Lead Scoring for Early-Stage SaaS: Fewer Fields, Clearer Rules

Early teams borrow scoring models from larger companies and then wonder why nobody uses them. Ten fields, a points engine, a "hot / warm / cold" badge that everyone ignores — that is not a system. It is a decoration. The version I would start with is closer to a sentence Adugbo Victory has used about the real work: who deserves a message today, and why. I would answer that with fit plus one signal, written as rules, not as a mysterious score.

Pramod Kumar, co-founder of Kingdom of Kumar. On X

·

Scoring does show up often in GTM engineer job posts. In Hlib S.'s November 2026 breakdown of 161 postings, enrichment and scoring were part of the common work (enrichment itself was in 71% of that sample). That is a labour-market observation, not a reason for a seed team to install a complex model. Hire or build for judgement first.

ICP rules are in How to define an ICP your team will actually use. Signals are in Types of buying signals. Once inbound is in the picture, see Inbound lead routing for a small team. The wider work is on GTM engineer and GTM engineering. If the product is not ready to sell, take the App Launch Score before you score anyone.

"Who deserves a message today, and why"

That sentence is the only score I care about at seed.

If a teammate cannot explain why this row is at the top of the queue, the score is a fog. I want a reason they can say out loud: "They fit the rule, and they asked for a demo this morning." Or: "They fit the rule, and a champion just moved there, inside our window." Or: "They fit the rule, and they are next on the hand-built list we committed to this week." Those are scores with nouns.

A number from 0 to 100, with no sentence attached, will not survive a busy Tuesday.

"Fit plus one signal"

Write two layers, and stop there for a while.

Fit (yes / no / never). Yes means they match the ICP page. No means they wait or they are researched further. Never means they are on the never-contact list — customers, opt-outs, wrong market, and so on. Do not let a "hot" signal override a never.

One signal (or none). A single, dated reason they should be in *today's* queue: inbound request, job change in window, hiring in the function you sell to, a closed-lost reopening you can explain, a referral. If there is no signal, they can still be contacted as part of a planned, capped outbound pass — but they should not jump the queue ahead of people with a real "why today."

If you have two signals, pick the one you will mention, or the one that is more time-sensitive. Do not add the points together to invent urgency.

I would store these as fields: `fit`, `signal`, `signal_date`, `why_today` (a short sentence). The "score" can be a sort order you derive from those fields. You do not need a product called a lead score to do this.

A simple sort order you can actually use

  1. 1Inbound requests and demo forms, newest first, if they fit or have not yet been checked
  2. 2People already in a conversation (these are not "leads to score"; they are replies to handle)
  3. 3Fit + a signal still inside its window
  4. 4Fit, no signal, on this week's planned list
  5. 5Everything else waits

If your CRM cannot sort that way, do it in a sheet for a month. A sheet you use beats a scoring app you do not.

Inbound and outbound can share the language, not the same "hot" label

A demo request and a job-change signal are both reasons someone might deserve a message today. They are not the same urgency, and they should not get the same badge if that badge will make a founder treat them as interchangeable. Inbound is usually a conversation the other person already started. Outbound is a conversation you are asking to start. If your score treats them as identical "80s," you will be late on the form-fill while you polish a cold first line.

Use the sort order above, and if you need a label at all, use words from the reason: "demo request," "champion moved," "planned list." Labels that mean something beat colours that mean "probably."

If you later add a product-led trial or a usage threshold, that becomes another named reason, with a window, not a hidden weight inside a black box.

When a points model starts to be worth it

Later, maybe, when you have enough volume that a human cannot see the queue, and you have history you trust — not survey data from another company, *your* history — you can test whether a weighted model beats the simple sort. Even then, keep the sentence. If the model cannot explain itself, it should not go live on sending.

I would also be wary of buying "intent scores" and dropping them into the same field as fit. Intent without fit is a notification. Intent on a never-contact is a hazard.

Mistakes that make scoring worse than a list

  • Letting a high score override "already a customer" or "asked to be left alone"
  • Scoring on stale data (title and email from last year)
  • Ten behavioural events that all mean "visited a page"
  • Recalculating scores so often that the queue reshuffles while someone is writing
  • Using the score as the first line of the email

The last one is how you get "I noticed you are high intent," which is not a fact the person can recognise.

A one-page scoring policy

  • Fit rules, with a link to the ICP page
  • Allowed signals, with windows
  • The sort order
  • Who may change a `why_today` sentence, and when
  • The rule that never-contact always wins
  • A Friday check: did we message people in the order we claimed, and did any "hot" row turn out to be wrong?

If the page needs a legend to be understood, it is too complex for the team you have.

If two cofounders argue about the queue

Do not settle it by raising a weight. Sit with the five rows at the top and the five that waited, and say the `why_today` sentence for each. If you cannot agree, the ICP or the signal window is still unclear. Tighten the written rule, pick it for two weeks, and look again on Friday. A points debate is often a fit debate in disguise.

FAQ

Should marketing and sales share one score?

At seed, you barely have two teams. Share one policy, one queue, and one language for "why today." A second model is how arguments start.

Do GTM engineers build scoring as their first project?

They can, but I would rather they make fit and data trustworthy first. A score on a dirty CRM is a confident mess. Hlib's sample shows the market asks for this work often; it does not say you should start there in week one.

Can you implement this for us?

Kingdom of Kumar builds and runs a scoped system on Grok and Claude for $500 a month, three-month minimum. You own it, and you approve what goes out. We will help you put `why_today` into the motion, not sell you a 100-point engine. That is not a RevOps platform, and it is not a $135,000 hire.

Contact

hello@work.kingdomofkumar.com