Ask ten businesses what a qualified lead is and you will get ten answers, most of which are a feeling. Interested. Warm. Serious. Ready. None of these can be counted, which means none of them can be improved, and it means two people in the same office will disagree about the same lead.
Start with the facts, not the score
A qualification set is a list of the things that actually decide whether a sale can happen. For a property business it looks roughly like this, and every industry has its own version:
| Slot | Why it matters |
|---|---|
| intent | Buy, rent, sell, invest. Everything downstream branches on it. |
| budget | The single biggest time waster when it is missing. |
| area or scope | Which of the things you sell are even relevant. |
| specification | Bedrooms, model, treatment, seat count, whatever “the right one” means here. |
| timeline | “This month” and “sometime next year” are different businesses. |
| financing | Cash, approved credit, or not started. The question humans skip. |
| contact | A name plus a way to reach them. |
| next step | A meeting booked, or availability captured. |
Note what is not on that list: enthusiasm, engagement, time on site, or how many messages they sent. Those are proxies for interest. Interest is not the constraint. A delighted visitor with no budget and no timeline is a pleasant conversation and nothing else.
Then count, and weight
The score is the count of filled slots, weighted by which ones your business cares about most. That is genuinely all it is. A workable set of bands:
- hot, intent, budget, area, timeline, financing and contact
- warm, intent, budget, area and contact
- cold, anything less
The specific rule matters far less than the fact that it is written down and applied the same way to every conversation. What you get in return is a number that can be reconstructed: anyone can look at a warm lead and see precisely which fact is missing, and go and get it.
Why the model must not do the scoring
This is the part that matters if there is an AI anywhere in the process.
A model asked to assess the quality of a conversation it has just had is being asked to mark its own homework, and it will award itself a good grade. It is agreeable by construction. Ask it whether that chat went well and it will tell you the lead seems very interested, because the transcript reads as friendly and it has no way to know that friendly and financed are different things.
So split the job. Extraction is a job for the model, reading a messy human sentence and working out that “we’ve got about four hundred to spend, maybe a bit more for the right one” is a budget. Scoring is a job for arithmetic. Count the slots. Apply the weights. The output is then a fact about what was established, not an opinion about how the conversation felt.
Why nobody trusts the lead score they already have
Most sales teams have a lead score and most of them ignore it, and the reason is not stubbornness. It is that they cannot see inside it. A number between one and a hundred, with no explanation, gives a rep two options: obey it, or dismiss it. After two bad calls from a ninety-scoring lead, everyone dismisses it, permanently.
A slot-based score can be argued with, “it says warm because we never got a timeline” , and a score that can be argued with is a score that gets used. Auditability is not a compliance feature here. It is the thing that makes the score real.
The questions your team skips
Budget. How it is being paid for. Who else has to approve it.
These are the three most predictive facts in most sales, and the three most likely to be left until “later in the process”, which frequently means never. They feel intrusive to ask in a first conversation. So they get deferred, and a week later somebody discovers there was never any money.
There is a reasonable case that this is one of the better arguments for automating the first conversation: an agent asks the awkward question in the same neutral tone as the easy one, every time, without the small social cost that makes a human put it off. Not because it is braver, because it is not embarrassed.
Only two bands should ever reach a person
Hot and warm go to a human. Cold stays in the conversation and keeps working, because a cold lead is not a bad lead; it is an unfinished one. The visitor who has not given a timeline yet has not said no, they have simply not been asked in a way that made answering worthwhile.
That distinction is worth guarding, because the alternative, sending everything to a salesperson and letting them triage, is how the highest-value enquiry ends up buried under forty that were never going anywhere. Independent mystery-shop work such as Pied Piper’s in the motor trade keeps finding the same shape: the enquiries that take longest to answer properly are the ones most likely to be dropped, and they are usually the expensive ones.
Questions people ask about this
What makes a lead qualified?
A lead is qualified when a specific, named set of facts has been established: typically intent, budget, area or scope, specification, timeline, how it is being paid for, a contact detail, and an agreed next step. Which facts matter is a decision each business makes once. What turns it into a score is counting how many of them you actually have.
How should lead scoring be calculated?
By counting filled slots and weighting them, in code, from the answers the customer gave. The output should be reconstructable: anyone should be able to look at a score and see which facts produced it. A score nobody can interrogate is a score nobody will act on.
Why do sales teams ignore lead scores?
Because most scores cannot be argued with. If a rep cannot see why a lead was called hot, their only options are to trust it blindly or ignore it, and after two bad calls they ignore it. A score derived from named facts can be disputed, which is exactly why it gets used.
Should an AI decide whether a lead is qualified?
It should collect the facts, not grade them. A model asked to assess the quality of a conversation it just had is being asked to mark its own work, and it will flatter itself. Extraction is a job for the model; scoring is a job for arithmetic.
What is the difference between a hot and a warm lead?
Only what you define it as, which is the point. A workable definition for property is that hot means intent, budget, area, timeline, financing and contact are all present, warm means intent, budget, area and contact, and anything less is cold. The specific rule matters less than having one that is written down and computed the same way every time.
Which qualification questions get skipped most?
The ones that feel intrusive: budget, how the purchase is being financed, and who else has to approve the decision. They are also the three that most often decide whether the sale can happen at all, which is why leaving them to the end of a human conversation is expensive.
The short version
Write down the six or eight facts that decide a sale in your business. Score leads by counting how many you have. That is the entire method, and it beats any lead score nobody trusts.
Ampa fills the slots, then hands you the brief.
It asks the full qualification set every time, in the same order, without sounding like a form, including the questions your team skips because they feel awkward.
Start free trial → or ask the agent something hard first