Sales Cloud · lead management architecture
Seven layers,
and the lead is lost in one of them.
Between a form submission and a converted opportunity there are seven places where a lead stops moving. Each has its own failure, its own check, and its own owner. The reason most fixes disappoint is that the wrong layer was rebuilt.
Reviewed 27 August 2026.
Before the seven
Lead management, routing and scoring are three different questions.
Most teams use the words interchangeably, and that is the single most common reason an audit produces findings that change nothing. Scoring decides which lead deserves attention. Routing decides who receives it. Lead management is the whole path, including the handoff and the feedback loop that neither of the other two touches.
The practical consequence: a routing project does not fix lead management, and a scoring rebuild does not fix the handoff to sales. Both are worth doing. Neither is a substitute for knowing which layer is actually broken.
Layer 01
Capture without source attribution
Forms multiply across the main site, microsites, partner pages, event registrations and content gates. Without governance, tracking parameters drift, hidden fields get copied wrong, and a year later nobody can say which channel produced which lead.
How to check it
Count the live forms, then count how many write a source value you would trust in a board report. The gap between those two numbers is the size of this problem, and it is usually larger than anybody expects.
Layer 02
Creation without a qualification gate
By default every submission becomes a lead record. Nothing checks whether the submission represents somebody who could buy. Students, competitors, job seekers and bots accumulate alongside real buyers, and every downstream measure is computed over the mixture.
How to check it
Take last quarter and count how many leads were never worked at all. If that share is large, the model is not filtering, it is collecting.
Layer 03
Routing that drifted
Assignment rules, round robin queues and territory logic are configured once and revisited rarely. Over a year or two you get queues owned by deactivated users, round robin entries for people who left, and territory rules describing a sales organisation that no longer exists.
How to check it
List the queues and check the owner of each is still active. Then look for leads with no activity since creation and group them by owner. The pattern is usually one or two queues, not a spread.
Layer 04
Scoring inflation with no decay
Points accumulate for activity and nothing removes them. Somebody who read a whitepaper eighteen months ago outranks somebody who opened the pricing page this week. The score keeps rising and stops meaning anything.
How to check it
This layer has its own page, because it is the one most often rebuilt on its own: the scoring rebuild is here. What matters for lead management is that fixing it alone does not fix the layers on either side.
Layer 05
The handoff with no agreed response time
Marketing raises a qualified lead. Sales is supposed to pick it up. There is no written time to first touch, no escalation when nobody does, and no requirement to record why. This is the most expensive gap in the list and the one audits skip, because it is not a configuration, it is an agreement.
How to check it
Measure time from qualification to first activity, and look at the distribution rather than the average. A good average with a long tail means most leads are fine and a specific queue is not.
Layer 06
Rejection with no feedback loop
When sales rejects a lead, three things should happen: the status changes, the reason is recorded, and the lead returns to marketing for nurture or is disqualified properly. In most orgs only the first happens, so nobody ever learns what marketing should stop sending.
How to check it
Pull the rejected leads from last quarter and count how many carry a reason. If most do not, the loop is open and the same unqualified leads will keep arriving.
Layer 07
Conversion that fragments the account
Converting a lead creates a new account, contact and opportunity by default. If a matching account already exists, the rep is meant to match it. Many do not, and the customer ends up as several accounts with the history split between them.
How to check it
Search for accounts with similar names and check whether opportunities are spread across them. This one is cheap to find and expensive to leave, because every report that groups by account is quietly wrong.
Order of work
Fix the earliest broken layer, not the most annoying one.
The layers depend on each other, so sequence decides whether the work compounds or cancels. Rebuilding scoring while capture writes no reliable source means the model is trained on data nobody can attribute. Fixing routing while the handoff has no agreed response time means leads arrive faster at somebody who was not going to call them.
So the order is the order of the list: capture, creation, routing, scoring, handoff, feedback, conversion. Find the earliest one that is broken and start there, even when a later one is louder.
The cheapest diagnostic
Three numbers, and all three come from reports you already have.
Time from qualification to first activity. The share of qualified leads that become an opportunity. The share of leads carrying a source you would defend in a board report.
Compare each against your own last four quarters rather than an industry benchmark. Published medians are gathered from companies whose definition of a qualified lead is not yours, which makes them interesting and not actionable. Your own trend is the only comparison that is definitely valid, and a decline in any of the three points at a layer rather than at a person.
Asked often enough to answer here
Questions
What is the difference between lead management, routing and scoring?
They are three layers of one system and confusing them is why audits produce findings that change nothing. Scoring decides which lead deserves attention. Routing decides who receives it. Lead management is the whole path from capture to a converted opportunity, including the handoff and the feedback loop. A routing project does not fix lead management, and a scoring rebuild does not fix the handoff.
Which layer should we fix first?
The earliest broken one, because the layers depend on each other. Fixing scoring while capture has no reliable source data means the model is trained on data nobody can attribute. Fixing routing while the handoff has no agreed response time means leads arrive faster at somebody who was not going to call them anyway.
How do we know whether we have a problem without hiring anybody?
Three numbers, and all three come from your own reports. Time from qualification to first activity. The share of qualified leads that convert to an opportunity. The share of leads carrying a source you would defend in a board report. Compare each against your own last four quarters rather than an industry benchmark, because the trend in your own org is the only comparison that is definitely valid.
Is this a Salesforce problem or a marketing automation problem?
Usually both, at the seam. Capture and scoring live on the marketing side, routing and conversion live in Salesforce, and the handoff is the boundary. That is why fixing one side in isolation moves a local number and not the pipeline.
What does the work cost?
The paid check is 500 euro against your own exports, delivered within three working days with a costed order of work. The scoring layer alone is 2,250 euro for 45 hours. Anything wider is quoted in hours at fifty euro from what the check finds, and every price is published.
The check is €500 against your own exports, delivered within three working days and credited against the work. The scoring layer alone is €2,250 for 45 hours. Anything wider is quoted in hours at fifty euro, and every price is published here.
Serhii Skrypnyk · Senior Salesforce Administrator and developer · 7 Salesforce certifications · on the platform since 2018. Reviewed 27 August 2026.