Two automations account for most of the return in a sales stack: scoring enquiries so the right ones get attention first, and following up on the ones that go quiet.
Both are straightforward to build and both are commonly built badly, in ways that are worth naming before you commission either.
Qualification scoring is only as good as the standard beneath it
A scoring model assigns weights to attributes. If nobody has written down which attributes matter, the weights encode whoever configured the tool.
The prerequisite is a one-page, sales-signed definition covering fit, trigger, budget signal and authority. With it, scoring is mechanical. Without it, sales overrides the score, trust collapses, and the system becomes decoration inside three months.
This is the single most common reason lead scoring implementations are abandoned.
If your team routinely overrides the score, the model is wrong — not the team. Log the overrides and use them to correct the weights rather than asking people to trust the number.
Enrichment matters more than the model
Most scoring runs on what somebody typed into a form, which is thin and often wrong. Company size, industry and technology stack are frequently blank or guessed.
Enriching at the point of capture — resolving the company from the email domain and appending firmographic data — changes what the model has to work with. A simple model on good data outperforms a sophisticated one on form entries every time.
It also improves the routing decision immediately, before any scoring sophistication is added.
The follow-up sequence nobody builds
Someone enquires, you reply, they go quiet. In most businesses nothing happens next, because the moment after a conversation stops belongs to nobody.
A prospect going quiet is rarely a decision. It is usually a busy week, a delayed internal conversation, or an email that arrived at a bad moment.
A four-to-six touch sequence over three to four weeks, with genuinely different content in each touch, recovers a meaningful share of these. It requires no new demand and no new budget.
A non-response sequence that works
Each touch adds something. None of them says 'just bumping this'.
- 1
Day 2
Short reply-to-thread nudge, restating the single next step.
- 2
Day 5
Send something useful — a relevant guide, not a chase.
- 3
Day 10
Address the objection you suspect is actually blocking it.
- 4
Day 17
Offer a lower-commitment option: a shorter call, a written answer.
- 5
Day 25
Close the loop honestly — 'assuming timing is wrong, I will stop here'.
- 6
Stop
Sequence ends. No indefinite drip, no restart without a new trigger.
Stop conditions are the whole difference
The line between persistent and irritating is entirely a matter of stop conditions, and they are the part most commonly left out.
A sequence must stop on: any reply, a meeting booked, an unsubscribe, and the end of the sequence itself. It must never restart automatically without a new inbound signal.
The failure everyone has experienced as a recipient is a sequence that continued after they replied. That single defect does more brand damage than the entire sequence recovers.
Personalisation, honestly
Automated personalisation works when it references something real — a funding round, a job posting, a published change. It reads badly when it references something generic dressed as insight.
"I noticed you're in manufacturing" is not personalisation; it is a merge field with a friendly tone, and recipients recognise it instantly.
The practical rule: personalise on a signal you could defend in conversation, or write a good generic message instead. A well-written generic message outperforms transparently fake personalisation.
What to measure
- Score-to-conversion correlation — do high-scoring leads actually convert better?
- Override rate — how often does a human disagree with the score?
- Sequence recovery rate — what share of non-responders re-engage?
- Unsubscribe rate per sequence — the early warning that tone is wrong.
- Time from enquiry to first human contact, still the dominant variable.