Lead prioritization: who gets the first call

An unsorted list is a random list

The lead list a salesperson opens in the morning is usually sorted one way: by date. Newest at the top, oldest at the bottom. That is not prioritisation — it is arrival order, and it has no relationship to the odds of a sale.

The consequence shows up daily. The rep starts at the top, runs out of energy by mid-afternoon, and the rest of the list rolls into tomorrow. That remaining part often holds the readiest customers: the people who called twice, asked about price, filled in the form.

Prioritisation fixes this and does not need a complicated model. What it needs is three dimensions written down, and a queue built on them.

The three dimensions of priority

Nearly every model that works in practice rests on the same three. The names change; the logic does not.

1. 1Recency — how fresh the enquiry isSomeone who filled in a form an hour ago is not in the same state as someone who filled it in three weeks ago. Recency is the strongest single dimension, because it tells you where the customer's attention currently is.
2. 2Intent signal — what they didVisiting the pricing page, downloading a catalogue, calling twice, naming a specific product — each is a stronger signal than general interest. The list of signals is particular to a company and has to be written once.
3. 3Fit — are they our customerGeography, sector, company size, product category. Fit does not grant priority on its own, but it is what stops an unsuitable lead rising to the top.

Using all three together matters. A queue driven by recency alone promotes poor leads; a queue driven by fit alone spends time on customers who have gone cold.

How to build a simple scoring model

The model should not be complicated. A five or six line table is the best form available, because it is the one the team will understand and therefore trust — and a model nobody trusts is a model nobody uses.

• Each signal gets points: form submission, price enquiry, repeat call, catalogue download, naming a specific product
• Recency gets a decay rule: the older the enquiry, the lower the score
• Fit becomes a filter: leads that do not fit never enter the queue
• The final score splits into three or four bands: now, today, this week, campaign
• Each band gets a response-time target

The exact point values do not matter in the first version. What matters is the separation of the bands: the «call now» group has to be small and unambiguous. If half your leads land in it, the model is not doing anything.

What does not belong in the model

Several things look like priority and are not, and they weaken the model.

• A rep's personal feeling — «this one looks good» cannot be measured or checked
• Brand recognition alone — a large name is not an intention to buy
• The general quality of a marketing channel — the signal comes from behaviour, not from the channel
• The lead saying they are «very interested» — only the next step tests that
• Lifting an old lead back to the top — without a reason, this breaks the queue

The last line deserves its own note. An old lead should rise only when it produces a new signal: visiting the site again, calling again, responding to a campaign. Raising it without a reason empties the queue of meaning.

Building the queue itself

Once the model is written, it has to live somewhere in the CRM. This part usually takes longer than the model and affects the result more.

1. 1Build the filtersOne CRM filter per band: source, date range, status and signal fields. A filter replaces a list assembled by hand.
2. 2Assign ownershipEvery band needs an owner. A band with no owner sits untouched, because everybody's job is nobody's job.
3. 3Write the stopping rulesAfter how many attempts a lead leaves the queue, which outcome code closes it, and which code moves it to the «later» list.
4. 4Choose what gets automatedThe bottom two bands — «this week» and «campaign» — are usually the first to be automated, because most of those calls are repeatable and nobody reaches them anyway.
5. 5Feed the outcome backThe result of a call should change the score: no answer lowers it, a real conversation changes the stage.

Priority is not the same in every segment

One scoring model does not fit an entire lead flow, and accepting that early saves a rebuild later. A lead from paid advertising and a lead from a referral do not produce the same signals: the first leaves a trail of behaviour, the second leaves almost none — and usually closes at a higher rate.

• Paid advertising: plenty of signal, recency is the dimension that matters most
• Referrals and word of mouth: little signal, fit and speed of contact carry it
• Existing customers: purchase history is the strongest signal, behaviour is secondary
• Cold lists: priority is built almost entirely on fit
• Event and trade-show leads: recency decays fast, the first 48 hours decide

The practical rule: do not write a separate model per source, but do write down which dimension carries the weight for each. One table, five rows — cheaper and more durable than five models, because a team can actually hold it in their heads.

Calling hours and the retry rule

Priority decides who gets called; the hour and the retry rule decide whether the call is answered. They are built separately.

• The best window per segment is found by observation over time, not by assumption
• The interval between attempts should not be constant: three calls at the same hour are worth less than one
• Attempts are capped in advance, and a lead that hits the cap moves to another channel
• Whatever hour the customer said suits them is recorded, and the next attempt follows it
• No automatic calls are scheduled at night

What to do with the ones that never answer is its own subject, and missed call lead recovery sets it out as a workflow.

How to check the model is working

The value of a priority model is not measured by feel. There is a simple, cheap way to test it: work a small part of the list the old way for a while and compare.

• A small share of leads is worked in the old order — by date
• The rest is worked with the new model
• Conversation rate, qualification rate and meeting rate are compared between them
• If there is a difference, keep the model; if there is none, rewrite the signal list

This comparison should not run for more than a week or two. The goal is not academic precision — it is seeing whether the model makes any difference at all.

Limits

Prioritisation does not create leads and does not fix a weak offer. The only thing it does is spend existing capacity better. If the team already reaches every lead in a day, the gain is small.

How Vexvon supports this

Vexvon's CRM and call side hold the fields a priority queue needs.

• Lead status, stage, category and priority are separate fields
• Contact attempts and reminders live on the lead record, so the attempt cap is not tracked by hand
• A customer who shares a number in a conversation is tagged automatically — one of the strongest intent signals there is, and it enters the queue directly
• A call campaign is built from a CRM filter, so a priority band becomes a target list as it stands
• Lead actions appear on one timeline, so why a score moved can be traced

How scoring is built on the chat channel is shown in lead scoring in a chatbot — the signals there are different because the channel is. The calling process as a whole is in sales call automation workflow.

First step

Starting takes a table and an hour: write the list of signals, give each one points, split four bands, and set a response-time target per band. Then work a week that way and read the outcome codes.

To discuss which part of the queue is worth automating, see the CRM page or get in touch.