Customizing sales stages: how many is the right number?

Thursday, 4:20 p.m. A sales manager opens the board ahead of the quarterly review and counts eleven columns: First Contact, Discovery, Needs Analysis, Technical Review, Demo, Post-Demo, Proposal Prep, Proposal Sent, Negotiation, Contract, Won. Of 23 open deals, 14 sit in the first three columns, Post-Demo has been empty for six weeks, and two reps are holding the same account in different places: one insists it is in Proposal Prep, the other moved it to Negotiation last Tuesday. The first twenty minutes of the review go not to where the revenue is, but to what a column is supposed to mean.

This article is about designing stages so that argument stops happening. We separate a stage from an activity, work out the constraints that set the number, count what surplus columns cost, look at where too few leave you blind, write exit criteria two people can read the same way, attach win probability to the stage, decide where losses belong, split one board into several, catch stalled deals early, and migrate an existing pipeline without losing the quarter.

What is a stage, and what is only an activity?

A stage is a change in the buyer's state. An activity is something the seller did. "Demo delivered" is an activity: it happened, it is logged, it is finished. "The technical team reviewed the solution and confirmed in writing that it fits the current stack" is a state. The second one deserves a column. Without that distinction the board slowly becomes a to-do list, and every new sales habit grows a column of its own.

There is a quick test. Can you read the column name as "where the buyer got to" rather than "what we did"? Proposal Prep is work sitting on the seller's desk; the buyer has no idea it is happening and may be sitting through a competitor's pitch that same afternoon. "Proposal received, review promised by Friday" belongs to the buyer. The first is a task, the second is a stage.

Activities do not disappear, they just live somewhere else: meeting counts, documents sent, the number of stakeholders pulled into the thread all belong in fields and on the deal timeline. Getting clear early on where custom fields end and stages begin is the cheapest way to keep the list short, because you stop opening a column for every metric you want to track.

How many stages is the right number, and what decides it?

The right number is not a matter of taste; it falls out of three constraints. First, how many genuinely separate decisions the buyer makes: if budget, technical and procurement sign-off sit with three different people, those are three stages; if one person owns all three, that is one. Second, cycle length, because a stage has to be wide enough to hold a deal for a few days at your median pace. Third, volume: if a column rarely holds more than one deal, its conversion rate is noise rather than statistics.

Those constraints land most B2B teams somewhere between five and seven stages. The table below gives a starting range by cycle length and the symptom you see when the count runs long. The figures are illustrative, not benchmarks; check them against your own closed deals first.

Average cycle length	Sensible stage count	Typical time in a stage	Symptom that the count is too high
7-14 days	3-4	2-4 days	Deals jump two columns in a single day
30-45 days	5-6	6-10 days	One column sits empty for months
60-120 days	6-8	10-20 days	Two reps file the same situation differently
6 months or more	7-9	3-6 weeks	Stage conversion swings quarter to quarter

Starting from a template saves time; leaving the template untouched does not. Trimming a ready-made pipeline template down to your actual motion beats inventing stages from a blank page: you get a skeleton on day one, and by the end of the first quarter your own data tells you which column is dead weight.

What do extra stages actually cost the team?

The first cost shows up in the forecast. Weighted forecasting assigns a probability to each stage; across eleven stages those probabilities end up four or five points apart and the difference stops meaning anything. A jump from 55 percent to 60 tells nobody anything once it becomes a number in a board deck. Fewer stages, spaced further apart, produce a coarser but far more honest forecast.

The second cost is data quality. Every extra column asks the rep for one more drag, and as the dragging gets tedious the record starts trailing reality: the deal is genuinely in negotiation while the board still shows demo. Nobody notices, because the board looks full. What is wrong is not the column but its date.

The third cost is behavioural. When stage progression is tied to pay, columns turn from a measurement into a lever and deals get pushed forward without the buyer moving at all. Tying commission to closed business rather than to mid-funnel movement is the most practical way to keep the board honest; intermediate stages exist to be measured, not to be paid on.

Where do too few stages leave you blind?

There is a floor as well as a ceiling. A three-column board of New, In Progress and Won says almost nothing. In Progress holds a deal with no budget yet and a deal whose contract is already with legal, side by side. Weighted identically, those two inflate the forecast systematically, and the correction lands in the last two weeks of the quarter when there is no time to react.

The second kind of blindness is diagnostic. If your loss rate is high but the pipeline is two columns wide, you cannot say where you are losing: at qualification, on price, or in procurement. A stage is an address for the problem, and without the address improvement is guesswork.

The practical test: every stage should have its own win rate, clearly different from its neighbours'. If two adjacent columns convert at nearly the same historical rate, they are one column wearing two names and should be merged. If the gap is wide and repeats across quarters, they have earned their independence.

How do you write an exit criterion for a stage?

The stage name is half the work; the criterion is the other half. It must be verifiable, and it must rest on something the buyer did. "Customer seems interested" is not verifiable. "The buyer confirmed the budget range verbally and shared the decision timeline" is: two different reps looking at the same record reach the same conclusion.

The first stage attracts the most argument because it is the front door. Putting the thresholds you defined in your ideal customer profile here keeps the board from filling with deals that will never close: if industry, size and geography do not match, no opportunity is created and the record stays on the lead side.

Entry also has a timing dimension, and when a record deserves to enter usually depends on the buyer's calendar rather than yours. Catching the buying signal at the moment it appears raises the quality of the first stage directly; a deal opened with no signal behind it stretches the average cycle no matter how well the stages are designed.

Keep the criteria short: two bullets per stage, three at the outside. The longer the list, the less anyone reads it, and the stage quietly reverts to opinion. A criterion that fits on one line is one the team can apply from memory during a call.

How should win probability be attached to a stage?

A stage needs a number attached, or it produces no forecast. In Rocketly's visual sales pipeline, the moment a deal is dragged into the next column its win probability comes from the stage definition automatically, so reps never type a percentage per deal. Probability then stops tracking personal optimism: two deals in the same stage are weighted identically regardless of who owns them.

Where that number comes from is a separate decision. The healthy route is to read it out of your own history: whatever share of deals that reached Proposal Sent over the last twelve months were won is the probability for that stage. Percentages assigned by instinct look reasonable for one quarter and distort the next, which is usually when people stop trusting the pipeline at all.

Once probability is bound to the stage, part of the movement can be handed to the system. Moving deals to the next stage automatically when a defined condition is met, such as flipping a record to won the moment the quote is approved, removes both lag and forgetfulness. Automate only on events the system knows for certain, never on a rep's judgement.

Is "lost" a stage or an outcome?

Keeping a Lost column on the board is common and misleading. Losing is not a stage, it is how the deal closed, and the right design stores the stage the record was in when it was lost. "Lost at Proposal Sent" and "lost at qualification" are two completely different stories; drop both into one column and they become indistinguishable forever.

Collecting the reason as free text does not help either: everyone phrases it differently and nothing aggregates. In Rocketly, loss reasons roll up into six fixed headings, namely price, competitor, budget, timing, no response and other, and are reported in win/loss analysis. A closed list is what lets "why are we losing" be answered at quarter end with a table instead of a debate.

Read stage and reason together and design faults surface. If most losses are marked no response and nearly all happen at the second stage, the problem is not price but the exit criterion of the first stage: records that never earned entry are getting in. The fix is a higher bar at the door, not a discount at the end.

One pipeline or several?

The most common reason a stage list bloats is that unlike motions share one board. New business and renewals do not belong in the same columns; a renewal has no demo step, it has usage data and a contract end date. Squeeze both into one board and you get an eleven-column hybrid that nobody uses fully and everybody works around.

Multiple pipelines exist for exactly this split: in Rocketly you create separate pipelines for a product line, a segment or a team, each with its own stages. Every board then stays short relative to its own job, forecasts are read per pipeline, and a bottleneck in one motion no longer drags down another's averages.

Where to draw the line is usually a question your customer segmentation already answers. Selling the same product to a fifty-person company and a five-hundred-person company means the second involves a buying committee and a vendor review; carrying those steps on a shared board lengthens the small-segment pipeline for no reason.

Direction of demand is another natural break. Inbound demand and outbound prospecting diverge completely in their first two stages: in one the interest already exists and the work is qualification, in the other interest has to be created. Sharing columns distorts both engines, because normal dwell time for one is an alarm threshold for the other.

How the team works matters too. The choice between inside and field sales determines whether site visits and on-premise discovery deserve stages of their own; in a motion that runs entirely over calls, those steps do not exist and the board gets shorter on its own.

How do you notice when a deal is stuck?

The real payoff of a well-designed stage list is that it makes stalling visible. A pipeline velocity report shows the average number of days spent in each stage, and the bottleneck almost always concentrates in one column, usually the one whose exit criterion is vaguest. The velocity report is pointing at a design fault, not just a slow deal.

You should not have to wait for someone to open a report. Automatic reminders raise a notification or open a follow-up task on any deal that has sat in the same stage longer than the number of days you set. Setting that threshold per stage is worth the effort: five days is a long time in qualification, while twenty in procurement can be entirely normal.

The deal health score aggregates the same signal at record level. Every open deal carries a score from zero to one hundred, and fourteen days of inactivity, thirty days of stagnation and a close date already in the past all push it down. When the stage design is right, low scores cluster in one column and where to intervene stops being a matter of opinion.

How do you change stages without breaking the board?

Changing stages is a data migration, and in the wrong order it costs you the quarter's reporting. Write the new list first and map every new stage to the old one it replaces. Merging is easy, splitting is hard: two old columns collapse cleanly into one, but splitting a column means deciding record by record which side each deal belongs on.

Customizable stages in Rocketly are arranged by drag and drop, and once order and names are settled, bulk stage change and assignment moves the existing records, up to a hundred deals in a single pass. Do the migration at the start of a quarter, not the end; changing columns underneath deals that are actively closing muddies both the forecast and the commission math.

Afterwards, pin down how the team looks at the board. Saved board views let you name and store pipeline, owner, stage and attention filters, and once everyone shares a filter the "it looks different on my screen" conversation ends. Stage-triggered automations need one pass of review against the new names too, or rules bound to a retired column stop firing silently.

Finally, give the work a cadence. A stage list is not something to revise monthly; folding a once-a-year review into your annual sales planning ritual is enough.

If you want to test this sequence against your own numbers, open a free account and start with one pipeline and five stages, widening it only where the data says so.

FAQ

If I cut stages later, does my historical reporting break?

Closed deals keep the stage they held when they closed, so past win rates are not erased. What breaks is comparability: the old conversion rates of two merged columns cannot be lined up against the new one. The cleanest approach is to note the change date and read reports as before and after it.

Should every stage have required fields?

One or two help; more than that pushes reps into not advancing records at all. The most valuable required fields are the ones that feed the forecast, namely expected close date and amount. Rather than making everything mandatory, set a rule that raises a reminder when a field is still blank a few days in.

Can a deal be moved backwards?

It can, and sometimes it should; if the buyer genuinely stepped back, the board has to reflect that. What matters is measuring how often it happens. A column that sees a lot of backward movement is a column whose exit criterion is firing too early.

Can different teams use different stage names?

Yes, once they run on separate pipelines each with its own stages. The one thing that must stay identical everywhere is the definition of won and lost; otherwise no single company-level win rate can be calculated.