Almost every team trying to fix a bad reply rate reaches for the same three levers: rewrite the subject line, change the offer, send more. The lever that usually matters more is the one nobody pulls, which is sending to fewer people.
Bad-fit leads are not a neutral cost. They are actively expensive, and not only in wasted sends.
What a bad-fit lead actually costs
- The send. Trivial, fractions of a cent. This is the cost everyone counts and the one that matters least.
- The deliverability. This is the real number. Poor-fit recipients delete without opening, never reply, and occasionally mark as spam. Inbox providers read that as evidence your mail is unwanted, and they apply the conclusion to your domain, not to that campaign. A bad list degrades the placement of your good campaigns.
- The mailbox. Sustained low engagement and elevated complaints is how mailboxes get flagged and domains get burned. Replacing them costs money and three weeks of warmup.
- The rep’s time. Every unqualified reply that turns into a call is an hour nobody gets back.
Put together, sending to the bottom third of a list frequently costs more than it earns, not in send fees, but in the placement penalty it applies to the top two-thirds.
ICP scoring vs lead scoring
These get conflated and they are not the same mechanism.
| Lead scoring | ICP scoring | |
|---|---|---|
| When it runs | After contact | Before enrolment |
| What it measures | Engagement, opens, clicks, replies | Fit, do they look like your best customers |
| What it decides | Who to prioritise | Whether to send at all |
| Protects deliverability | No, the sends already happened | Yes, bad fits never get sent to |
Lead scoring tells you who to call back. ICP scoring stops you from emailing people who were never going to buy. Only one of them protects your sender reputation.
Building a score that actually predicts fit
1. Start from closed-won, not from intuition
Export your last 50 closed-won accounts and find what they genuinely had in common. This almost always contradicts the ICP written on the strategy deck. Teams routinely discover their best customers are smaller than they target, in an industry they deprioritised, or all running one specific tool.
2. Weight three layers
Firmographic, industry, headcount, revenue band, geography, growth rate. The baseline, and the easiest to get right.
Technographic, the tools they already run. Frequently the strongest single predictor, because it tells you whether they have the problem you solve and whether they buy software to solve problems at all.
Timing, hiring for a role adjacent to your product, recent funding, new leadership in the buying function, rapid headcount growth. Fit tells you whether they should buy; timing tells you whether they will buy this quarter.
3. Score negatives explicitly
The most useful part and the most commonly skipped. Competitors, current customers, companies in a hiring freeze, industries you have never closed, these should score down, not merely fail to score up. A firmographically perfect account that has churned from you twice is not a good lead.
4. Set the threshold by capacity
Do not pick 70 because it sounds right. Work out how many prospects your team can genuinely work in a week, then set the threshold so the volume above it matches that number. If your capacity is 300 a week and 900 clear the bar, raise the bar.
Where this sits in the workflow
The order matters more than the scoring model. Fit has to be evaluated before enrolment, not as a report you read afterwards.
- Search, build the list in Lead Finder using firmographic and technographic filters. Filters do the coarse work.
- Score, ICP Score rates each lead against your profile. Filters get you to “plausible”; scoring ranks within it.
- Gate, only leads above your threshold can be enrolled. This is the step that protects deliverability, and it has to be enforced by the system rather than by discipline.
- Personalise by tier, high scorers get deeper AI personalization and a multichannel cadence; mid scorers get a lighter touch.
- Feed results back, when a deal closes, that account’s attributes should strengthen the model. When high scorers consistently do not reply, the model is wrong.
That last loop is why scoring living in the same platform as your sequencing matters. If your data sits in one tool and your replies in another, you can build the score once but you can never improve it.
What to expect
Teams that put a real fit gate in front of enrolment usually see the same pattern: send volume drops, reply rate rises more than proportionally, and deliverability metrics improve within a few weeks because engagement per send went up. The uncomfortable part is that it means sending to fewer people, which feels like doing less outbound. It is doing less bad outbound.
Next: how to define an ICP properly, or building a targeted lead list.
Frequently asked questions
What is ICP scoring?
ICP scoring rates each prospect against your ideal customer profile before they enter a sequence, producing a single fit score from firmographic, technographic and behavioural signals. It is different from lead scoring, which usually measures engagement after contact. ICP scoring happens first, so you can decline to send at all.
Why does sending to bad-fit leads hurt deliverability?
Poor-fit recipients ignore, delete or report your email at far higher rates than good-fit ones. Inbox providers read those signals as evidence that your mail is unwanted and adjust placement for your whole domain accordingly. A bad list degrades the deliverability of your good campaigns too.
What signals should an ICP score use?
Start with the firmographics that describe your best existing customers, industry, headcount, revenue band and geography, then layer technographic signals such as tools they already run, and timing signals such as hiring activity or funding. Weight them by what your closed-won accounts actually had in common, not by what feels intuitive.
What is a good ICP score threshold?
Set it by capacity rather than by an abstract number. Work out how many prospects your team can genuinely work per week, then set the threshold so the volume above it matches that. Most teams find their reply rate improves sharply when they cut the bottom third of their list.
Does Salesforge or Apollo have ICP scoring?
Neither offers fit scoring as a gate before enrolment in the way described here. Apollo has filters, and most platforms have engagement-based lead scoring after contact. Scoring fit before the first send is a different mechanism and a meaningful gap in most outbound stacks.