How to score an inquiry from free-text answers

Scoring an inquiry from free text means turning what a visitor wrote in their own words into a few plain facts. Then each fact gets points from a rule you set. These five steps work for any business whose inquiries arrive as sentences instead of ticked boxes.

Who is this for?#

This is for a business that gets inquiries as typed words. They might come from a message box on a contact form, an email, or a conversational form. It fits a design studio reading project briefs, a contractor reading job descriptions, and a venue reading event plans.

It assumes you already know what makes an inquiry worth a call. If you do not, start with the guide to qualifying leads from your contact form. It covers that first decision.

It also assumes a score out of 100, built from a few weighted answers. That is the usual shape of lead scoring, and it is the shape the example on this page uses.

If every inquiry you get is a set of dropdowns and checkboxes, you do not need this page. A rule can already check those answers.

Why can't you score free text as it is?#

A rule can only check something exact. "Is the budget at least $10,000?" has a clear answer when the budget is a number. It has no answer when the budget is "depends on what you quote." Free text mixes the fact with maybes, side notes, and tone.

Three common shortcuts each go wrong in a different way:

1. Keyword lists miss the meaning. "No rush" and "rush job" both contain "rush."
2. Message length rewards writing, not fit. A long message tells you the visitor likes to explain. It does not tell you their budget.
3. Asking an AI for a number gives a score you cannot repeat. Ask twice and the number can change. You cannot compare it with last month. You cannot explain it to someone who asks why their inquiry ranked where it did.

The fix is to split the work in two. First, read the text for a few facts. Then score the facts with rules written ahead of time.

1. What fact should each answer become?#

For each question you score, decide the one kind of fact its answer turns into. Most answers become one of five kinds.

These rows are examples, not a fixed list. Write the kind beside each question you score.

A question that could become two facts is really two questions. "Tell us about your project" often holds a category, a size, and a deadline. Score each one on its own. That way one long answer does not count three times.

For category answers, keep a short list of allowed answers. Five to eight categories are easier to score than new labels made up for each inquiry.

2. How do you write a rule for each fact?#

Write each rule before you read any inquiries. That way the rule decides, not your mood that day.

A rule has three parts: the fact it reads, the check, and the points it gives. Give each question a weight, and make the weights add up to 100. The question that matters most should carry the most. For a design studio that is often budget. For a plumber it is often whether the job is inside the area they cover.

Then write point bands for each fact, not just pass or fail. A budget of $9,000 against a $10,000 minimum is close. A budget of $2,000 is not. The score should show the difference. Here is an example rule for a budget worth 40 points:

The last row matters. Step 4 explains why "not known yet" gets its own points.

3. How do you turn a sentence into a fact?#

Write down how to read the common ways people answer. Then read them the same way every time. Most of the trouble comes from five patterns:

1. A range ("8 to 10k"): use the low end. The score then never promises more than the visitor did.
2. A rough word ("a few thousand," "soon"): turn it into the band it most likely means. If two bands are just as likely, mark it unknown.
3. A date in words ("after the holidays," "in the spring"): turn it into a month, counted from the day the inquiry arrived.
4. A condition ("10k if you can do it all"): take the number. Write down the condition for whoever replies.
5. A refusal ("I'd rather talk about that on a call"): mark it unknown, not zero.

Having these rules matters more than the exact choices in them. Two people who follow the same written rules should get the same fact from the same sentence.

In Qualiform, the AI writes down what it learns from each reply as the conversation goes. Each scored question on the Questions tab then checks one saved answer in code. So the same saved answers always get the same score. The code reads the saved words itself. A budget written as "15k to 20k", "$15,000", or "about 20 grand" is read as a range, and your smallest project is checked against its low end, so a lead never needs a best-case number to clear it. What counts as a good answer is yours to set on each scored question, such as your smallest project. An answer with no number in it, like "not sure yet", stays unanswered rather than being read as zero.

4. What happens when an answer is vague or missing?#

Ask once more, in plain words, before you score a gap. "Soon" often becomes a month when someone asks, "Roughly which month?" One extra question is fair. Asking a third time starts to feel pushy, so stop after one.

If the answer is still unclear, mark it unknown. Unknown is its own band with its own points. It is not zero, because "I'm not sure of the budget yet" is not the same as "I have no budget." A middle number, like the 10 of 40 in the example rule, keeps a promising inquiry from sinking because of one gap. You can fill the gap later, on the call or in an email.

Answers also go missing when a visitor leaves partway through. In Qualiform the lead is saved on every turn, not only at the end. An inquiry that stops after two answers is still there to read. The AI can bring a question up again while it still has no usable answer, up to the number of times you allow for that question, from one to three. A scored question with no answer is left out of the score, not counted as zero. The score is worked out over the questions that were answered, so a missing answer never sinks an inquiry.

5. How do you keep each score explainable?#

Save three things for every scored answer:

1. The visitor's own words, so someone can check the reading.
2. The fact you read from them, so inquiries can be compared.
3. The points the rule gave, so you can see which answer moved the total.

Keep a one-line reason with the total too. For example: "Budget $15,000 or more, 40 points. Start within three to six months, 25 points."

When a visitor asks why they were offered a call next week instead of tomorrow, the answer is already written down. When a score looks wrong, the reason shows whether the reading or the rule is wrong. That tells you which one to fix.

In Qualiform, every lead carries a line saying how it got its score, so the reason sits next to the number.

What does a scored free-text inquiry look like?#

Here is one example inquiry to a design studio, scored with example rules. The studio, its rules, its weights, and the visitor are all made up for this page. The inquiry arrived in October.

Hi, I run a small bakery and my website is ancient. I'd like a new one with online ordering. I was thinking 8 to 10k. Ideally it's live before spring. My partner and I would decide together.

Budget held the score back. The low end of the visitor's range sits in the $5,000 to $9,999 band. Reading the high end would have put it at $10,000 and added 15 points. That is why the range rule in step 3 is written down. It keeps the reading the same, whoever does it.

Under the example score ranges in the guide to qualifying leads, 75 lands in the 60 to 79 range and is offered any open time.

What are the common mistakes?#

1. Scoring on keywords. A word can show up in an answer that means the opposite.
2. Scoring how much someone wrote. Length measures the writer, not the fit.
3. Treating unknown as zero. It sinks inquiries that were never asked a second time.
4. Asking an AI for the final number. You cannot repeat it or explain it.
5. Changing a rule without scoring again. When a rule changes, score last month's inquiries again, so old and new scores mean the same thing.
6. Scoring side details. The name of a visitor's current supplier helps the reply. It rarely belongs in the score.
7. Keeping the reading rules in one person's head. If the rules are not written down, a second reader cannot follow them.

When is this not worth doing?#

When you get a handful of inquiries a week, just read them. Written rules pay off when there are too many to read quickly, or when more than one person decides who gets a call.

It is not worth it when every inquiry gets the same reply anyway. A score that changes nothing is a number nobody reads.

It is also not worth it when a fixed choice would do. If budget is the one thing that decides, a dropdown with four ranges gives you the fact with no reading at all. A plain form with that dropdown can be the simpler fix. Free text is worth it when answers do not fit in a list, such as the kind of job, the shape of an event, or a deadline with conditions.

Last, do not score text you cannot write a rule for. "Seems serious" is a feeling. A feeling turned into points is still a feeling.

How do you know it is working?#

Run a two-reader test. Take twenty inquiries from last month. Give the written reading rules from step 3 to two people. Have each one turn the text into facts without talking to the other. Then compare the facts line by line. Where they disagree, a rule is missing or unclear. Add or fix it, and test again.

Then check the scores against what happened. Line up a month of scored inquiries with what became of each one: a call, a project, or nothing. If a project you won scored low, the reason line from step 5 shows which answer pulled it down. Fix that rule, not the total.

For the questions that feed these rules, see which questions a contact form should ask to qualify a lead.