If your sales team is tired of chasing “hot” leads that go nowhere, you’re not alone. The promise of lead scoring is simple: focus on the people most likely to buy. But most out-of-the-box models are generic, and the fancy AI stuff? Usually just a black box. This guide is for sales or ops folks who want a clearer, more practical way—using Rev—to build a lead scoring model that actually reflects what closes in your business.
Why bother with custom lead scoring?
Let’s be honest: most default scoring models in CRMs are, at best, a guess. They lump together website visits, job titles, and industry, then spit out a “score” that barely moves the needle. If you’ve ever followed up with a “95/100” lead who ghosts you, you know what I mean.
Custom lead scoring is about building a model based on your historical data and real signals—not someone else’s template. Done right, it means:
- Less time wasted on tire-kickers
- More focus on accounts that actually fit your ideal customer
- A scoring system your reps trust (not just roll their eyes at)
What is Rev, and why use it?
Rev is a lead intelligence tool that pitches itself on finding your best-fit prospects using AI and “lookalike” modeling. Unlike most, it doesn’t just match firmographics or scrape LinkedIn. The big sell: You feed it a list of your current best customers, and it finds more prospects that actually resemble them in subtle ways.
It’s not magic, but if you have enough closed-won data, it can help you find patterns you’d miss by hand—especially for complex B2B sales where the right signals aren’t obvious.
What Rev isn’t: A silver bullet. It’s only as good as the data you feed it, and if your sales process is a mess, no tool will fix that.
Step 1: Get your data together
You can’t build a decent lead scoring model if you don’t know what your best customers look like in the first place.
What you’ll need
- A clean list of your “closed-won” customers (the more, the better)
- Key data fields: company name, industry, size, revenue, location, and anything unique to your business (e.g., tech stack, use case)
- A list of “bad fits” (closed-lost or churned customers) if you want to sharpen the model
Pro tip: Quality is better than quantity. If your CRM is full of junky data, take the time to clean it up. Garbage in, garbage out.
Step 2: Define your “ideal customer” with actual evidence
Skip the marketing personas for now. Instead, dig into closed deals:
- Are your best customers a certain company size or industry?
- Do they use a particular piece of software?
- Is there a pattern in location, growth rate, or even tech stack?
If you can, talk to your sales team. They usually know which deals were a pain and which ones were a breeze.
Ignore: Vanity metrics (like website visits) unless you know they truly correlate with buying. Don’t overweight superficial stuff—focus on what actually predicts a good deal.
Step 3: Set up Rev’s lookalike model
Once you have your data, it’s time to get it into Rev. Here’s how to do it without getting lost in the UI.
- Upload your “closed-won” customer list. Rev will ask for a CSV or spreadsheet. Map the fields as closely as you can (company, domain, etc.).
- (Optional) Upload “bad fit” accounts. This helps the model know what not to look for.
- Let Rev analyze your data. It’ll go hunting for patterns, some obvious (like industry), some weird (like companies that use a specific technology).
- Review the “lookalike” profile. Don’t just accept the first draft. Check if the signals actually make sense. Is Rev highlighting things that track with your experience? If not, tweak or re-upload better data.
Honest take: Rev’s model is only as smart as your input. If you feed it random deals or a mix of product lines, you’ll get a mushy model. Be picky.
Step 4: Build your custom scoring criteria
Rev gives you a list of signals and a similarity score for new leads. Don’t just use these scores blindly—combine them with what you know from your own sales process.
How to set up scoring:
- Weight the signals: Maybe industry is a must-have, but tech stack is just “nice to have.” Assign points accordingly.
- Factor in your own data: Layer in engagement signals from your CRM (e.g., last activity date, inbound requests).
- Set thresholds: Decide what a “high score” actually means for your team. Is it top 10%? Top 20%?
- Test before rolling out: Score a batch of known leads. Does the model surface your real best customers? If not, adjust.
What to ignore: Don’t get obsessed with tiny differences (is a 79 really worse than an 82?). Group leads into buckets: high, medium, low. That’s usually more useful for reps.
Step 5: Push scores to your CRM and train your team
A model is worthless if your reps don’t see or trust the scores. Make sure the workflow is dead simple.
- Sync scores to your CRM: Rev can push scores to Salesforce, HubSpot, etc. Put the score where reps can’t miss it.
- Explain the “why”: Show your team what the score means and which signals matter. If they don’t buy in, adoption will be a slog.
- Use the scores to prioritize, not dictate: The model should help reps focus, not replace their judgment. Encourage feedback—if the score is off, you want to know.
Pro tip: Don’t hide the logic. If reps know what drives the score, they’ll trust it more (and spot when it’s off).
Step 6: Measure, tweak, and repeat
No model is perfect out of the gate. The real world is messy, and sales cycles change.
- Track conversion rates by score bucket: Are your “high score” leads actually closing more often? If not, dig in.
- Get rep feedback: If your team says the “good” leads are duds, listen.
- Update your model as you learn: Add new data, adjust weights, and rerun the analysis every quarter or so.
Warning: Don’t fall into the trap of overfitting—making the model so tailored to old deals that it misses new patterns. Keep it simple and pragmatic.
A few things to watch out for
- Don’t expect miracles: Lead scoring can help, but it won’t fix a broken sales process or a bad product.
- Be skeptical of “AI” magic: If Rev (or any tool) gives you a result that seems off, trust your gut and double-check.
- Avoid black-box models: If you can’t explain why a lead is scored a certain way, you’ll lose your team’s trust.
Keep it simple. Keep improving.
Custom lead scoring with Rev isn’t about building a perfect model—it’s about getting a little smarter every month. Start with the basics, get your team’s input, and don’t overcomplicate it. If it’s working, great. If not, tweak it. The best models are the ones your team actually uses.
If you’re still stuck, remember: a simple, transparent model almost always beats a fancy black box. Good luck—and don’t let the hype distract you from what actually drives sales.