Which unsold estimates should we chase first?

Size is the wrong sort order

Sorting the follow-up list by quote amount feels rational and performs badly, because the probability of closing varies far more than the amount does. A quote from two days ago on urgent work for an existing member is a different animal from a large quote from six weeks ago for optional work, and the second one dominates a size-sorted list.

The right ranking multiplies expected value by probability of recovery. You do not need a model to do this — a rules-based score built from four fields beats an unsorted list by a wide margin.

The four fields that carry most of the signal

• Age. The decay is steep and fastest in the first week. Any follow-up system that does not touch a quote within days is optimizing the wrong end of the curve.
• Contact since. A quote nobody has followed up is uninformative. A quote followed up three times with no response is telling you something.
• Urgency of the work. Failed equipment, safety findings and active leaks recover. Optional upgrades and efficiency work go cold and stay cold.
• Relationship. Existing customers, members and repeat buyers respond at meaningfully different rates than first-time callers.

Segment the dead from the deferred

The most common pipeline pathology is that nothing is ever marked lost, so the open-estimate total grows forever and stops meaning anything. Split the pipeline into active, deferred and dead, with a written rule for each transition — an age threshold plus a contact-attempt threshold works fine.

Deferred is not dead. Seasonal work quoted in the wrong month should come back onto the list at the right one. That requires a wake-up date on the record, which most systems will hold in a custom field, and a scheduled process that resurfaces it. Building that resurfacing loop is a small piece of workflow automation with an unusually direct payoff.

What the follow-up should say

Prioritization only helps if the contact is worth taking. Generic checking-in messages perform poorly. Referencing the specific finding, the specific equipment and what changes if it waits performs better, and that content already exists in the technician's notes and the original call.

Assembling it automatically — pulling the job details into a message a human reviews and sends — is a straightforward use of automation, and it is the difference between a follow-up queue that gets worked and one that gets ignored because the work is tedious.