A lead sitting at 90 points can be worthless. One at 20 can turn into the order of the quarter.
That is not an argument against lead scoring. It is the reason most point systems in smaller companies go unread after six months: they were built before anyone knew which signals actually precede a signed order in that particular business. After that the system reliably computes something that means nothing.
What follows is the part that comes first: when scoring starts to pay off at all, which signals carry weight, and how you notice that your model is wrong.
Scoring answers a running order, not a verdict
Lead scoring means rating enquiries by points so it is clear who sales calls first. Nothing beyond that.
The restriction sounds modest and it heads off the most common mistake. A score does not say whether someone would be a good customer. It says who you call today when you only have time for three of twenty contacts. Read as a verdict, it filters out people who would have bought later.
Two very different things earn points: who someone is, and what they do. Keeping them apart matters more than any individual number, and it is the first thing that disappears from almost every model.
The threshold: below roughly twenty enquiries a month it does not pay
A point system is a shortcut for a volume you can no longer hold in your head. At fifteen enquiries a month you know your prospects by name anyway – scoring is then effort without insight. The same yardstick governs marketing automation as a whole, for the same reason.
There is a second threshold that gets mentioned far less often: a model cannot be checked until enough deals have closed. Win two orders a month and after half a year you have twelve cases – too few to see whether high scores really do convert more often. Until then any point system is a guess in numeric form.
If what you lack is enquiries rather than order, the work sits earlier. What produces enquiries is covered in the twelve levers of website lead generation and, for the business-to-business case, in B2B lead generation from your website.
Two numbers, not one: fit and behaviour
Every usable model answers two questions separately.
Fit: does this contact belong to the target group at all?
Industry, company size, location, role in the buying process. These change rarely and are known before the first interaction. For a Swiss business the market belongs in here: an enquiry from Hamburg is not worse, it simply matches a different offer. Which roles actually occur in your process comes out of your customer journey map, not out of a template.
Behaviour: how strong is the interest right now?
Page views, downloads, opened emails, meeting requests. These move daily and say something about timing, not about suitability.
Adding the two into a single number is the mistake that makes the whole thing unusable. Seventy points can mean perfect fit and no interest – a contact for the next campaign. Or high interest and the wrong target group – someone writing a dissertation. The two cases call for opposite responses, and the combined number hides which one you are looking at.
Two values side by side cost no extra software. They cost one more column in the CRM.
A signal counts when it cost the prospect something
That is the simplest yardstick for weighting, and it replaces most point tables.
Booking a meeting costs a prospect a fixed slot in the calendar. Requesting a quote costs the willingness to name figures. A repeat visit to the pricing page costs at least the decision to come back. An opened email costs nothing.
Ask that question and nearly everything sorts itself:
- Strong: quote request, booked meeting, repeat visits to pricing or service pages, several people from the same company within a few days.
- Medium: downloading a document that matches the specific problem, replying to an email, visiting a landing page after an ad.
- Weak: newsletter signup, opened email, a click from social media, reading a general blog post.
The most common build error is being generous with weak signals. Give every click five points and within three weeks you have a stack of apparently hot contacts, none of whom wants to buy – and sales stops looking at the list. A model does not lose its effect by computing wrongly. It loses it when nobody believes it any more.
The age of a signal weighs more than its size
A contact who looked at the pricing page the day before yesterday is a different case from the same contact seven months ago. The points are identical, the situation is not.
So every model needs decay. It does not have to be clever: halve behavioural points after thirty days and drop them to zero after ninety, and you have enough for most businesses. Fit does not decay – a company stays in its industry.
Without decay the same thing always happens. After a year the top of the list holds the contacts who have been around longest, not the ones searching now.
MQL and SQL: the threshold belongs to sales
A marketing qualified lead fits the target group and shows interest but is not yet in a buying conversation. A sales qualified lead is far enough along that sales takes over.
Where exactly the line sits cannot be answered from outside the business. There is no score at which a contact becomes ready to buy – 40 and 70 appear in every guide and appear there without derivation.
The number is determined backwards, not estimated forwards. Take the last twenty to thirty won deals and look up the score they carried when sales picked them up. The median of those cases is your threshold. Everything else is guesswork.
And the handover is an agreement between two teams, not an automation. What is worth automating is the notification – an internal message when a contact crosses the threshold. Whether and how anyone acts on it is a human decision.
Under Swiss law a score is a profile
This point is missing from almost every guide, and it is not a detail.
Swiss data protection law defines profiling as "any form of automated processing of personal data consisting of the use of such data to evaluate certain personal aspects relating to a natural person, in particular to analyse or predict aspects concerning [...] personal preferences, interests, reliability, behaviour [...]" (Art. 5 let. f of the Federal Act on Data Protection, FADP, SR 235.1, in force since 1 September 2023).
That is a description of lead scoring, clause by clause. Not by analogy – by definition.
What does not follow matters as much as what does. No consent requirement arises: the FADP does not require agreement for ordinary profiling by private companies. And the rule on automated individual decisions (Art. 21 FADP) does not bite as long as a person decides at the end who to call – it covers decisions taken exclusively by automated means that have a legal effect or significantly affect the person.
Two practical consequences remain:
The privacy notice has to cover it. The duty to inform on collection (Art. 19 FADP) requires the purpose of processing to be stated. If you evaluate behaviour, write that down – as a sentence, not as a clause.
The score is subject to access requests. Under Art. 25 FADP any person may request "the personal data as such" that is being processed. The stored score belongs to that, and so do the notes next to it. The information is free of charge, is normally provided within 30 days, and nobody can waive the right in advance.
Which yields a rule that is easy to remember: do not write anything into a scoring field that you would not read out to the contact. "Budget probably too small" is a sentence you may have to hand over. "No budget stated" is an observation.
How to check whether the model measures anything
A point system never held against outcomes is decoration. The test takes an hour and needs no additional data.
Compare the scores of won deals against lost ones. If the won ones sit clearly higher, the model measures something. If they sit level, it measures nothing – and then the fault is in the criteria, not in the thresholds.
After that, four metrics are worth tracking, and no more:
- share of MQLs that become SQLs
- close rate per score band
- time from first interaction to order
- share of handed-over contacts that sales actually accepts
The last one is the most honest. If sales returns half of what it receives, the threshold is too low – regardless of how good the other numbers look. Which events you can measure in the first place depends on how your analytics is set up. What a contact is allowed to cost follows from customer value and close rate; that calculation sits in Google Ads costs in Switzerland.
Does a smaller company need software for this?
Not at the start.
The first usable version is two fields in the CRM, maintained by hand: fit from 1 to 3, and the last strong signal with its date. That answers the question of who to call today, and it forces you to name the criteria in the first place.
Automation pays off once three things come together: enough enquiries, several people in sales, and signals nobody can note down by hand – repeat page views, for instance. That is the point at which you need a connection between website, forms and CRM, which is what system integrations provide.
In both cases the order is the same: sales process first, then criteria, then the tool. Expensive software on top of an unverified model only computes the wrong answer faster.
Where scoring sits inside lead management
Scoring is one step, not the whole of it. Lead management covers the full stretch: generating enquiries, capturing contacts, scoring them, handing them over, following up – and feeding what happened back into the scoring.
Scoring sits in the middle. It replaces neither the part before it nor the part after.
Before it comes generation. Without enough enquiries you are ranking a set you can see anyway – which is why the threshold sits near the top of this article rather than in a footnote.
After it comes development. A low score rarely means a bad contact. It usually means an early one. How to stay with someone like that over months without crowding them sits in the marketing funnel, including its section on lead nurturing.
Have all three parts and you are running lead management. Have only the middle one and you are keeping a sorted list.
Frequently asked questions
How does lead scoring work?
Each contact earns points for two things: how well they fit the target group, and how actively they are engaging with the offer right now. The two values produce a running order for sales – not a statement about whether someone would make a good customer.
At how many enquiries does lead scoring start to pay off?
As a rule of thumb, from around twenty enquiries a month. Below that you know your prospects by name and scoring costs more time than it saves. More important than the number of enquiries is the number of closed deals: without them you cannot check whether the model holds.
How many points does a lead need to qualify?
There is no universal number. Derive it backwards from the last twenty to thirty won deals: what score did they carry when sales took over? The median of those cases is your threshold, and it deserves recalculating every few months.
What is the difference between MQL and SQL?
A marketing qualified lead fits the target group and shows interest but is not yet in a buying conversation. A sales qualified lead has crossed the agreed threshold, so sales takes over. Where that line sits is agreed jointly by marketing and sales.
Is lead scoring lawful in Switzerland?
Yes. Scoring is profiling under Art. 5 let. f FADP but needs no consent. What is required is transparency about the purpose in the privacy notice (Art. 19 FADP) and disclosure of the stored score on request (Art. 25 FADP). For a specific setup, a legal review belongs in the plan.
What is the difference between lead scoring and lead management?
Lead management is the whole run from enquiry to order: capture, score, hand over, follow up. Lead scoring is the scoring step inside it. Score only and you have a sorted list; connect the steps and you have a process.
Does lead scoring work outside B2B?
In principle yes, in practice rarely. It pays off where several weeks and several touchpoints sit between first contact and purchase – the rule in B2B, the exception in direct sales.
Conclusion: the score orders your day, it does not decide it
Lead scoring for smaller companies is not a points collection. It is an answer to one very small question: who do I call first today?
What it takes is little. Two separate values instead of one combined number. Weighting by effort spent, not by whichever signal is easiest to capture. Decay, so old contacts do not drift to the top. And a threshold derived from closed deals rather than from a guide.
That leaves the sentence a model has to earn: if your won deals carried no higher score than your lost ones, the system measures nothing – and then it needs changing, not extending. How that scoring fits into one continuous path from enquiry to order is what ALPENIQ Growth is built around.