[Intermediate → Advanced] — the specific analyses that are the analyst job: funnels, cohorts, retention, and the unit economics that decide whether a business works.
The kid version first
Most analyst work isn’t abstract — it’s a handful of standard, high-value analyses that every product and growth team needs: where do users drop off (funnels), do they come back (retention/cohorts), and is each customer worth more than they cost (unit economics)? These aren’t exotic techniques; they’re the bread and butter, and knowing them cold is most of being a product/business analyst.
Funnel analysis — where do users drop off?
A funnel is a sequence of steps toward a goal, and the analysis is measuring how many users make it through each:
Visited 10,000 ██████████
Added to cart 3,000 ███ ← 70% drop here. Biggest leak
Checkout 2,400 ██
Purchased 1,800 ██ ← 3,000 → 1,800 = 60% checkout conversion
The point is finding the biggest leak — the step with the worst drop-off is where a fix has the most leverage. A 10% improvement at the step that loses 70% of users is worth far more than perfecting a step that already converts 90% → funnels in SQL.
The subtleties that matter:
- Time window — do steps have to happen in one session, or within N days? A “funnel” over an unbounded window overcounts
- Order — strict (must go A→B→C) or any-order? Real users skip and backtrack
- The denominator — conversion “of visitors” vs “of those who reached the previous step” are different numbers that answer different questions. Be explicit about which
- Segment the funnel — the overall funnel hides that mobile drops at checkout while desktop drops at cart → segmentation
Cohort analysis and retention — do they come back?
A cohort is a group of users bound by when they started, and cohort analysis tracks each group over time — which separates “are we growing?” from “are we retaining?” (a business can grow by acquisition while every cohort churns, which is a leaky bucket that eventually empties).
Retention by signup cohort (% still active):
Month 0 Month 1 Month 2 Month 3
Jan cohort 100% 45% 38% 35%
Feb cohort 100% 48% 41% 39% ← Feb retains better — what changed?
Mar cohort 100% 52% — — ← improving. Good sign
Reading a retention curve:
- The shape is the product’s health. A curve that keeps falling to zero = users never stick (a real problem, no acquisition fixes it). A curve that flattens = you’ve found a core of users who stay — the “retention plateau” is the single most important signal of product-market fit
- Compare cohorts to see if changes helped — did users who joined after the redesign retain better?
- Retention beats acquisition — retaining an existing user is far cheaper than acquiring a new one, and a product that doesn’t retain can’t grow no matter how much it spends → cohorts in SQL
Churn is retention’s mirror: the rate users leave. Analysing churn — who churns, when, and the leading signals that precede it → is one of the highest-value analyses, because a leading churn indicator lets you intervene before the customer is gone → leading indicators.
Unit economics — does the business work?
The analyses that decide whether a business is viable, not just growing:
- CAC (Customer Acquisition Cost) — what it costs to get one customer (marketing + sales ÷ customers acquired)
- LTV (Lifetime Value) — the total profit from a customer over their lifetime (roughly: average revenue per period × gross margin × expected lifetime, where lifetime ≈ 1/churn rate)
- The LTV:CAC ratio — the number that decides if a business model works. If you spend more to acquire a customer than they’re ever worth (LTV:CAC < 1), you lose money on every sale and growth makes it worse. A healthy ratio is often cited as ~3:1
- Payback period — how long until a customer’s revenue recovers their acquisition cost. Shorter is better; it’s a cash-flow reality, not just a profitability one
Why analysts own this: these numbers connect the product metrics (retention, conversion) to whether the company survives. A retention improvement isn’t just “nicer” — it lengthens lifetime, raises LTV, and improves the ratio that justifies the whole growth engine. Connecting product analytics to unit economics is what makes an analyst a business partner rather than a report generator → the deliverable is a decision.
Acquisition and engagement
The rest of the standard toolkit:
- Acquisition analysis — which channels bring users, at what cost and quality. Cheap channels often bring worse-retaining users, so cost-per-acquisition without quality is misleading → segment acquisition by downstream retention, not just volume
- Engagement metrics — DAU/MAU (daily over monthly active users — the “stickiness” ratio; high means people use it most days), session frequency and depth, feature adoption. These are the leading indicators of retention
- Segmentation and personas — grouping users by behaviour (power users vs dabblers) to understand who your product actually serves, often via RFM (recency, frequency, monetary) or clustering
The connective tissue
These aren’t separate — they form one story:
acquisition → activation (funnel) → retention (cohorts) → revenue (LTV)
│ │ │ │
CAC conversion churn LTV:CAC
A good analyst reads across the whole chain: a funnel fix improves activation, which feeds retention, which raises LTV, which justifies more acquisition spend. Understanding how a change in one propagates to the others is what turns a pile of metrics into a growth model → the metric tree.
The tools
Beyond SQL, product analysts often use product-analytics platforms — Amplitude, Mixpanel, PostHog, GA4 — which specialise in funnels, cohorts and event-based analysis over user behaviour, so you don’t rebuild them in SQL every time. They’re event-tracking systems (the event stream) with a funnel/cohort UI on top. Know both — the platform for speed, SQL for anything custom or when you need to trust the number yourself.
Key insight
Product and business analytics is a small set of standard, high-value analyses — funnels (where users leak), cohorts and retention (whether they come back, and the plateau that signals product-market fit), and unit economics (LTV:CAC, the ratio that decides if the business works) — connected into one chain from acquisition through revenue. The craft is reading across that chain: a retention improvement lengthens lifetime, raises LTV, and justifies acquisition spend. Knowing these cold, and connecting product metrics to the unit economics that determine survival, is what makes an analyst a business partner rather than a report generator.
Related
- SQL for analysis — funnels, cohorts, retention in SQL
- metrics and KPIs — the metrics these analyses produce
- diagnostic analysis — explaining when these move
- A/B testing — proving a change caused an improvement
Source: [reference] — Sep 2026.
Software engineering
Infrastructure & security
Computer science
Hardware & engineering