Business track 🧭 · four sessions, no code
For BD and sales leaders who read the pipeline and sign the forecast. Stages as buyer commitments, the four numbers and the velocity identity, coverage and stale deals, the forecast contract that splits the number from the method, and the evidence that says it is time to hire.
What a pipeline is evidence of
Stages as things the buyer did, not things the rep did; the four numbers every pipeline rests on; the velocity identity and the trap in its denominator.
🔍Business session 2Reading a pipeline review
Coverage ratios and where "3x" comes from, stale deals, deals pulled forward and pushed out, and the stage mix behind a full-looking pipeline.
🤝Business session 3Who owns the forecast
BD owns the number and the commit; data owns the method, the range and its error rate. The one-page forecast contract.
🪜Business session 4From pipeline to plan
Rep capacity and ramp, and the evidence in your own numbers that says scale: win rate, cycle time and forecast error over past quarters.
Analyst track 🧮 · six sessions, hands-on in the browser
For the data analysts who build pipeline reporting and forecasts. Every widget computes in your browser on real rows of Olist's public B2B funnel: 8,000 leads and the 842 that signed. The bench forecasts one quarter three ways and scores each against what really closed.
The pipeline data model
Opportunity, stage history and weekly snapshot; why the CRM's current state overwrites what happened; rebuild any past week with an as-of join.
🧮Analyst session 2Conversion done honestly
Cohort-based, not period-based; by lead origin; the logging lag that makes whole months look dead; survivorship in won-only tables.
⏱️Analyst session 3Velocity and cycle time
Distributions, medians and percentiles, and a survival curve for deals still open, with censoring explained without heavy maths.
🔭Analyst session 4Three forecast methods
Rep commit, stage-weighted and history-based on the same quarter, and what a vendor's default stage probabilities do to the forecast.
🎯Analyst session 5Bias, error and the track record
Forecast against actual over many quarters, bias against error, per-rep over- and under-calling, and ranges that admit how wrong they can be.
⚖️Analyst session 6 · bench nightThe forecast honesty bench
The ladder of three methods against what really closed, twenty replays, the happy-ears anti-lever, and the whale that slips a quarter.
Your passport 🎫
Clear all three quiz questions in a session and it gets stamped. Ten stamps means you can say what a pipeline proves, rebuild it as it stood on any date, measure conversion and cycle time honestly, and score a forecast against what really closed. Stamps live in this browser only.
What this is not, so you land in the right place 🧭
| If your question is | Go here instead |
|---|---|
| How do reps use AI to sell, prepare calls and keep the CRM clean? | AI Sales ↗ |
| Which statistical model should forecast my time series? | Timeseries Forecasting ↗ |
| How big is the market we are selling into? | Market Sizing ↗ |
| Whose job is which analysis between data, marketing and BD? | Marketing Analytics ↗ |
| What happens after the deal: renewal and churn? | Customer Retention ↗ |
| Can we trust the pipeline, and how wrong is the forecast? | you are in the right place |
Pipeline analytics from the data side. Kettle & Co is constructed; the funnel rows are Olist's public dataset, read at source; vendor claims about forecast accuracy are taught as contested.
Choose your path 🗺️
Three doors into the same conviction: a pipeline is evidence, and a forecast is only as good as its measured track record.
What is honest here 🧾
The running team is constructed and says so on every page. The funnel rows are real and named. Where a claim is contested, the page teaches both sides.
| Claim | Status |
|---|---|
| Kettle & Co, its reps, stages, deal amounts and the 300-deal quarters | Constructed, and labelled each time |
| Lead origins, first-contact dates and won dates (Olist marketing funnel, 8,000 leads, 842 won) | Read at source from the public dataset |
| Every number on the widgets and the bench | Computed in your browser from those rows and the constructed layer; a Python reference prints the same numbers |
| CRM vendor defaults for stages, probabilities and forecast categories | Read at source where the page could be opened, otherwise reported, and dated |
| Kaplan-Meier, scaling readiness | Reported from the original paper and the author's book, paraphrased |
| "3x coverage" and vendor surveys of forecast accuracy | Convention and contested, and taught as such |
The knowledge map 🧠
Both tracks at a glance - hover a session to spotlight its concepts, click any node to jump in.