Lead Scoring for Funding Shops Without a Data Team
You do not need a model or an analyst. You need a short list of signals your best brokers already use by instinct, written down and applied the same way every time.
Ask your top closer how they decide which callback to make first and you will get an answer within seconds. Time in business, monthly deposits, whether the merchant answered the phone last time, whether they already sent statements. They are scoring leads. They just keep the model in their head, which means nobody else on the floor benefits from it.
Lead scoring for a funding shop is simply that instinct turned into a few rules everyone shares. Done well, it sorts the queue so newer brokers call the most promising merchants first and do not burn their best hours on files that were never going to go anywhere.
This guide walks through building a point model on one page, testing it against deals you already closed, and keeping it honest as things change.
Pre-flight summary
- Capture the signals your best closers already use, then share them as written rules.
- Keep fit and engagement scores separate so you can act on each properly.
- Test the model against past funded and dead deals before trusting it.
- Re-score automatically when calls connect or statements arrive.
- A low score means a later call, not a dismissed merchant.
Leg 01Start by interviewing your closers
Sit with two or three of your strongest brokers and ask them to sort twenty recent leads into call-now, call-later and probably-not piles. Then ask why for each one. Write down every reason they give, word for word.
You will notice the same handful of signals coming up. Those are your starting point. You are not inventing a theory of what makes a good lead; you are capturing what already works on your floor and making it repeatable.
Also ask what makes them give up on a lead. The negative signals are often more useful than the positive ones: a merchant who will not say what the money is for, a business type your panel rarely accepts, a phone number that has gone to voicemail six times. Those belong in the model too, as point deductions, so the queue quietly pushes them down instead of each broker rediscovering them.
Leg 02Two kinds of signals: fit and engagement
Fit signals describe the business and whether it is likely to match your funders' guidelines. Engagement signals describe how the merchant is behaving toward you right now. Keep them separate, because a perfect-fit merchant who has gone silent and a so-so merchant who just uploaded statements need different handling.
- Fit: time in business, stated monthly revenue, industry (and whether your panel restricts it), state, requested amount relative to revenue.
- Fit: existing positions disclosed, which affects which funders could consider the file.
- Engagement: answered or returned a call, replied to a text, opened the application.
- Engagement: uploaded bank statements or a signed application through the portal.
- Engagement: requested a specific callback time, which is a strong sign of intent.
Leg 03A point model that fits on one page
Keep it small. Five to eight criteria with whole-number points is plenty. Here is an illustrative version; your values should come from your closers and your own results, not from this example.
- Time in business over two years: +20. One to two years: +10. Under one year: 0.
- Stated monthly revenue within your panel's typical range: +20.
- Industry your panel commonly accepts: +10. Restricted industry: -20.
- Statements uploaded: +25. Application signed: +15.
- Conversation in the last 48 hours: +15. No response to five attempts: -15.
- Sort the queue by total, highest first, and review the cutoffs monthly.
A scoring model nobody can explain in one breath is a model nobody will trust on a busy morning.
Leg 04Test it against deals you already closed
Before rolling it out, score the last few months of leads retroactively. Put the funded files in one column and the dead ones in another. If funded deals mostly land in your top band and dead ones in the bottom, the model is directionally right. If funded deals are scattered evenly, one of your criteria is pulling the wrong way, and you can usually spot which by looking at the outliers.
Say a high-scoring lead died because the merchant was in a restricted industry your form did not catch. That points to a missing question at intake, not a broken model. Fixing the intake form often helps more than fiddling with point values.
Leg 05Re-score as the deal moves
A score set at intake goes stale fast. Engagement signals should update the moment something happens: a call connects, statements arrive, the merchant stops answering. Fit can change too, once statements show that true revenue is lower than the merchant estimated or that there are more existing positions than they mentioned.
This is where having the score live in the CRM helps. In FundPilot, statement underwriting pulls deposits, balances, NSFs and existing positions from the uploaded statements, each figure linked to its source page, so a broker can update fit criteria from real numbers rather than the merchant's guess. The AI assistant can also answer questions like which of my leads uploaded statements this week straight from live CRM data.
Leg 06Guardrails so the score helps people instead of replacing judgment
A low score means call later, not never. Merchants with short histories or uneven revenue still deserve a straight answer about whether you can help, and sometimes the honest answer is that they should wait or look at other options. Do not let the score turn into a reason to ignore people or to push higher-scoring merchants toward more funding than they need.
Review the model monthly with the same closers who helped build it. Retire criteria that stopped predicting anything, and add new ones when your panel's guidelines change. FundPilot's lender matching scores files against funder guidelines too, so when a funder tightens a rule, it is worth checking whether your lead score should follow.
Questions from the floor
How many criteria should a funding lead score use?
Usually five to eight. Fewer than that and you lose useful distinctions; more and nobody can explain why a lead scored the way it did, so brokers stop trusting it. Start small with the signals your best closers mention most often, test the model against past deals, and only add criteria that clearly improve how well it separates funded files from dead ones.
Should we use AI to score leads instead of points?
A simple point model is often the better starting point because everyone can see how it works and challenge it. AI can help by answering questions about your pipeline and summarizing calls, but the scoring rules themselves should stay understandable to your managers. You can always add sophistication later once the basic model has proven useful.
How often should we change the point values?
Review monthly and change only what the results justify. If funded deals keep appearing in your middle band, look at which criteria those files share and adjust. Avoid weekly tinkering; funding cycles take time, and changing values constantly makes it impossible to tell whether any change actually helped.
What if a merchant's statements contradict their application?
Update the fit score using the statements, which are the more reliable source, and have a straightforward conversation with the merchant about the difference. Often there is an innocent explanation, such as a transfer between their own accounts counted as revenue. Never adjust figures to make a file look stronger than the statements support.
General information only — not legal, tax or financial advice. Rules for telemarketing, texting, recording, disclosures and data retention vary and change; confirm current requirements with qualified counsel.
Connecting flights