What is the flow waterfall?

The flow waterfall breaks an automation — a welcome series, an abandoned-cart sequence, a win-back flow — into its individual messages and lines them up in order: sends, opens, clicks, unsubscribes and attributed money per step. Flows are sequences, and their totals hide where the sequence works or dies; the waterfall is the step-by-step view that shows it.

Formula

one row per (flow, step): sends, deliveries, opens, clicks,
                          unsubscribes, rates, attributed revenue/margin

Worked example

A three-step abandoned-cart flow sends 8,000 / 5,200 / 4,700 messages. The 35% drop into step 2 is mostly good news — buyers exit the flow after converting on step 1. But step 3's click rate collapses from 4.1% to 0.9% while its unsubscribe rate doubles: by message three the remaining audience isn't buying, just tiring. Attributed margin per step confirms it — step 3 earns a tenth of step 1. The fix is a better step 3 or a two-step flow, and only the waterfall view could tell you that.

How Saldo Metrics computes it

fact_messaging_performance stores flow rows at message grain: each row carries its flow_name and flow_message_name (the step). v_flow_message_performance aggregates per (flow, step) — totals first, rates recomputed from the summed counts — with attributed revenue, margin and send cost per step; v_flow_performance provides the whole-flow rollup above it. Step volumes descend naturally as recipients convert, unsubscribe or fail a flow's filters between messages; the waterfall shows the platform-reported reality of that without modeling why each exit happened.

Why it matters

Automations earn a large share of messaging revenue and run unattended — a broken step keeps quietly underperforming for months in a way no campaign would. Step-level economics also answer the perennial flow question, "should this sequence be longer?": the marginal step's attributed margin against its unsubscribe cost is the actual trade, and the waterfall puts both on one screen.

Common mistakes

• Reading all drop-off as failure. Exits after conversion are the flow succeeding; exits with rising unsubscribes are the flow overstaying. The unsubscribe column separates the two.
• Judging steps by their rates alone. Later steps mail a pre-filtered, less-interested audience; their rates are structurally lower. Compare a step against its own history, not against step 1.
• Crediting the whole flow's revenue to its existence. Attribution here is last-touch matching, not a holdout experiment; a flow's true incremental value is smaller than its attributed total.

Where you see this in the app

Marketing → Messaging, as the flow waterfall section with one row per automation step.