learn-sales-pipeline-analytics-with-phoebe / Business session 1 of 4
Learn B2B Sales Pipeline Analytics with Phoebe · Business session 1 of 4

What a pipeline is evidence of

Every Monday somebody reads out a pipeline number and the room nods. Tonight asks what that number is evidence of. A pipeline is a list of claims about buyers: this one confirmed a need, that one let us run a trial, this one asked for a contract. When the stages record what the buyer did, the pipeline is evidence. When they record what the rep did, it is a diary. Then the four numbers every pipeline rests on, and the velocity identity that ties them together, with the trap hidden in its denominator. Kettle & Co's office-tea team is constructed; the cycle times come from real B2B funnel rows.

🟢 Business BD and sales leaders who read the pipeline, and the analysts who build it No code 45 min
0-3 · Welcome 3-18 · Stages, four numbers, velocity 18-40 · Build-along: your stage definitions 40-45 · Q&A
Part 0

Where we start

The Business track is for the people who own the sales number and the people who have to explain it: the BD lead, the sales manager, the founder who signs off the plan, the head of data asked "is this forecast real?". Four sessions: tonight, what a pipeline is evidence of; then how to read a pipeline review without being steered; then who owns the forecast and who owns its error rate; then how to get from pipeline to a hiring plan. No code. The Analyst track builds the same numbers from the rows. Kettle & Co's office-tea arm, a BD lead and four account reps selling tea subscriptions to offices in Japan, is constructed: its people, deals and amounts are invented and labelled. The timings and outcomes underneath come from real rows of Olist's public B2B funnel, a Brazilian marketplace recruiting sellers.

🎫 Clear all three quiz questions to stamp your passport Live = presented in the 45 minutes Self-study = the full depth, after class ★ Build-along = everyone works it through

★ What you walk out with tonight. A stage list for your own pipeline where every stage names something the buyer did and where it is recorded, the four numbers your pipeline rests on with their honest definitions, and one velocity figure computed with the median cycle time, not the mean. The stage list is the first page of the forecast contract in Business session 3.

Part 1 · covers stages as buyer commitments

A stage is something the buyer did

5 min live

Most pipelines start life as a list of things the rep does: called, emailed, demo given, proposal sent. Every one of those can happen while the buyer does nothing. A stage that a rep can reach alone measures effort, and effort is not evidence that a deal will close. Rewrite each stage as an act by the buyer, and record where the proof lives.

THE SAME FOUR STEPS, WRITTEN TWO WAYS left: who can move it alone · right: what the buyer did, and where it is recorded REP ACTIVITY Called the office the buyer may not have picked up Demo given to whoever sat in the room Proposal sent nobody asked for it all reachable by the rep alone 2 · Need confirmed office manager named the problem · call note 3 · Tasting trial the office let us set up a station · trial log 4 · Proposal buyer asked for a price for their headcount 5 · Contract out procurement holds the contract · e-sign log Stage 1, Contacted, is the only one a rep reaches alone, and it is labelled as such. Constructed Kettle stages.
🔍 Click to zoom - on the right, every stage needs the buyer to move, and names the record that proves it
LiveThree tests for a stage definition3 min▶

Who moved? If the rep can reach the stage without the buyer doing anything, it is an activity, not a stage. "Proposal sent" fails; "buyer asked for a price" passes.

Where is the proof? Each stage names one record a second person can check: a call note with the buyer's words, a trial log, an e-signature request. If the only proof is the rep's memory, the stage will drift with the rep's mood.

Can it go backwards? A trial that ends with "not this year" moves the deal back. A definition that only ever moves forward hides every stall. Analyst session 1 shows why most CRMs lose the backward moves unless somebody keeps the event log.

Self-studyWhat the CRM vendors do by default2 min▶

HubSpot's default deal pipeline, read on its knowledge base in October 2026, ships seven stages: Appointment scheduled, Qualified to buy, Presentation scheduled, Decision maker bought-in, Contract sent, Closed won and Closed lost. Two of the five open stages are things a rep can do alone (schedule, present). Salesforce's defaults are reported by third-party guides to follow a similar sales-process naming. Defaults are a starting point for a team that has not thought about it yet, not a definition of evidence; rename them to buyer acts before the first forecast relies on them. Re-verify the vendor pages before you teach from them: they change.

Real world

The demo that was a stage. Kettle's pipeline used to have "Demo given" as stage 3. Ben, the most energetic rep, gave tasting demos to anyone who would sit down, including receptionists with no budget. His stage-3 count was the highest on the team and his win rate the lowest. Renaming stage 3 to "Tasting trial: the office agreed to a station for two weeks" cut his stage-3 deals by half overnight and made the stage worth reading. Constructed Kettle & Co example; the pattern is common.

Part 2 · covers the four numbers

Four numbers carry every pipeline

5 min live

However many dashboards sit on top, a pipeline rests on four numbers: how many deals are in it, what share of them win, how big a won deal is, and how long a deal takes to close. Each has an honest definition and a flattering one, and the gap between them is where most pipeline arguments live.

THE FOUR NUMBERS, WITH THE HONEST DEFINITION OF EACH DEALS count only deals past a buyer act (stage 2+) "Contacted" is a lead, not a deal WIN RATE won / entered deals that entered together, decided by entry month, Analyst session 2 DEAL SIZE median won what was signed, not what was asked one whale moves a mean CYCLE TIME 14 days median 48.4 days mean real Olist rows, 842 won deals The cycle-time mean is 3.5 times its median: a few deals that took over 180 days pull it up. Report the median.
🔍 Click to zoom - three cards are definitions you choose; the copper one is a real distribution with a long tail
LiveThe flattering version of each number3 min▶

Deals. Counting every contact as a deal makes the pipeline look full and the win rate look terrible. In Kettle's quarter 153 of 300 open deals sit at stage 1, Contacted; none of the 1,022 stage-1 deals in the previous quarter closed inside that quarter. Report stage 1 separately as leads.

Win rate. "Won this month divided by deals created this month" mixes two groups of deals. Win rate is the share of deals that entered together and have now been decided. Analyst session 2 shows how the period version swings by tens of points on the real rows.

Deal size. The average asked price includes deals that will be discounted or never signed. Use the median signed amount, and report the largest deal separately, because one whale can move a quarter (Analyst session 6 breaks the bench with exactly that).

Cycle time. On the real Olist rows, the median won deal closed 14 days after first contact; a quarter of them took 54.75 days or more, a tenth took 161.9 days or more, and the mean is 48.4 days. Any formula that uses "average cycle time" is using the 48.4.

Self-studyWhy the real rows are a Brazilian marketplace, not a tea company2 min▶

Public B2B pipeline data is rare: companies do not publish their CRM. Olist, a Brazilian marketplace, published 8,000 marketing-qualified leads from businesses that asked to sell through it (first contact June 2017 to May 2018) and the 842 that signed, with dates. That is a real B2B funnel with real timings. It has no lost deals and no stage history, so every stage, amount and rep on these pages is constructed and labelled. The same rows feed the funnel bench in Marketing, session 4 ↗ for budget and acquisition cost; here they are used for cycle time and forecasting.

Part 3 · covers the velocity identity

Velocity: the four numbers multiplied out

5 min live

Pipeline velocity is the four numbers in one line: deals times win rate times deal size, divided by cycle length. The answer is revenue per day the pipeline produces. It is an identity, not a forecast: it says what the four numbers imply if they hold. Its use is to show which lever moves revenue, and its trap is the denominator.

VELOCITY = DEALS x WIN RATE x DEAL SIZE / CYCLE LENGTH yen per day · counts and amounts constructed 40 deals stage 2 or later 25% win rate 10 wins x 480,000 yen cycle length: the pipe median, 14 days 342,857 / day mean, 48.4 days 99,174 / day 4.8m yen of wins either way. Only the cycle length changed, and velocity moved 3.5x. Say which cycle number you used. The median describes a typical deal; the mean is pulled by the slow tail.
🔍 Click to zoom - the copper box is the velocity most dashboards print, because "average cycle time" means the mean
LiveWhat velocity is good for, and what it is not3 min▶

Good for: choosing a lever. Because it is a product, a 10 percent gain on any one of the three top terms is worth the same 10 percent of velocity. Cutting cycle length by 10 percent is worth a little more (one over 0.9 is 1.11). That makes a useful argument about where to spend effort, and it is arithmetic, not evidence.

Not good for: a forecast. Velocity assumes the four numbers hold, that every deal behaves like the typical one, and that the deals in the pipe today are like the ones that closed before. Analyst session 4 builds forecasts that test those assumptions instead of assuming them.

No single origin. The velocity formula circulates in sales-operations writing without one source to cite. Treat it as a convention whose arithmetic you can check, which is a better position than citing a name.

Self-studyWhy the median is still not the whole answer2 min▶

The median describes the typical won deal. Revenue, though, often sits in the slow tail: big offices take longer to sign. If your largest deals close slowest, a velocity built on the median can overstate how soon the money arrives. The honest version is two numbers, median and 75th percentile, and a sentence on which deals sit in the tail. Analyst session 3 draws the full distribution and the curve for deals still open.

Real world

The board slide that tripled overnight. Kettle's BD lead presented velocity using the CRM's "average days to close". A new analyst rebuilt it with the median and velocity jumped about 3.5-fold, with nothing in the business changed. The right response was neither number alone: both, labelled, plus the 75th percentile. Constructed Kettle & Co example using the real Olist median and mean.

Build-along

Rewrite your stages as buyer acts

★ 22 min · everyone works it through

No code, one page. Take your own pipeline's stage list, or Kettle's old one (Called, Demo given, Proposal sent, Negotiation). Five steps turn it into stages that are evidence. Do step 1 together, steps 2 to 4 in pairs, then read your step 5 velocity line aloud. Your stages will differ from Kettle's, which is the point.

THE STAGE SHEET one row per stage · Kettle's trial row filled in as the example · constructed stage buyer act that moves it in record that proves it can move back? 3 · Tasting trial office agreed to a station trial log, signed yes, "not this year" your stage your stage A row with no buyer act, or no record, is an activity. Move it to the activity log.
🔍 Click to zoom - four columns, and the last one is the column almost every pipeline leaves out

Step 1 · List what you have. Write your current stage names in order, exactly as the CRM shows them. Mark each one R if the rep can reach it alone, B if the buyer has to do something.

★ Step 1 · today's stagesstage 1: ______________ R / B stage 2: ______________ R / B stage 3: ______________ R / B stage 4: ______________ R / B Kettle's old list: Called (R) · Demo given (R) · Proposal sent (R) · Negotiation (B)

Step 2 · Rewrite every R as a buyer act. Ask: what would the buyer have to do for this to mean something? "Demo given" becomes "the office agreed to a two-week trial station". Keep stage 1 as the one rep-only stage and call it what it is.

★ Step 2 · buyer acts1 Contacted (rep only; report as leads, not deals) 2 ______________ buyer act: ________________________________ 3 ______________ buyer act: ________________________________ 4 ______________ buyer act: ________________________________ 5 ______________ buyer act: ________________________________

Step 3 · Name the record. For each stage, the one record a second person could open to check it. If no such record exists today, write what would have to be logged.

★ Step 3 · the proofstage 2 proof: ____________________ (e.g. call note with the buyer's words) stage 3 proof: ____________________ (e.g. signed trial log) stage 4 proof: ____________________ (e.g. buyer's price request email) stage 5 proof: ____________________ (e.g. e-signature request sent to procurement)

Step 4 · Allow the backward move. For each stage, write what sends a deal back, and confirm somebody records the date it happened. Without dates you cannot rebuild last month's pipeline, which is the first thing Analyst session 1 needs.

★ Step 4 · backward movesa deal leaves stage 3 backwards when: ______________________ the date of that move is stored in: ______________________ Kettle example: trial ends "not this year" -> back to stage 2, logged with date

Step 5 · One velocity line, two cycle numbers. Fill in your four numbers with the honest definitions, and compute velocity twice: with the median cycle and with the mean. Read the two aloud and say which one your dashboard shows today.

★ Step 5 · velocity, twicedeals (stage 2+): ____ win rate (by entry period): ____ median deal size: ____ velocity, median cycle (__ days): ____ per day velocity, mean cycle (__ days): ____ per day Kettle example (constructed counts, real Olist cycle): 40 x 25% x 480,000 yen / 14 days = 342,857 a day ; / 48.4 days = 99,174 a day

Keep the sheet. The stage list is page one of the forecast contract in Business session 3. If a stage has no record, the analyst cannot measure it, and a forecast built on it inherits the rep's mood.

Homework

Before Business session 2

30 min
  1. Count your stage-1 share. What share of your open pipeline sits in a stage the rep can reach alone? Write it down; Kettle's is 153 of 300.
  2. Find your cycle-time definition. Ask whoever owns the CRM report whether "average days to close" is a mean or a median, and from which date to which.
  3. Pick one stage to rename as a buyer act, and ask one rep whether the new name would change where their deals sit today.
  4. Bring last Monday's pipeline report. Business session 2 reads it for coverage, stale deals and deals that moved quarters.
Check yourself

Three questions

Clear all three and Business session 1 gets stamped in your passport.

1 · Which of these is a pipeline stage that counts as evidence?

A stage the rep can reach alone measures effort. A buyer act with a record a second person can check is evidence.

2 · On the real Olist rows the median won deal took 14 days and the mean 48.4. What does that do to a velocity formula that uses "average cycle time"?

The few deals that took months pull the mean up. Report which cycle number you used, and show the 75th percentile with it.

3 · What is pipeline velocity?

It says what the four numbers imply if they hold. It has no single origin, and it assumes away everything a forecast has to test.

Sources covered

What this session teaches, and where it came from

✓842 won deals; cycle median 14 days, 75th percentile 54.75, 90th 161.9, mean 48.4Olist, "Marketing Funnel by Olist", Kaggle dataset, CC BY-NC-SA 4.0; computed by the engine in Analyst session 3, read at source
✓HubSpot's default deal stagesHubSpot Knowledge Base, "Set up and customize pipelines", read at source October 2026; re-verify before teaching
◐Salesforce's default stage namingSalesforce Help, opportunity stages; reported from third-party guides
◐The velocity identityA sales-operations convention with no single origin; taught as arithmetic
◐Kettle & Co stages, 153 of 300 deals at stage 1, the 40-deal velocity exampleConstructed and labelled; the stage counts are printed by the bench in Analyst session 4

Every fact on this page, and its verification tier, is recorded in the course's source map.

Business session 1 cheat sheet · pin this

A stage is a buyer actIf the rep can reach it alone, it is activity. Log it, do not stage it.
Name the recordEvery stage points at one record a second person can open.
Let deals go backAnd store the date. Without it, last month's pipeline is gone.
Four numbersDeals (stage 2+), win rate by entry period, median signed size, cycle time.
Velocity is arithmeticDeals x win rate x size / cycle. Compare levers with it; never forecast with it.
Median and mean14 and 48.4 days on the real rows: say which one, show both.