Build it with AI
Create a lead tracker that follows prospects from first touch to next action.
Build a lightweight lead system for source, status, owner, activity, follow-up, and conversion.
Introduction
What a lead tracker has to hold.
Most teams end up with a lead tracker the same way: an inbox, a spreadsheet, and somebody's memory of who they promised to call back. One person handling twenty leads a week can hold the state of all of them. At sixty a week across three people, the state exists only as the union of three inboxes, and nobody can name which leads have gone quiet.
Leads arrive faster than anyone updates the sheet, so the ones nobody claimed are indistinguishable from the ones already handled. There is no record of when a lead was last touched, no single owner that survives a handoff, and no rule that makes an untouched lead visible before it is cold rather than after.
What follows covers building a lead tracker: the records it holds (source, status, owner, first-touch time, activity history, and the next follow-up), the systems it reads (Gmail and HubSpot), and what it does not fix.
The problem
Leads arrive faster than anyone can decide what to do with them.
A spreadsheet models leads as rows and has no opinion about time, which is the only dimension that matters — a lead is not a record, it is a clock. Marketing automation tools have the opposite problem: they model the clock beautifully and hide the individual lead inside a campaign.
The records are source, status, owner, first-touch time, activity history, and the next follow-up, and the authoritative copy of most of them already lives in Gmail or HubSpot. The version in the sheet says contacted, the version in the inbox says the reply never went out, and the lead has been sitting between the two for eleven days.
The cost is not the inconvenience: a lead that asked a specific question gets a generic reply four days later, or none.
You're likely here because
- Nobody can say how many leads from last month were never contacted at all
- Leads arrive faster than anyone updates the sheet, so the ones nobody claimed are indistinguishable from the ones already handled.
- When it is wrong, a lead that asked a specific question gets a generic reply four days later, or none
What gets built
Launch builds it, Grow operates it.
Built in Launch
- • Lead database
- • Status views
- • Owner queues
Operated through Grow
- • Lead enrichment
- • Follow-up
- • Scheduling
Systems it reads
- • Gmail
- • HubSpot
- • Google Calendar
The record model
What a lead record has to hold.
- Source and campaign
- Captured at creation, never inferred later. Attribution reconstructed from timestamps is a guess dressed as a measurement, and it always favours the channel that is easiest to see.
- Created timestamp
- The clock starts here, not on the next working day. Weekend arrivals are the classic silent loss, and a queue that ages from creation makes the Monday backlog visible as a backlog.
- Owner
- Assigned at creation rather than claimed later. An unowned lead should be a system error, because in practice it is a lead two people contact or none do.
- Status and its last change
- Both, because a lead that has been Contacted for eleven days is a different problem from one contacted yesterday, and the status alone cannot tell them apart.
- Last outbound contact
- Observed from the mailbox rather than marked by hand. The most common status lie is a lead marked contacted whose reply never actually went out.
- Qualification answers
- The specific answers rather than a computed score. A score cannot be argued with; the answers can, which is what lets the definition improve.
- Consent basis
- How this person came to be contactable, recorded at capture. The question arrives later and there is no way to reconstruct the answer.
How it runs
From an inbox to a queue with owners.
Step 01
Describe what a lead tracker has to do
Define what qualified means for this business, in a form somebody can apply without asking. Most lead tracking failures are definition failures wearing a tooling costume.
Step 02
Connect the systems of record
Email supplies whether a reply actually went out, the calendar supplies whether a meeting happened, and the CRM supplies whether the lead already exists as a contact. Reading them removes the three most common status lies.
Step 03
Build the operating surface
A lead record with source, owner, status, last-touch timestamp, and next action — plus the queue views that make an untouched lead impossible to miss.
Step 04
Start narrow
A single queue of leads with no outbound contact in the last three days, sorted oldest first, with an owner against each. Nothing else in version one.
Step 05
Route the exceptions
A lead that passes its response threshold with no activity reassigns or escalates automatically, because the failure mode is not bad follow-up but absent follow-up.
Step 06
Measure median time from lead arrival to first genuine contact attempt
Time from lead created to first genuine human contact, measured at the median and at the ninetieth percentile. The tail is where the lost revenue is, and the average hides it.
Implementation path
Standing up lead tracking in a fortnight.
- 01
Write down the qualification definition and test it on twenty real leads from last month with two different people applying it. If they disagree on more than three, the definition is the problem, not the tracking.
- 02
Baseline first-response time at the median and the ninetieth percentile, plus the share of leads from last month with no recorded contact at all. The second number is usually the shock.
- 03
Decide whether the CRM or the tracker owns a lead once it converts, and make the transition explicit. Leads living in both is how the same person gets contacted twice by two people.
- 04
Run it on one source first — the highest-volume channel — before pulling in the rest. A tracker that covers one source well beats one that covers all sources partially.
- 05
Build the narrowest useful version first: a queue of leads with no owner or no contact attempt inside the response-time target.
- 06
Writing the qualification definition and testing it on twenty real leads with two people applying it takes a morning and prevents most of the failures. Connecting the mailbox so first contact is observed rather than logged is the technical core. One queue, one source, live within a fortnight; extend to other sources only once the response-time baseline has moved.
- 07
Once first-response time has moved, add source-level conversion so the channel conversation is settled with data rather than by whoever remembers last quarter more confidently. Enrichment comes after that, and only as hints a person confirms rather than facts the workflow acts on.
Controls
Controls that matter.
Control 01
Ownership assigned at creation rather than claimed later, so an unowned lead is a system error rather than a normal state
Control 02
Outbound messages staged for human review before they send, because a mistimed automated follow-up costs more than a slow one
Control 03
Source and consent basis recorded on every lead, so a later question about how someone entered the list has an answer
Examples
Three ways a lead stops going quiet.
The lead that came in on a Friday
Weekend arrivals are the classic loss. A queue that ages from the creation timestamp rather than from the next working day makes the Monday backlog visible as a backlog rather than as a normal Monday.
Two reps contacting the same person
Ownership set at creation and enforced on the record removes the duplicate outreach that costs a deal before the first call. It is a permissions problem rather than a coordination problem.
The monthly source review
Source on the lead record and conversion attached to it means the channel argument is settled with the tracker open, rather than by whoever remembers last quarter more confidently.
How it goes wrong
Three ways lead tracking degrades.
The tracker is fed by web forms only, and the phone enquiries that convert best are invisible.
Build a manual creation path that takes under thirty seconds and works from a phone. A tracker covering one channel produces confident management of that channel and silence about the rest, which is worse than knowing you cannot see them.
Automated follow-up is switched on, and warm leads receive a sequence that reads as a sequence.
Draft automatically, send after a human has looked. The cost of a two-hour delay is far below the cost of a message that tells a warm prospect they are on a list, and the two get confused because only one of them is measurable.
Statuses multiply until there are eleven, and half the leads sit in one called Nurturing.
Every status needs an exit condition and an owner. Nurturing without either is the field where leads go to be forgotten with a timestamp, and it will absorb the majority of the pipeline within two quarters.
Limitations and considerations
What tracking a lead does not do.
- Faster response does not fix a lead source that produces unqualified leads. A tracker will tell you which source that is, which is often an uncomfortable answer about a channel someone owns.
- Consent and marketing-contact rules vary by jurisdiction and by how the lead arrived. Recording the basis is straightforward; deciding what it permits is a question for whoever owns compliance.
- Enrichment fills gaps and also introduces confident errors. Treat an enriched field as a hint that a person confirms, not as a fact the workflow acts on.
- Below about twenty leads a week with one owner, a shared inbox with a discipline is genuinely adequate and a tracker is overhead. If leads are plentiful and conversion is the complaint, the constraint is qualification or the offer rather than tracking, and a tracker will document that faster than it will fix it.
- Connector coverage varies: Gmail, HubSpot, Google Calendar are representative rather than guaranteed, and the fields exposed depend on your workspace permissions.
FAQ
Build a lead tracker with AI: common questions.
Is this just a CRM with fewer features?
It is the part of a CRM that runs before an opportunity exists, which most CRMs model badly because their object graph starts at the account. A tracker cares about elapsed time and ownership on records that may never become accounts at all.
When should a lead move into the CRM?
At the point your qualification definition is satisfied — which is exactly why the definition has to be written down first. Leaving the boundary implicit produces the two-systems-one-person problem within a month.
Can it chase leads automatically?
It can draft and stage the follow-up, and it should not send it unreviewed. The cost of an automated message that reads as automated to a warm lead is higher than the cost of a two-hour delay, and the two are frequently confused.
What about leads that arrive by phone?
They need a manual creation path that takes under thirty seconds, or they will not be recorded. A tracker that is only fed by forms produces a confident picture of the channel it can see and silence about the one that converts best.
What should the first version contain?
A queue of leads with no owner or no contact attempt inside the response-time target. Everything else waits until that one is genuinely used.
How will we know whether it worked?
Measure median time from lead arrival to first genuine contact attempt against the baseline taken before anything changed.
Start with ARIA
Ask ARIA to build it.
Describe the website, application, workflow, or operating surface you need. ARIA plans, connects, builds, tests, and keeps refining it — inside the permissions you set.
- ARIA acts only through the systems and permissions you connect.
- Connections use scoped credentials you can change or revoke.
- Actions are recorded, and consequential ones can require approval.
Start here
Build a lead tracker around the process you actually run.
Start with one queue — leads untouched for three days — and let the response-time baseline tell you what to build next.