Mortgage broker follow-up automation

An unreturned fact find and a missing payslip are the same problem wearing different clothes. Something was asked for, the client meant to do it, and it slipped. Nudging works. The question is how many, how far apart, through which channel, and when to stop.

Most firms get the wording right and the schedule wrong, which is why sequences either fizzle out after one polite reminder or grind on until someone unsubscribes.

Design the schedule around how people actually forget

The forgetting is not malicious and it is not evenly distributed. It happens in the first forty-eight hours, when the request competes with everything else in an inbox, and it happens again after about a week, once the task has fallen out of mind entirely.

A schedule that reflects that looks roughly like this. The original request. A short prompt after two working days. A change of channel after four or five. A direct human contact after a week or so. Then a decision.

What does not work is a daily drip. Daily reminders about the same item teach the recipient that your messages are ambient, and they are the most reliable way to generate an opt-out from someone who was never annoyed with you in the first place.

Nor does a single reminder followed by nothing, which is what most firms actually do. The item then sits in a pipeline report marked outstanding for six weeks while everyone assumes someone else is on it.

Change the channel, not just the wording

The second reminder should not be another email with a firmer tone. It should be somewhere else.

Email is fine for the request itself, because it carries links and detail. A text message is far better as a nudge, because it is short, it is read, and it costs the client nothing to reply with a question. A voice call is better again at the point where the sequence has failed twice, because the reason for the silence is often something the client has not wanted to put in writing.

That last point is worth dwelling on. A client who has gone quiet about bank statements is quite often embarrassed about what is on them. A client who has not returned a fact find is sometimes having second thoughts, or has had a change in circumstances they have not told you about. No amount of automated chasing will surface either. A two-minute call will.

So the sequence's real job is not to extract the document. It is to get to the phone call quickly when the document is not coming.

Stop rules, which matter more than the sequence

Every automated chase needs conditions that end it, and they need to be built at the same time as the sequence rather than added after the first complaint.

• Stop when the thing arrives. Obvious, and routinely broken, because the sequence is watching a task status that nobody updates when the document is uploaded. Tie the exit condition to the actual upload event if you possibly can.
• Stop when the client replies at all. Any reply means a human is now in the conversation, and continuing to send scheduled messages alongside a live exchange is the single most damaging thing these systems do.
• Stop on request, everywhere, immediately. An objection to further contact has to be honoured across every system, including whatever sends your newsletter. Under UK GDPR that is not optional.
• Stop after a defined number of attempts regardless of outcome. Three or four is plenty. After that it is a person's job or it is a case that should be marked dormant.
• Suppress entirely for clients who have disclosed illness, bereavement or financial difficulty. They need a named person, not a cadence.

Escalate to something that has authority

When the automated attempts are exhausted, the case has to land somewhere with a name on it.

The usual failure is that it lands in a queue everybody can see and nobody owns. Assign it to an individual with a due date, and report on the age of those tasks weekly. A pipeline where the oldest outstanding item is nine days is a healthy one. Where it is nine weeks, the sequences are producing activity rather than progress.

It also helps to give the escalating adviser something more than a reminder to chase. A note of what has already been sent, and when, saves a client from being asked as though nothing has happened, which is a specific and avoidable irritation for someone who has already had four messages.

Reduce the need for nudges in the first place

The best sequence is the one that fires less often, and the fastest wins are usually upstream of it.

Ask for fewer things at once. A request for seven documents produces partial completion and a long tail. Two or three at a time completes faster.

Say why each item is needed. People act on requests they understand and defer ones they do not.

Make the doing easy. A link that opens on a phone, accepts a photo, and does not require a password reset will outperform any reminder schedule. A meaningful share of document delay is a login problem rather than a motivation problem.

Set the expectation at the appointment. A client who has been told in person that these five items are needed by Friday, and why, needs far less chasing than one who receives the same list cold two days later.

What to watch

Track how long each item takes from request to receipt, and look at the distribution rather than the average. Averages hide the cases that took a month, and those are where the pipeline value is leaking.

Then check how many items were resolved by the automated stages versus by a person picking up the phone. If almost everything needs the call anyway, your sequence is theatre, and the honest response is to shorten it and get to the call sooner.