Gantt Chart Dependencies Explained: FS, SS, FF and SF

The project we will use throughout

Dependency diagrams are easy to nod along to and hard to apply. So: one real project — a billing system cutover — that every type below is applied to.

Finish-to-Start (FS) — the one you will use

Task B cannot start until task A finishes: pour the foundation, then frame the walls.

Task 2 is FS after task 1: you cannot dry-run a script nobody has written. Task 3 is FS after task 2, because the defect list does not exist until the dry run produces it. That chain — 1 → 2 → 3 — is the spine of the project and sets the 3 April finish.

FS is the default in every scheduling tool, and if you are unsure it is almost certainly right. Reaching for an exotic type usually means the tasks are wrongly broken down.

Start-to-Start (SS) — work that runs together

Task B cannot start until task A starts, and the two then run in parallel.

The runbook (task 4) is SS after the dry run. The writer needs the dry run underway to have anything to document, but waiting for it to finish would waste a week — so both begin Monday 16 March. The tell for SS is a shared trigger, not a shared deadline.

Finish-to-Finish (FF) — work that must land together

Task B cannot finish until A finishes.

The runbook also has an FF link to task 3: it cannot be signed off while defects are still landing, because each fix rewrites a step in it. So it stays open until Friday 3 April — started by the dry run, closed by the fixes, which is the shape of a support activity. Testing, documentation and QA sign-off are all FF for the same reason: B may start whenever, it just cannot cross the line first.

Start-to-Finish (SF) — the rare one

Task B cannot finish until task A starts. It reads backwards, and most planners never need it.

Task 6 is the case it was invented for. The legacy system must run until the new one is live: its finish is pinned to go-live's start. Link it SF and the legacy bar stretches to meet whatever date go-live lands on.

That is the test: a handover, where the outgoing thing ends because the incoming one begins. Anywhere else, reordering the tasks usually says the same thing with a plain FS.

The four types side by side

Type	Real-world trigger	How often it applies	The mistake it invites
FS Finish → Start	A's output is B's input — a defect list, a signed drawing, a poured slab	~90% of links	Used for preference, not constraint, until every date is pinned
SS Start → Start	Two activities share a trigger, then run together	Common; most plans have a few	Reacts only to the predecessor's start: if A's finish slips, B never moves
FF Finish → Finish	A wrapper that cannot close before the work it wraps — testing, docs, sign-off	Occasional, mostly support tasks	B's start left unconstrained, so it looks able to begin absurdly early
SF Start → Finish	A handover — the old shift or system ends because the new one begins	Rare; many correct plans have none	A clumsy backwards FS that reviewers misread

When the wrong type makes the schedule lie

Now task 5, training. A planner links it SS after task 3: training starts when defect-fixing starts. It looks fine — both bars begin 23 March, everyone trained by 27 March, well before go-live.

So test for it: extend the task most likely to overrun by a week, and see what moved. Anything you expected to shift and did not is a wrong type.

Lag and lead — a lag is a wait, not slack

A lag adds waiting time to a link. "FS + 3 days" means B starts three days after A finishes. Here, go-live is FS+1d after the defect fixes: one clear day for a final data sync.

The thing people get wrong: a lag is not slack. Slack is room you can spend when something slips. A lag is time already committed — concrete curing, paint drying, a contractual review window — and nobody gets to "use" it when they run late. Padding links with lag because a task "might" overrun creates contingency nobody can find or cut.

A lead is negative lag: "FS − 2 days" overlaps the tail of A. Leads compress a schedule — this is what fast-tracking is — but they buy time with risk. An overlap you have not reasoned about is a rework queue.

Both are written 3FS+2d — row 3, finish-to-start, two days lag — which gantts.app takes in the Runs after column.

Total float and free float are not the same number

Every task now has float: how long it can slip before it hurts something. There are two kinds, and confusing them is expensive.

Total float is how far a task can slip before the project finish moves. Free float is how far before its own successor moves. Free float is never larger.

Training has total float — the finish date is set by the defect chain. But total float is shared: if three tasks in a row each show eight days, there are eight days between them, not twenty-four. A lead who reads only that number and delays a week has spent float belonging to the whole chain.

Total float of zero is the definition of the critical path. Tick Critical path in gantts.app and that chain draws striped. That is where the types have to be right.

Creating and editing a dependency in gantts.app

1. Drag from bar to bar. Grab the small dot at the edge of a task bar, drag onto the bar you want to link and release. The link is created as FS with zero lag.
2. Change its type. Click the arrow between the bars. The link editor opens with a Type dropdown: FS · Finish → Start, SS · Start → Start, FF · Finish → Finish, SF · Start → Finish.
3. Add a lag. Same editor, Lag (days) field: 1 for the FS+1d before go-live, a negative number for a lead. Both undo with Ctrl+Z.
4. Or type it. The grid's Runs after column takes predecessors by row number: 3, 2SS, 3FF, 3FS+2d, comma-separated. Fastest way to link a plan you just pasted in.
5. Or use the task card. Click a task, open Runs after (predecessors), and pick from the + Start this after… dropdown. Each existing link has its own type selector.
6. Remove one. 🗑 Remove this link in the link editor, or the ✕ in the task card.

One behaviour worth knowing: adding a dependency here can only push a task later, never pull it earlier. The scheduler is "as-placed" — a task starts at the later of where you dragged it and where its predecessors allow. Link a task whose bar already sits well after its predecessor and nothing moves; the gap stays, because a bar you placed on purpose should not be yanked about by a link added for reporting. To compact everything, press ⚡ Auto-schedule, which pulls each dependent task back to the earliest date its predecessors allow, leaving unlinked tasks as anchors.

Templates that use this

• Project Management Gantt Chart Template
• Free Construction Gantt Chart Template
• Browse all Gantt chart templates

Keep reading

• Critical Path Method (CPM): Steps, Formula & Example
• What Is a Gantt Chart? Definition, Examples & Uses
• Baseline vs Actual: Tracking Schedule Variance on a Gantt Chart

Frequently asked questions