How to build a sales pipeline from scratch

To build a sales pipeline from scratch, first write down who you sell to and what problem you solve. Then define five to seven stages, each with an exit criterion the buyer can confirm. Work back from your revenue target to the number of first meetings you need. Choose two or three sources of pipeline, such as referrals, targeted outbound and inbound content. Record every deal in a CRM from day one, and review it weekly. Expect the first quarter to be about learning conversion rates, not hitting them.
We make Nagent, whose AI Sales team is led by SERA. This guide is neutral; where Nagent fits is near the end.
What do you need before the first deal?
Four things, each fitting on one page:
An ideal customer profile. Industry, company size band, region, the trigger that makes them buy now, and the job title that feels the pain. If you cannot name ten real companies that fit, the profile is too vague.
One clear offer. The problem, the outcome, the price range. A pipeline cannot be measured if every deal is a different product.
A number. New revenue this quarter, and the average deal size you expect.
A CRM. Even a free one. A spreadsheet works for ten deals and fails at forty, when two people edit it and nobody knows which version is true.
Which stages should a new pipeline have?
Most guides agree that stages should follow what the buyer does, not what the seller does. HubSpot's pipeline guide makes the same point and lists generic stages from prospecting to close (HubSpot), and Salesforce describes a typical seven-stage pipeline from prospecting to post-purchase (Salesforce). What many guides skip is the exit criterion for each stage. Here is a starting set:
Target. A company that fits your profile. Exit: you have a named contact and a reason to reach out.
Engaged. They replied or booked. Exit: a first meeting is on the calendar.
Discovery. You have met. Exit: the buyer has confirmed the problem, its cost and who else decides.
Evaluation. They are testing whether you solve it (demo, trial, pilot). Exit: the buyer agrees the success criteria were met.
Proposal. Price and terms are on the table. Exit: the buyer has seen the proposal and named their approval steps.
Commit. Verbal yes. Exit: contract sent with a signing date.
Won or Lost. With a reason picked from a short list.
Keep it to these until you have twenty closed deals. Then look at where deals stall and split or merge stages based on evidence.
How many first meetings do you need?
Work backwards. Here is our worked example, labelled as an example:
Target: $120,000 new revenue this quarter.
Average deal: $10,000, so you need 12 wins.
Win rate from qualified opportunity: assume 20%, so you need 60 qualified opportunities.
Share of first meetings that qualify: assume 30%, so you need 200 first meetings over the quarter, about 15 a week.
Every assumption here is a guess until you have data. Replace each with your real rate after the first 30 to 50 opportunities. The point of the exercise is to show which number is unrealistic: if 15 meetings a week is impossible, the target, the deal size or the conversion rate has to change.
Where should the first pipeline come from?
Pick two or three sources and run them properly rather than five badly:
Founder network and referrals. Fastest and highest win rate, but it runs out. Use it to learn the sales conversation.
Targeted outbound. A list of companies that match the profile, researched one by one, with a short message about a specific trigger. Slower to start, but it scales and you control it.
Inbound content. Pages that answer the questions your buyers ask. Slow to build, cheap per lead once it works.
Partners. Agencies, consultants or adjacent products that already serve your buyer. Good for credibility, slow to set up.
Track the source on every deal from day one. By the end of the quarter you will know which one to double.
How do you decide which deals are real?
Qualify every opportunity against a short checklist at the Discovery exit:
Is there a problem the buyer will say out loud?
Does it cost them something measurable?
Who signs, and have you spoken to them or to someone they trust?
Is there a date that matters to the buyer?
How do they buy (procurement, security review, legal)?
Simple, short-cycle deals need only the first four. Larger deals with several decision makers need a fuller method such as MEDDIC, which adds metrics, decision criteria, the decision process and a champion.
What should you track from week one?
New qualified opportunities per week, by source.
Stage-to-stage conversion, to find the stage where deals die.
Win rate: won deals divided by qualified opportunities closed in the period.
Average deal size.
Sales cycle length: days from qualified to won.
Pipeline velocity: opportunities × win rate × average deal size ÷ cycle length in days. It gives revenue per day and shows which lever moves most.
Pipeline coverage: open pipeline divided by the remaining target.
What does a weekly pipeline review look like?
Thirty minutes, same time every week:
New this week (5 minutes): what entered, from where.
Moved (5 minutes): what advanced, and on what evidence.
Stuck (15 minutes): deals with no movement or a past close date. For each: the next step, its owner and its date, or close it.
Numbers (5 minutes): coverage, conversion and the meeting count against the weekly need.
What mistakes kill a new pipeline?
Counting hope as pipeline. A friendly call is not an opportunity until it passes the Discovery exit.
No close date or next step. Deals without them never close; they just age.
Too many stages too early. Twelve stages for eight deals means nobody knows where anything is.
One source. If referrals dry up, the pipeline stops.
Never closing lost deals. A pipeline full of dead deals hides the real gap.
Which tools do you need at the start?
A CRM. HubSpot's free CRM covers up to two users and 1,000 contacts (HubSpot); Salesforce's Starter Suite lists $25 per user per month, checked 24 October 2026 (Salesforce).
A way to send and track email from your own domain.
Later: conversation intelligence such as Gong once you have recorded calls worth analysing, a sequencer such as Outreach once outbound volume grows, and forecasting such as Clari (now part of Salesloft) once you have a forecast to defend.
AI agents help most with research and first drafts, the slowest parts of targeted outbound.
Where Nagent fits
We make Nagent. Nagent is multiplayer AI for growth: people and AI agents in one shared context, delivering the number. Nagent's CRM starts every workspace with thirteen pipeline stages, from New to Won and Lost, which you can rename and reorder; adding or removing stages is not available yet. On the Sales team, NORA finds accounts that fit your customer profile, researches each into a sourced dossier and drafts the first email, which waits for a person's approval. SERA, the AI Sales Chief of Staff, answers where the pipeline stands from the record, stage by stage, and for a stuck deal names the next step and who owns it. A booked meeting becomes a handoff card owned by a named person.
What it does not do: SERA does not draft a forecast yet, and a new workspace needs your customer profile, your knowledge and a sending account before NORA can do useful work. Agent work is metered in credits; see pricing.
Frequently Asked Questions
How do you set up sales pipeline stages in a CRM?
Name each stage after a buyer action, write one exit criterion per stage, and make the exit criterion a required field or checklist where your CRM allows it. Start with five to seven stages and change them only on evidence.
Who should own pipeline generation, AEs or SDRs?
At the start, the founder or first account executive owns it all. Add an SDR when the account executive spends more time finding meetings than running them, and give the SDR a meeting target, not an activity target.
How do you define an ideal customer profile for B2B?
List your best five customers or prospects, then write what they share: industry, size, region, the trigger that made them buy and the title that felt the pain. Test it by naming ten more companies that fit.
Is a spreadsheet enough for an early-stage pipeline?
For the first ten deals, yes. Move to a CRM before two people need to edit the same list, because version conflicts are where deals get lost.
Want to see a pipeline built and reported from one record? Book a demo.