each rule, and the free ICP template gives you the fields, including a window for each buying signal.

2. Build the target account list

List the companies that pass the rules, with a source and a year for size and ownership. Filters alone are a weak start: in a SalesOne test on Sep 29, 2026, 28.9% of 499 companies pulled from one company-data API with industry and size filters were clear misfits. How to build a target account list covers sources and checks step by step.

3. Add a dated buying signal to every account

Choose the five to ten events that came before your past wins, such as a new executive, a new owner, a new site or a system being replaced, and give each a window. Then search for those events and attach the event, date and source to each account. Signal-based selling explains the approach; trigger events in sales lists event types with their windows and where to find each one for free.

4. Qualify every account, with a reason

Keep an account only if it fits and its signal is still inside the window. Write one line on why it stays, and one line on why every other company goes. This is account qualification. The reasons become your best feedback: a familiar name missing from the pipeline has a recorded cause, and a rule that turns down too many good companies shows up quickly.

5. Map the buying committee

Name the person whose job the event affects most, then the people around them: the economic buyer, the people who will use what you sell, and anyone who can block it. Use leadership pages, press releases and filings, and keep the date of each source so you do not write to someone who has left. The buying committee guide covers the six roles and who to contact first.

6. Reach out with a point of view

The first message names the event and its date, the problem it usually creates for someone in the reader’s role, and one useful offer. Send it while the window is open, and have a person approve it. B2B sales prospecting covers channels and sequencing, and the cold email guide covers writing and delivery.

7. Measure each stage, then change one thing

Count accounts at each stage every week and look at where they stall. The measures that matter in a new pipeline are below.

The number of accounts you need comes from working backward from the customers you need, using one conversion rate per stage, until you reach the number of companies you must check.

Pipeline math is division. If you win one opportunity in four, you need four opportunities for every customer, which is the same as carrying pipeline worth four times your target. Repeat that for each stage. The only rate SalesOne can give you from its own data is the first one: S1 agents kept 249 of 1,051 companies checked, about 1 in 4.2. Every other rate depends on your market, your offer and your message, so use your own, and use rough guesses only until you have a few weeks of data.

Two things help. Every account you keep has a dated reason to hear from you, so compare your later rates with any list-based campaign you have run. And every company you turn down keeps its reason, so you do not check it again next quarter unless something changes.

With no customer history, build the pipeline in small batches from a written guess about who you help, and let each batch correct the guess.

1. Write the problem you solve in one sentence, then the kinds of companies that have it most, and the events that make it urgent. That is your first profile and your first signal list.
2. Research a batch of 25 to 50 companies against it. Small batches keep the research cheap while the profile is still a guess.
3. Contact the accounts you keep within their windows, and talk to every company that replies, even when it is a polite no. Each conversation tests one rule.
4. After each batch, read the reasons for every company you turned down and every account that went quiet. Tighten or widen one rule at a time.
5. Keep the founder, or whoever knows the problem best, in every first conversation until the profile holds. Founder-led sales covers when to hand over.

Expect the profile to change in the first batches. That is the point of starting small: each change is cheap while the pipeline is still short.

Each account needs enough recorded evidence that someone else could pick it up tomorrow and know why it is there, who to talk to and what happens next.

A spreadsheet is enough for the first few hundred accounts. Move to a CRM when more than one person works the pipeline, and keep the same fields.

Measure how accounts move between stages and why they stop, not how many emails went out.

• Keep rate: companies kept out of companies checked. A falling rate means the profile or the signal list is drifting.
• Reject reasons, by share. Mostly timing means you need more or different signals; mostly size or ownership means the search, not the profile, needs fixing.
• Accounts contacted inside their window, as a share of accounts qualified. Accounts that expire before anyone writes are research thrown away.
• Conversations per account contacted, counted by account rather than by message.
• Meetings held per conversation, and opportunities per meeting held.
• Time in stage. An account that sits in one stage for more than a few weeks needs a next step or a decision.
• Lost reasons, in one line each, read every month.

Compare your own numbers month to month rather than against published reply-rate benchmarks; we have not found an independent one with a stated method and sample.

• Starting with a contact export and adding the reasons later.
• Counting companies that fit as pipeline, when most of them have no reason to buy this quarter.
• Losing the reason a company was turned down, so it gets researched again.
• Writing to one person when the decision belongs to a committee.
• Letting a signal’s window close while the account waits in a queue.
• Changing the profile, the signals and the message all at once, so no one knows what worked.

SalesOne does steps 1 to 5 as research, and your team keeps the conversations. S1 deep research agents read your own material into a company profile and an ideal customer profile, which your team confirms. They then search your market for dated events, accept or reject each company against the profile, and map the buying committee for each account they keep.

Every buying trigger comes with its source and the date it happened. In SalesOne research runs, the newest trigger on a typical account was 59 days old at delivery, and 85% were 90 days or under. Every company S1 agents turn down keeps the rule it failed and the evidence. For each account kept, they find about four people, each backed by a dated public source.

Contact data comes only for the people your team picks, from 20+ data providers. A person approves every message, and every approval is recorded with who gave it.