What orchestration adds over a single flow
A single linear flow sends step one, then step two, then step three, on a fixed clock. Orchestration adds a control layer on top: it decides the next best step per person, picks the channel for that step, holds messages until the moment is right, and suppresses anything that would over-message someone. The node taxonomy itself — entry, send, wait, branch, exit — lives in what is a customer journey. This guide is about the layer that makes those nodes behave as one coordinated conversation rather than several independent ones.
Branching and wait-for-event
Two mechanisms make a journey react instead of march. A branch reads a condition — plan tier, segment membership, last action — and routes the person down one path or another. A wait-for-event pauses the journey until a specific signal arrives (a purchase, a login, a reply) or a timeout expires, then continues. A trial that converts on day two should not get the day-three nudge to convert. Events come from your own tracking, so orchestration is only as good as your data; if the events you want to branch on are not instrumented, fix that first in first-party data and tracking.
A/B splits: testing the path, not just the copy
A subject-line test compares two versions of one message. An A/B split inside a journey compares two versions of the path. You can split on timing (send now versus wait a day), on channel (email first versus push first), or on structure (two-step nurture versus five-step). The split randomly assigns each person at the split node, then you measure which branch reaches the goal more often. The unit under test is the orchestration decision itself, not the wording.
Cross-channel suppression and handoff
Suppression is the part people skip, and it is what separates orchestration from spam. Without it your lifecycle email, your win-back SMS, and your in-app prompt can all fire at the same person on the same day, because each was built in isolation. Cross-channel suppression and frequency caps apply at the person level across every channel at once, so total volume is governed, not just volume per channel — Braze frames the rule plainly: customers experience a brand, not separate channels. Handoff is the related decision of which channel carries the next step: escalate rather than repeat, starting with the cheapest, least intrusive touch and only moving to a more direct one when it gets no response. Knowing when in-app beats push or email is covered in web push and in-app messaging.
Tie every journey to a goal metric
A journey without a goal node is just automation. The goal is the metric the journey exists to move — activation, reactivation, a repeat purchase — and it should map to your north star metric for startups. The goal does two jobs: it gives the A/B split something to compare branches against, and it lets a person exit the moment they reach it, which is itself suppression. Orchestration is the mechanism; omnichannel is the maturity level you reach when it works across every touchpoint as one — a coordination outcome, not a feature checkbox, as covered in multi-channel vs omnichannel. Hosted platforms like Customer.io and Iterable center their products on a visual journey builder for exactly this reason. When you build, start narrow: one goal, the events it needs instrumented, the linear happy path, then a branch, a wait-for-event, and a global frequency cap before you ever turn it on. Send at the person's local active hour, not your server's midnight, and respect quiet hours, which for SMS are often a legal requirement rather than a courtesy. Orchestration earns its keep through restraint — the best flow is often the one that decides to stay quiet.