Lead scoring gets sold as a black box — "our AI identifies your hottest leads" — which is convenient for vendors because it prevents you asking how. The useful version is not mysterious. It is a weighted sum of a handful of signals, and knowing which signals and how they are weighted tells you immediately whether a given implementation is real or decorative.
What actually predicts a good lead
In rough order of predictive weight, across most categories:
Stated budget
The heaviest single factor almost everywhere. Someone who tells you what they are willing to spend has already decided to buy something. Worth the most points and worth asking for on every form.
Timeline
"This week" and "sometime this year" are different businesses. Second heaviest, and the cheapest question to add.
Prior visits
Someone who has been to your site before is further along than someone meeting you cold in a feed.
Income band
A real signal in high-ticket categories, noise in low-ticket ones. Weight it by what you sell.
Age
Weak alone, useful combined — and heavily category-dependent. A cosmetic clinic and a coaching institute care about opposite ends of it.
Device
Genuinely predictive in some markets and pure noise in others. Also the input most likely to encode something you did not intend, so treat with care.
The currency bug that inverts the whole thing
This one is worth a section because it is invisible, expensive, and we shipped it ourselves.
If the budget factor compares a stated figure against a fixed threshold, that threshold has a currency. Set it in rupees and apply it to a lead who stated a budget in dollars, and the numbers differ by roughly 85 times. Every foreign lead comes in far below the threshold and scores zero on the heaviest factor in the model.
💡 The result is not slightly wrong — it is inverted. Your best foreign leads rank below your worst domestic ones, and the ranking looks plausible because the scores are all real numbers. If a tool scores leads, ask which currency the budget test uses and what happens when the lead's currency differs from it.
The correct implementation evaluates the budget in the ad account's own currency, or converts before comparing. Anything else silently penalises every market you are not based in.
Scoring matters less than the alert
The uncomfortable truth about this whole category. A perfectly scored lead sitting in an inbox nobody opens until evening has been scored for nothing.
Response time predicts conversion more reliably than almost any other variable. A lead contacted within minutes converts far better than one contacted an hour later, and one contacted the next day is usually gone — while you paid exactly the same for it.
- 1The alert must reach a phone, not an email address. Something that interrupts, carrying the name, number and what they asked about, so you can judge in ten seconds whether to stop what you are doing.
- 2It must fire the moment the lead lands, not on a sync schedule. A tool that pulls leads every fifteen minutes has built a fifteen-minute delay into your most time-sensitive asset.
- 3If it goes over WhatsApp, it must be sent on a utility-category template rather than a marketing one. Marketing templates are rate-capped per recipient and delivered differently, so a lead alert on the wrong category arrives late or not at all.
The design rule nobody asks about
💡 Notifying must never put the lead at risk. Save the lead first, then attempt every alert inside something that cannot fail loudly. The worst acceptable outcome is an unsent notification; a lost lead is never acceptable. Ask a vendor what happens to a lead if their messaging provider is down — the answer tells you whether they have thought about this properly or bolted it on.
What scoring is genuinely for
Not deciding which leads to ignore. Almost every business should contact every lead. Scoring decides the order and the effort.
Order
Call the high scores first. When six arrive on a Monday morning, that ordering is worth real money.
Effort
A high score justifies a call; a low one might justify a message. Same lead volume, better use of your day.
Feedback
The most underused benefit. If one audience segment produces consistently low-scoring leads, that is a targeting signal — stop buying that audience.
That last one is where scoring pays for itself over months rather than days. Cost per lead tells you what you paid; score distribution tells you what you bought.
How to evaluate an implementation
- 1Ask what the factors are. A vendor who cannot list them is selling a black box, and a black box you cannot inspect is a black box you cannot debug.
- 2Ask which currency the budget test uses.
- 3Ask how fast the alert fires and by what channel.
- 4Submit a test lead through a live ad and watch what arrives — the score, the fields, the timing. Five minutes, and it is the only way to actually know.
Scored in your own currency, on your phone in seconds
Adyft scores every lead 0–100 on budget, timeline, prior visits, income, age and device — evaluated in your own account currency — and alerts you immediately. 14-day free trial.
Start free — no card needed