Hear it call your own phone
Half an hour. We build a flow for your enquiries and run it against your number.
Linear sequences vs branching flows: why a fixed follow-up schedule loses enquiries
A sequence decides in advance what happens on day one, day three and day seven. A customer decides in the moment. The difference between the two is where most follow-up goes wrong.
By the Unibound team6 min read
Almost every sales and service tool has a feature called sequences, cadences or drips. You write a list of steps, set the gaps between them, and enrol contacts. It is easy to build and easy to understand, and for a three-email welcome series it is all you need.
It breaks down the moment follow-up involves a phone call, a second channel or a person. Then the fixed list starts doing foolish things: texting someone who is already talking to your sales team, emailing a customer who said no on the phone yesterday, ringing a fourth time when a WhatsApp would have been answered in a minute.
What a linear sequence assumes
A sequence has one question: did they respond? If not, do the next thing. That hides three assumptions.
- Everyone wants the same channel. They don't. In YouGov's survey of British preferences, the phone was first choice for 50% of the oldest generation and 16% of the youngest.
- Every non-response means the same thing. A call that rang out, a call that was declined, a voicemail and a wrong number are four different situations. A sequence treats them as one.
- The only way out is a reply. So a customer who books by phone keeps receiving the emails, unless someone remembers to remove them.
Every step has more than one outcome
Take a single AI call to a new enquiry. It can end at least five ways:
| Outcome of the call | What a sequence does next | What should happen next |
|---|---|---|
| Booked | Sends step two anyway, or needs manual removal | Stop. Send a confirmation. Start the reminder flow |
| No answer | Waits two days, sends step two | Text or WhatsApp now, while they remember enquiring |
| Asked for a call back later | Sends step two | Wait until the time they named, then call |
| Wants to talk to a person | Sends step two | Put them in the team's call queue with the transcript |
| Not interested | Sends step two | Stop, and mark the record |
| Asked not to be contacted | Hopefully stops | Opt out on every channel, permanently |
Six outcomes, six different right answers, and the sequence has one. Multiply that by each step and each channel and you can see why linear follow-up either annoys people or is kept so timid that it gives up after two attempts.
What a branching flow looks like
A flow is a diagram. Each step ends in named outcomes, and each outcome is a line to the next step. Here is a new-enquiry flow: a call first, then WhatsApp if it rings out, a wait for a reply, an AI chat if they answer and an email if they don't.
Notice that nobody receives every message. A customer who answers the first call sees one step. One who never answers sees five, across three channels, and then the flow ends. The work scales with how hard each person is to reach.
Five things branching makes possible
1. Changing channel when one isn't working
Half of UK mobile users say they are unlikely to answer a call from a number they don't recognise, according to Ofcom. A flow can follow the missed call with a message in the same minute, and stay on whichever channel gets a response. We compare the channels in WhatsApp, SMS, email or a call?
2. Waiting for the customer, not the calendar
A sequence waits a fixed time. A flow can wait for something: a reply, for up to 24 hours, then take one path if it comes and another if it doesn't. Nothing else is sent while it waits, so a reply at 9:15pm is answered at 9:15pm and not trampled by the scheduled 9am email.
3. Deciding on meaning
Events such as "replied" or "no answer" only go so far. An AI decision step reads the conversation and chooses between outcomes you have defined: ready to book, not yet, worth a call from sales. This is how an old lead who writes "the grant has finally come through" gets a phone call from a person, while one who writes "still thinking" gets a note to try again in six months.
One rule keeps this safe: the AI only picks from the outcomes you wrote. It can't invent a new path.
4. Bringing a person in at the right point
A sequence has no place for people except "a task is created". In a flow, a call or reply by a person is a step like any other, with outcomes of its own. The customer reaches your salesperson when they are ready, and the flow carries on from whatever the salesperson decides. Timing matters: field evidence from Taobao in 2026 suggests late handovers are far harder to rescue than early ones. More in AI-to-human handoff, done properly.
5. Stopping
The most underrated branch is the exit. A flow has goals, and reaching one ends everything else for that contact: no reminder to book after they have booked. It has exits for "not interested". And an opt-out on any channel stops all of them. Under UK rules you must not call someone who has told you they don't want your marketing calls, and a sequence that carries on regardless is a compliance problem as well as a rude one.
Does branching get better results?
You will find claims that conditional sequences lift reply rates by 35 to 50%. We could not trace them to anything but vendor blog posts about email and LinkedIn outreach, so we won't repeat them as fact.
What can be said with more confidence:
- Persistence without adaptation has limits. The 2007 Lead Response Management study, based on about 100,000 call attempts, found that after 20 hours each additional dial reduced the chance of making contact. Doing the same thing again stops working quickly.
- People differ in the channel they answer. That is measured, by Ofcom and YouGov among others.
- Irrelevant messages have a cost. Each one sent after a customer has acted is a small reason to opt out.
The honest claim is modest: a branching flow sends fewer wasted messages and has fewer dead ends. Whether that is worth five points of conversion or fifteen in your business is something to measure, which a flow also makes easier, because you can see how many contacts took each path and where they dropped out.
When a simple sequence is fine
Not everything needs a diagram. A linear series is the right tool when there is a single channel, the goal is information and not a decision, and nothing a recipient does should change what comes next. Onboarding tips, a newsletter, a post-purchase care guide. Use the simple thing for the simple job.
How to design a flow that stays simple
- Start from the goal. What counts as done? A booked appointment, an accepted quote, a resolved question.
- List the outcomes of the first step before adding a second. Give every outcome somewhere to go, even if it is "exit".
- Use two or three channels, in a deliberate order. The likeliest first, then a fallback.
- Add waits that listen. Wait for a reply, with a time limit, and branch on both results.
- Mark where a person comes in, and what they will see when they do.
- Cap it. Decide the maximum attempts and the point at which you leave people alone.
- Test it on your own phone. Play the awkward customer: don't answer, reply late, say no.
- Read the report after a fortnight and change one step at a time.
A flow that fits on one screen and can be read aloud, "we call; if there is no answer we WhatsApp; if they reply the AI answers; if they are ready, sales rings", is one your team will trust and improve. That is how flows in Unibound are meant to be used. For a worked example, see the AI lead follow-up playbook.
Questions people ask
What is the difference between a sequence, a cadence and a flow?
A sequence or cadence is a fixed list of touches on a timetable: email on day one, call on day three. A flow, sometimes called a workflow or journey, is a diagram where each step has outcomes and each outcome leads somewhere different. A sequence asks only whether they replied. A flow asks what happened.
Is a drip campaign the same as a flow?
No. A drip campaign sends a pre-set series of messages on a schedule, whatever the recipient does, short of unsubscribing. A flow reacts: it waits for a reply, changes channel, hands over to a person or stops when the goal is met.
When is a simple linear sequence enough?
When there is one channel, one goal and little variation: a three-email welcome series, for example. Once you add a second channel, a phone call or a person who needs to step in, you need branches, because the outcomes multiply.
What is an AI decision in a flow?
A step where an AI reads the conversation so far and picks one of the outcomes you defined, such as ready to book, not yet or wants a person. It lets a flow branch on meaning, not only on events like a reply or a missed call. It should only ever choose from outcomes you wrote.
How many steps should a follow-up flow have?
Usually six to ten. Enough to cover the main outcomes of each contact, a second channel, a wait, and an exit for people who say no. If a flow needs thirty steps, it is probably two flows.
Sources
- InsideSales and Oldroyd, Lead Response Management Study (2007). mortech.com
- Ofcom, experiences of suspicious calls, texts and app messages (2024). ofcom.org.uk
- YouGov, a deep dive into British customer service preferences (April 2025). yougov.com
- Wang et al., agentic AI in customer service: a field experiment (2026 preprint). arxiv.org
- ICO, guidance on live direct marketing calls. ico.org.uk
