CRM Sales Pipeline: How to Set It Up Right

Most B2B pipelines look tidy in the CRM and lie to the people reading them. Deals sit in "Proposal Sent" for six weeks. Half the stages mean different things to different reps. The forecast says one number on Monday and another on Friday, with no event in between that explains the swing.

A pipeline is a model of how your buyers move from first contact to signed contract. When the stages match what actually happens in a deal, the CRM becomes a tool you trust for forecasting and coaching. When they do not, you get a database of optimistic guesses.

This guide walks through building a pipeline that reports the truth: how to name stages around buyer actions, which fields to require, how to set probabilities that mean something, and which reports tell you where revenue is leaking. The examples use HubSpot, Pipedrive, and Salesforce conventions, but the logic holds in any CRM.

Start with how buyers actually buy, not how you sell

The most common mistake is naming stages after seller activities: "Call Made", "Demo Done", "Quote Sent". Those describe what your team did, not where the buyer is. A demo can be "done" while the prospect has zero intent to move forward.

Stage names should describe a buyer commitment that you can verify. "Demo Scheduled" is weak. "Buyer confirmed budget and timeline" is a real state you can check. The test for any stage: could two reps look at the same deal and agree on whether it belongs there? If the answer is no, the stage is too vague.

Map your real sales motion first. Sit with two or three reps and trace a recent won deal and a recent lost one, step by step. You will usually find five to seven distinct buyer states between "we have a lead" and "they signed". More than eight stages and reps start guessing; fewer than four and the pipeline hides too much. For a deeper look at where deals stall between those states, our breakdown of where leads leak in the B2B funnel pairs well with the stage design below.

Here is a pipeline structure that works for most B2B teams selling considered purchases. Treat the conversion and duration figures as illustrative placeholders until you measure your own.

Stage	Buyer state (exit criteria)	Typical win probability*	Median time in stage*
Qualified	Fit confirmed, real need, right to buy	10%	3 days
Discovery	Pain, scope, and decision process documented	25%	10 days
Solution agreed	Buyer confirms the proposed approach fits	45%	14 days
Proposal / negotiation	Pricing under review, terms being discussed	65%	12 days
Verbal commit	Decision-maker said yes, contract in legal	85%	7 days

*Illustrative figures. Measure your own from closed-deal history.

Notice that "Qualified" sits at the entry, not "New Lead". Raw leads belong upstream, in marketing or an SDR queue, scored before they ever reach the pipeline. If you push every form fill into the deal pipeline, your conversion math becomes meaningless. The pipeline starts when a human has confirmed the lead is worth a salesperson's time. Our guide to lead scoring by buying readiness covers that handoff in detail.

Define exit criteria for every stage

A stage without exit criteria is a holding pen. Reps drag deals forward when the quarter looks thin and leave them parked when they are busy. The fix is a short, written rule for what must be true before a deal advances.

Write one or two checkable conditions per stage and put them where reps see them. In HubSpot you can add stage guidance text; in Salesforce, use validation rules or a "Path" with key fields per stage; in Pipedrive, document them in a shared note and a required-field rule.

Good exit criteria are observable:

Discovery exits only when the deal has a recorded decision-maker name, a budget range, and a target go-live date.
Solution agreed exits only when the buyer has confirmed (in writing or on a call you logged) that the approach matches their need.
Proposal exits only when a numbered quote has been sent and the buyer has responded to it.

The point is to remove judgment from stage movement. When advancement requires a fact rather than a feeling, your forecast stops swinging on optimism. This is also where most "stuck deal" problems get exposed, and it connects directly to finding the bottlenecks in your funnel.

Set probabilities from data, not gut feel

Many teams accept the default probabilities a CRM ships with, then wonder why the weighted forecast is always wrong. Default percentages are placeholders. Yours should come from your own closed-deal history.

The method is simple. Pull every deal that entered a given stage over the last 12 months. Count how many eventually closed won. That ratio is your real probability for that stage. If 40 deals reached "Proposal" and 26 closed won, the stage is roughly 65%, whatever the CRM defaulted to.

Recalculate every quarter. Win rates drift as your market, pricing, and lead quality change. A stage that converted at 50% last year may sit at 35% after you opened a new segment with a longer cycle, where deals drag for months before they resolve.

One caution: weighted forecasting (probability times deal value) smooths out fine for a team with hundreds of deals. For a team closing ten large deals a quarter, it produces nonsense, because no single deal is "65% closed". It either closes or it does not. Small-volume, high-value teams should forecast deal by deal and use stage probability only as a sanity check.

Decide which fields are required, and keep the list short

Every required field is a tax on your reps and a risk to your data. Ask for too much and reps fill in garbage to get past the validation. Ask for too little and your reports have holes.

Require only the fields you will actually report on. For most B2B pipelines that means:

Deal amount. Without it, no forecast and no revenue attribution.
Expected close date. The backbone of every forecast and the trigger for stalled-deal alerts.
Lead source. So you can tie won revenue back to the channel that produced it.
Primary contact and company. For activity tracking and account history.

Make those required at deal creation. Add stage-specific required fields only at the stage where the data first exists. Asking for "budget" on a brand-new deal forces a guess; asking for it as the exit criterion for Discovery forces a real answer.

Lead source deserves a note. If you want to know which marketing actually pays back, the source field has to survive from the first touch all the way to closed won. That is the whole premise of closing the loop between marketing and revenue, and a sloppy source field breaks it.

Build the automation that keeps the pipeline honest

A pipeline maintained entirely by hand decays. Reps forget to update stages, and within a month the data drifts from reality. A few automations keep it clean without adding work.

Rotting-deal alert:
  IF deal stage unchanged for > 14 days
  AND deal is open
  THEN notify owner + flag deal "At risk"

Stale close-date cleanup:
  IF expected_close_date < today
  AND deal is open
  THEN require owner to set a new date before next edit

Stage-entry task:
  WHEN deal enters "Proposal"
  THEN create task "Follow up on quote" due in 3 days

The rotting-deal rule alone surfaces more lost revenue than most reporting dashboards, because stuck deals are invisible until someone looks for them. Speed matters at the top of the pipeline too: a deal that sits a day before first contact converts far worse, so first-response timers belong in your automation rules, not just your training deck.

Keep automation visible and few. A pipeline drowning in silent workflows becomes impossible to debug when a deal behaves strangely. Start with three or four rules, watch them for a month, then add more.

The reports that tell you what to fix

A pipeline you cannot read is just storage. Four reports turn it into a management tool.

Stage conversion. What share of deals moves from each stage to the next. A sharp drop at one transition points to a specific problem: a weak proposal, a discovery step that skips qualification, a price objection no one is handling.

Deal aging by stage. How long open deals have been sitting. A cluster of old deals in one stage is a bottleneck you can name and work.

Pipeline coverage. Open pipeline value divided by your target for the period. A common rule of thumb is 3x to 4x coverage, though the right ratio depends on your win rate. If you close 30% of qualified pipeline, you need a bit over 3x to hit target.

Win/loss by source and reason. Tie closed deals back to lead source and a required loss-reason field. This is where you find out that a channel producing plenty of deals produces almost no revenue.

Review these weekly with the team and monthly with leadership. The weekly view drives coaching on specific deals; the monthly view drives decisions about where to spend. Keep a benchmark for each stage transition so you know what "healthy" looks like before you call a number a problem.

Common mistakes that quietly break a pipeline

Too many stages. Eight or more and reps stop agreeing on where deals belong. The data gets noisy and the forecast gets soft.

Activity stages. "Demo Done" tells you what you did, not where the buyer is. Name stages after buyer commitments you can verify.

One pipeline for every motion. A new-business deal and a renewal move through different steps. Run separate pipelines rather than forcing both into one set of stages.

Probabilities no one updated. Defaults straight from the box make your weighted forecast fiction. Pull your own numbers and refresh them quarterly.

Treating the CRM as a sales reporting chore. A pipeline that only serves management theater gets filled in carelessly. Reps maintain it when it helps them work their deals, so design it for them first.

FAQ

How many stages should a B2B sales pipeline have? Usually five to seven. Fewer than four hides too much; more than eight and reps start guessing where deals belong. Map your real deals first and let the stage count fall out of that, rather than picking a number and forcing deals into it.

Should marketing leads go straight into the sales pipeline? No. Raw leads belong upstream, scored and qualified before a salesperson touches them. The deal pipeline should start at "Qualified", once a human has confirmed the lead is worth selling time. Pushing every form fill into the pipeline destroys your conversion math.

What is the difference between a sales pipeline and a sales funnel? The funnel is the marketing-and-sales view of volume at each stage, useful for spotting where the largest share of people drops off. The pipeline is the CRM object your reps work deal by deal, with owners, amounts, and close dates. They describe the same journey at different resolution.

How do I set realistic win probabilities for each stage? Pull every deal that entered a stage over the last 12 months and count how many closed won. That ratio is your real probability. Ignore the CRM defaults. Recalculate quarterly, because win rates drift as your pricing, lead quality, and market change.

Which fields should be required in my CRM? Only the ones you report on: deal amount, expected close date, lead source, and the primary contact and company. Add stage-specific fields at the stage where the data first exists, so reps record facts instead of guesses. Every extra required field is a tax that pushes people toward junk data.

How often should I clean up the pipeline? Run a rotting-deal check weekly: any open deal with no movement in two weeks or a close date in the past gets reviewed and either advanced, re-dated, or closed lost. A pipeline full of zombie deals inflates your forecast and hides the real picture. Automate the alerts so the cleanup is a five-minute review, not an audit.

Checklist and where to start

Before you call the pipeline finished, run through this:

Stages are named after verifiable buyer commitments, not seller activities.
Five to seven stages, each with written exit criteria reps agree on.
The pipeline starts at "Qualified", with raw leads scored upstream.
Required fields are limited to what you report on, added at the right stage.
Win probabilities come from your own closed-deal history, refreshed quarterly.
Rotting-deal and stale-date automations run quietly in the background.
Four reports are live: stage conversion, deal aging, coverage, win/loss by source.

If your current pipeline is a list of optimistic guesses, fixing it is less about software and more about the model behind the stages. Start by tracing three recent deals, two won and one lost, and writing down the real buyer states between first contact and signature. That single exercise usually exposes half the problems in an existing setup.

If you would rather not rebuild it alone, that is the kind of work we do every day. Book a short pipeline review with Lead The Way, and we will walk your CRM stage by stage and show you where deals are leaking and what to change first. Bring your last quarter's closed-won and closed-lost list; an hour with real data beats a month of guessing.