AiformleadsLead scoring: how to score and qualify your leads automatically from the form
Team at AiformleadsWritten with AI assistance21 min readIn summary
- Lead scoring is a sum of points, not a guess: each answer on your form is worth a number, the total decides who you call first. The difficult part is not the formula, it’s choosing which questions score.
- Scoring without tagging leaves the job half done: an 85 tells you the lead is good; a tag like
high-budgetorno-decisiontells you why and what to say on the call. - Thresholds are evaluated from highest to lowest: the first one to reach its minimum wins. With the default distribution from Aiformleads (Hot ≥ 80, Warm ≥ 50, Cold ≥ 0), a lead of 80 points is Hot and one of 79 is Warm. You decide that boundary and you will move it.
- An uncalibrated model is worse than having no model: if 70% of your leads come out Hot, your scale is broken. Measure it with the distribution of scores and with the actual drop-off per step, not with the percentage promised by a brochure.
- Scoring people has legal limits: legal basis, prior information, and right to erasure are not optional (Regulation (EU) 2016/679). This article explains how the tool works, it does not substitute the judgment of your legal advisor.
If you receive forms and then manually decide which deserve a call, you are doing lead scoring: you are just doing it mentally, late, and without it being recorded. This guide explains how to turn that judgment into a model of points and tags that applies automatically, at the time of submission, and how to know if it works. The examples are built on Aiformleads, the lead capture mini app from AIsToHire (Aiformleads sheet and demo), and everything described here has been verified in its version from October 2026: there is no promised function that the tool does not perform.

What exactly is lead scoring?
Lead scoring is assigning a numerical score to each contact based on what they have answered and how they have behaved, to prioritize whom you attend to first. In its most practical form, each response option on your form carries an associated number of points; when the form is submitted, the points from the chosen options are summed, and the total falls within a range (Hot, Warm, Cold) that determines commercial priority.
It is not artificial intelligence, and it does not need to be. It is arithmetic based on criteria that you already apply mentally:
- Someone with a budget is worth more than someone without one.
- Someone who needs something this month is worth more than someone who "is looking at options".
- Someone who can sign is worth more than someone who has to ask three people.
The difference between doing it mentally or with a model is that the model applies to all leads, in the zero second, in the same way, and leaves a trace for you to review later.
Explicit and implicit scoring: the distinction that matters
| Type | Where it comes from | Example | Does a form cover it? |
|---|---|---|---|
| Explicit | What the person declares | Budget, company size, urgency, role | Yes: it is exactly what a form collects |
| Implicit | What the person does | Pages viewed, emails opened, downloads | Partially: needs behavioral tracking |
This article focuses on explicit scoring, which is what you can set up today with a form and which is the least likely to break: a declared data point does not depend on whether cookies work or whether the visitor uses the same browser.
Where do the points come from? From the options the person chooses
In Aiformleads, points live within each option of single and multiple selection questions: next to the text of each option, there is a numeric box where you write how many points it is worth. When the form is completed, the backend goes through the answers and sums the points of all the selected options. The result is saved in the lead as score.

Three practical consequences of that design, stated plainly:
- Only questions with options score. An open text field, a slider, or a date do not add points by themselves. If you want the budget to score, ask it in ranges (options), not with an open number.
- Points can be negative. Writing
-20in the option "I am a student / personal project" is perfectly valid: it sums as is. Subtracting is the cleanest way to prevent a bad lead from slipping through by accumulation. - In multiple selection, all selected are summed. If a person chooses three services and each is worth 10, they get 30. Keep this in mind or set a maximum number of selections in the field.
What an open or numeric field cannot do for points, it can do for tags: tagging rules allow numeric operators (greater than,less than or equal…) on any response. A budget slider does not add points, but it can trigger the taghigh-budget.
How to build your scoring table in five steps
The method that works is the opposite of what seems logical: don’t start with the questions, start with the clients you already have. Set aside an hour, open your list of closed clients from the last year, and do this.
- Write down the five characteristics of your best client. Not "large companies", but verifiable things: "invoices over €2M", "has a marketing manager on staff", "contacted us with a deadline".
- Write down the three characteristics of the client that made you lose money. The ones you swore you wouldn’t repeat. Those will be your negative points.
- Turn each characteristic into a question with closed options. If you can’t convert it into 3-5 options, don’t score it: ask it anyway, but as an open text for context.
- Distribute 100 points among the questions, not among the options. First, decide how much each question weighs (e.g. budget 40, urgency 25, role 20, size 15) and then distribute the weight of each question among its options. This forces you to prioritize.
- Set the thresholds with your hand on your heart: At how many points would you call within an hour? That is your Hot.
Five scoreable questions is a good ceiling to start. Each extra question reduces the likelihood that the form will be completed, and a perfect lead that doesn’t come in scores nothing.
A complete example model: four questions, 100 points
This is a real and usable model for a B2B agency or consultancy. Copy it, change the texts, and you will have a defensible starting point.
| Question | Option | Points |
|---|---|---|
| What budget do you handle? (weight 40) | More than €10,000 | 40 |
| €5,000 – €10,000 | 30 | |
| €1,000 – €5,000 | 15 | |
| Less than €1,000 | 0 | |
| When do you need it? (weight 25) | This month | 25 |
| This quarter | 15 | |
| Later on | 5 | |
| I’m just researching | −10 | |
| What is your role in the decision? (weight 20) | I decide | 20 |
| I decide with my team | 12 | |
| I recommend, someone else decides | 6 | |
| None, it’s for a class project | −20 | |
| How many employees are you? (weight 15) | More than 100 | 15 |
| 51 – 100 | 12 | |
| 11 – 50 | 8 | |
| 1 – 10 | 3 |
With this table, the maximum score is 100 and the minimum is −30. Two examples of how it falls:
- Marketing director of a company with 120 people, €12,000 budget, needs it this month:
40 + 25 + 20 + 15 = 100→ Hot. - Freelancer with €800 who "is researching" and recommends another:
0 + (−10) + 6 + 3 = −1→ Cold.
Notice what the negative points do: without them, the second lead would have scored 9 points just for filling it out; with them, it explicitly falls below zero.
Thresholds: turning the number into a decision
A loose score is worthless; what matters is the range it belongs to. In Aiformleads, you define ranges with a name, a minimum, a color, and an emoji, and they are evaluated from highest to lowest: the first whose minimum is reached wins. The default ranges are:
| Range | Minimum | Suggested operational meaning |
|---|---|---|
| 🔥 Hot | 80 | Call the same day, by phone, by a person |
| ⚡ Warm | 50 | Personal email within 48 hours with a concrete proposal |
| ❄️ Cold | 0 | Automatic content sequence, without commercial effort |
Three details worth knowing before touching them:
- The lowest range must have a minimum of
0or negative, or leads that score below will remain uncategorized and will not appear in the score distribution. - Each range can have an associated tag, so that scoring and tagging are synchronized without writing a separate rule.
- Alerts allow for minimum scoring: you can set that the alert to the team, the message to the chat, or a webhook only triggers if the lead exceeds X points. It’s the difference between your team looking at notifications and silencing them after a week.
Scoring isn't enough: the label is what makes a lead actionable
A number organizes; a label explains. 92 points tells you to call; high-budget + urgent + current-competitor tells you with what argument to start the call. That's why automatic labeling is not just an embellishment of scoring: it is the other half of the same work.

A labeling rule is a IF… THEN: if certain conditions about the responses are met, certain labels are assigned. The conditions are grouped with the usual logic —within a group all must be met (AND), between groups it is enough that one is met (OR)— and accept ten operators: is equal to, is not, contains, does not contain, greater than, greater than or equal, less than, less than or equal, is empty, is not empty.
What makes the difference between a helpful labeling and one that clutters the inbox are these five controls:
| Control | What it serves for | Example |
|---|---|---|
| Order of the rules | They are evaluated in the order you place them | Putting the discard rule first prevents labeling a student as hot |
| Stop if applies | Cuts the evaluation at the first rule that is met | A discarded lead does not continue accumulating business labels |
| Default label | Assigned only if no rule has applied | unclassified: no lead is left orphaned in the inbox |
| Exclusion rules | Two labels that cannot coexist | current-client and new-lead: if both appear, one is removed |
| Hierarchy (parent label) | When assigning a child, its parent is added, up to three levels | madrid adds spain; filtering by spain also retrieves Madrid |
And a metric that almost no one has and that comes out on its own here: coverage. The editor calculates what percentage of all the options from your questions appears in some rule. If your coverage is 30%, seven out of ten possible responses are not saying anything about the lead. It is not necessary to reach 100%, but a very low coverage usually means you are asking questions that you then do not use.
Contactability: 90 points with a disposable email are worth zero
An excellent lead that you cannot call is not a lead. Therefore, in addition to the qualification score that you define, Aiformleads calculates a contactability from 0 to 100 with a fixed formula that does not depend on your model: it measures whether the contact data is useful for anything.
| Signal | Points |
|---|---|
| Valid format email | 40 |
| The email domain is not disposable | 10 |
| The domain is not a free email (gmail, hotmail, outlook, yahoo, icloud) | 10 |
| Phone with 9 digits or more | 20 |
| Name with more than 2 characters | 10 |
| One more answered question (up to 10 questions) | 1 each |
These are two numbers with different jobs and it is advisable not to mix them:
- The score answers "Am I interested in this lead?".
- The contactability answers "Can I talk to him?".
A lead with a score of 90 and contactability of 40 (Gmail email, no phone) is not a bad lead: it is a lead that is missing a piece of information. The correct action is not to discard it, but to ask for the phone. Both values travel together in the lead's record and in the export.
How do I know if my model is poorly calibrated?
A lead scoring model is poorly calibrated when its distribution does not resemble the reality of your business. The clearest signal is the distribution by qualification: if more than 25-30% of your completed leads come out Hot, you do not have a great product, you have the threshold too low. If in three weeks not a single Hot has come out, you have it too high.
Four alarm signals and what to look at in each:
| Signal | What is happening | What to review |
|---|---|---|
| Almost all leads are Hot | Low threshold or inflated points | Distribution by qualification; raise the minimum of Hot |
| No Hot ever comes out | Theoretical maximum below the threshold | Add up the maximum points of your table and compare them with the threshold |
| The Hot do not buy and the Cold do | You are scoring what does not predict the sale | Response map: what option did those who closed really choose |
| Many leads without label | Incomplete rules | Coverage of rules and default label |
For the last one, the panel calculates a response map: for each question with options, how many completed leads chose each option and what percentage it represents. It is the quick way to discover that 80% of your leads choose the lowest budget range and that, therefore, that question is hardly discriminating anything.
Don't believe the percentages in brochures: measure your own funnel
Anyone who promises you that a multi-step form converts X % more than a one-page form is selling you the average of other companies. The only thing that matters is where your people drop off, and that is measured with the funnel of your own form, not with an external study.

The Aiformleads funnel counts, for each step, how many sessions reached it and calculates the drop-off with an explicit formula:
drop-off at step = (arrivals at previous step − arrivals at this step) / arrivals at previous step × 100
Next to that percentage, you will see the average time each person spent at the step. The combination is very informative:
- High drop-off + high time → the question is difficult or uncomfortable (budget, phone). Move it later or make it optional.
- High drop-off + low time → people get scared when they see the step. It is usually a form that suddenly asks for too many fields.
- Low drop-off + very high time → the question is misunderstood, but people insist. Rewrite the statement.
If you test two versions of the form, the panel also shows the 95 % margin of error for each variant, calculated as 1.96 × √(p(1−p)/n) based on the conversion rate. It serves for the only thing that matters in an A/B test: knowing if the difference you see is real or just noise. If variant A converts at 22 % ± 6 and B at 25 % ± 6, you still haven't gained anything: you lack data.
Six mistakes that ruin a lead scoring model
- Scoring intention instead of fit. "I want to start now" is free to say. Budget and role are not.
- Not using negative points. Without deductions, the score only goes up, and after a few weeks everyone is Hot.
- Asking what you won't use. Each question costs drop-off. If a question does not enter the score or a tagging rule or the business conversation, delete it.
- Setting it up and not looking at it again. A model is reviewed when the offer, price, or market changes. At a minimum, once a quarter.
- Confusing the score with a promise. The score orders a work queue; it does not predict a sale or replace the conversation.
- Scoring without having prepared the legal part. Collecting data to classify people requires legal basis and prior information; that is prepared before publishing, not after.
What should be considered legally when scoring and tagging leads?
Scoring and tagging a person is processing their personal data, so it falls squarely under Regulation (EU) 2016/679 (GDPR) and, if you send commercial communications in Spain, Article 21 of Law 34/2002 (LSSI-CE). In practice, this means having a legal basis for processing, informing why you collect the data before collecting it, and being able to address the rights of the person. This is not legal advice: Aiformleads is a form tool, not a legal service, and the responsibility for processing lies with the person who publishes the form.
What the tool provides to comply with that part, verified in October 2026:
| Usual obligation | What the app offers |
|---|---|
| Informed and separable consent (arts. 6.1.a and 7 GDPR) | Granular consents at the contact step, stored with the lead |
| Prior information (art. 13 GDPR) | Consent text and configurable links in the form |
| Right of access (art. 15 GDPR) | Export of all data from a specific lead |
| Right of erasure (art. 17 GDPR) | Deletion of the lead and their associated data on demand |
| Minimization and retention periods (art. 5.1.e GDPR) | Configurable retention in months: the oldest leads are automatically deleted |
Regarding Article 22 of the GDPR (automated decisions): a score that only serves to order the call queue of a salesperson is usually not an "automated decision with legal or similar effects," because the real decision is made by a person. If your case is one —for example, automatically denying a service based on the score— consult your advisor before setting it up. The tool does not decide anything for you: it calculates, tags, and leaves it in your inbox.
How to set this up in Aiformleads, tab by tab
These are the actual steps in the app, with the names you will see on screen.
- Create the form. Start from scratch or from one of the 10 built-in templates. A form is a cover page, the steps of questions you need, and a final contact step.
- Add the scoreable questions in the Steps tab. There are 13 types of fields (short and long text, single and multiple selection, scale and NPS, slider, matrix, date, phone, address, file, signature, and calculation). For scoring use single or multiple selection and write the points in the numeric box of each option.
- Activate scoring and define the ranges in the notifications section: name, minimum, color, emoji, and, if you want, an associated label.
- Write the tagging rules in the Tags tab. First define the glossary (name, color, category, parent label) and then the rules with their conditions. Order the rules, check "stop if applies" on the discard ones, and set a default label. Watch the coverage indicator.
- Check it in the preview. It is the same renderer that the person filling out the form will see, so what you see is what there is. In preview mode, no session is created, emails are not sent, and webhooks are not triggered.
- Publish it. There are seven ways: public link, iframe, script with automatic height, modal window, sidebar, floating button, and downloadable QR code for printed materials.
- Manage the inbox. Filter by label, by lead status (new, contacted, in process, qualified, converted, discarded) or by score range; sort by score; add notes; act in bulk. Export to CSV or Excel selecting columns, including score, rating, contactability, labels, country, device, referrer, utmsource, utmmedium, utm_campaign, and variant.
- Review and recalibrate. Go back every month to the distribution by rating and to the funnel, and move the thresholds.
What Aiformleads does and what it doesn't
Honesty here saves disappointments. This is what is included and what is not in the scope of the app, as of October 2026.
| Does it do | Does not do |
|---|---|
| Score with the points of each selection option | Scoring by behavior on your website (pages viewed, time on page) |
| Rate by ranges and tag with rules when the form is submitted | Decide for you: the call is still made by a person |
| Calculate contactability with a fixed formula | Verify with the operator if a phone exists |
| Funnel, drop-off by step, response map, sources, countries, devices, and variants | Complete multichannel attribution of your marketing |
| Notifications to the team, confirmation to the lead, and conditional email sequences | Be a complete CRM with opportunities and sales forecasting |
| Outgoing webhooks and REST API with its own key | Bidirectional synchronization with Google Sheets via OAuth |
| Export to CSV and Excel with column selector | Own domain for the public form |
Frequently Asked Questions
How many points do I need to give to each answer?
There is no universal table. The method that works is to first distribute 100 points among your questions based on what truly predicts a sale in your business (for example: budget 40, urgency 25, role 20, size 15) and then distribute the weight of each question among its options. Start with round numbers, let a few weeks pass, and adjust by looking at which leads ended up purchasing.
Can I subtract points instead of adding them?
Yes. In the points box of an option, you can write a negative number and it will be added as is. This is the best way to ensure that profiles you are not interested in — "it's for a class project", "I'm just researching" — fall below the threshold instead of accumulating points simply for filling out the form.
What about questions that are not multiple choice?
They do not accumulate points: the score is only calculated with the options chosen in single or multiple selection questions. However, any answer, including sliders, numbers, and texts, can trigger a tagging rule, because the conditions allow for numerical and text operators. If you need something to score, turn it into closed options.
How many questions should the form have?
As few as possible that allow you to decide. Each question adds abandonment, and that you can measure: the funnel tells you exactly at which step people drop off and how much time they spent on it. Start with four or five scoreable questions, monitor abandonment by step for a couple of weeks, and trim those that are not contributing anything to the score or the tags.
When is the score calculated, upon answering or upon submission?
When the form is completed. At that moment, the backend validates the submission, sums the points of the chosen options, determines the scoring tier, applies the tagging rules, resolves the final screen, and triggers the corresponding alerts and webhooks. Partial answers from an abandoned session are saved, but that session is not scored or tagged.
How often should the lead scoring model be reviewed?
At least once a quarter, and whenever you change your offer, pricing, or market. The quick review consists of two data points: the distribution by rating (if almost everything comes out Hot or none at all, the threshold is wrong) and the response map (what the leads who ended up purchasing really chose).
Can I notify the team only when a good lead comes in?
Yes. Team notifications, chat messages, and webhooks allow for a minimum score: below that threshold, they do not trigger. This is the way to ensure that notifications continue to mean something after the first month instead of ending up silenced.
Can leads be exported to another tool?
Yes. You can export to CSV or Excel by choosing which columns to take (score, rating, contactability, tags, status, source, UTM, variant…), send each lead to your systems with outgoing webhooks, and query forms, leads, and statistics from a REST API with your own key.
Is it legal to score people who fill out my form?
Classifying leads is a processing of personal data and is subject to the GDPR: you need a legal basis, inform before collecting the data, and be able to address the rights of access and deletion. The app provides the technical pieces (granular consents stored with the lead, export and deletion of a specific lead, configurable retention in months), but the responsibility for processing is yours, and the specific configuration should be reviewed with your legal advisor.
In a nutshell
Aiformleads is the mini app from AIsToHire that turns a multi-step form into already scored and tagged leads at the moment of submission, and it's at Aiformleads — Forms that capture leads and classify them automatically. Lead scoring is not magic: it's deciding once, in writing, what makes a good customer, and letting that decision apply automatically to everyone who comes knocking at your door.
Set up your lead scoring model in Aiformleads: create the form, assign points to each option, define your ranges, and start receiving each lead already scored and tagged in your inbox.
Set up your lead scoring model in Aiformleads: create the form, assign points to each option, define your ranges, and start receiving each lead already scored and tagged in your inbox.Keep reading
View all articles →
Almost half of your form submissions are bots: how to detect fake leads in 2026

Why Your Car Insurance Increased in 2026: Average Premium of €1,412 and Historic Record

Discussing via WhatsApp or talking in person: what science says about couple conflicts through messages
