A falling close rate isn’t a summons to a sales pep talk. It’s a signal that something in the path from inquiry to accepted work changed. Your team may be slower. Your offer may be worse. Or one part of town may now be a different market than the other.
Call the missing control a catchment cohort: leads grouped by where the buyer is, how the lead arrived, and when it entered the pipeline.
Without it you see one blended close rate, your sales manager hears “sell harder,” and you change the script for a problem that began across a bridge. Efficient, in the way a leaf blower is efficient at sorting paperwork. The point isn’t to deny a sales problem. It’s to locate the broken stage before you pay to amplify it.
Start with your last 90 days, split the records by customer-origin catchment and source, and follow each entry cohort from response to close.
The Denominator
“Close rate” isn’t a measure until you name its denominator. One person can make several inquiries, receive one quote, become one opportunity, accept one job, and make several payments. A report that blends those units will always find a number. It won’t always find an answer.
Build a short measurement contract for every segment:
- State the admitting event: inquiry received, qualified lead, site visit, or released quote.
- State the success event: two-way contact, appointment held, quote accepted, or closed-won with acceptance evidence.
- State the date basis, treatment of duplicates and reopenings, prevailing system record, and permitted recording delay.
Use entry cohorts for conversion. A lead received in the first part of the 90-day window stays in that cohort even if it closes later. Use closing periods for booked orders or collected cash. Those are different reports. A late-stage deal that closes this week may have entered before the local condition changed, and treating it as new demand is how your dashboard starts making up folklore.
Keep count and value side by side. A territory can close a high share of small repairs while losing the installations that make your month. Then segment by source, offering, new versus existing customer, and field versus office sale before you assign a cause. A citywide percentage is a summary. It isn’t a diagnosis.
Where the Buyer Is
Define the catchment around the actual purchase occasion. For emergency or routine work, response time, travel, and service availability can make a short physical distance expensive. For specialist or infrequent work, a buyer may tolerate far more travel.
Credit-card research cited in the corpus finds median transaction distances of about 4 km for food stores, 12 km for eating and drinking, more than 20 km for durable goods, and 33 km for amusement and recreation. Those figures aren’t service-radius settings. They show why one universal circle is fiction with a map pin.
Use the address where the work happens, not just the account’s mailing address. Add the route condition that changes serviceability: a bridge, ferry, parking restriction, service-day boundary, or access rule.
Add the customer origin and the competitor named at loss. A municipal boundary, postal code, franchise territory, and tax jurisdiction can all sit on the same map while answering four different questions.
Don’t manufacture a catchment from resident population. County Business Patterns gives structural detail down to 6-digit NAICS levels, but it arrives roughly 18 months after its reference year and excludes some industries. It’s a baseline, not a live competitor register.
OnTheMap can show worker inflow, outflow, and workplace density at census-block detail. That can explain a lunch route or a weekday demand generator, but it can’t define your service catchment without your own customer-origin records.
The practical comparison looks like this. The values are the fields to inspect, not universal performance targets.
| Catchment | Lead source | Travel boundary | Competitor signal | Response rate | Appointment rate | Close rate | Average ticket |
|---|---|---|---|---|---|---|---|
| Core service area | Tagged referrer, repeat customer, directory, or paid source | Actual route and access conditions | Buyer-confirmed alternative or no stated alternative | First substantive attempt and two-way contact kept separate | Appointment booked after two-way contact | Accepted evidence divided by the stated cohort | Accepted value divided by wins, shown beside count |
| Contested edge | Same controlled source codes | Bridge, congestion, delivery charge, or scheduling constraint | Named local, national, or remote substitute where supplied | Compare with the same source and entry cohort | Compare after contact, not after all inquiries | Compare only with an identical denominator and maturity | Compare offering and value band as well as average |
| Outer or uncertain area | Untagged sources quarantined until classified | Boundary still a hypothesis | Loss reason and supplier reach need evidence | Missing timestamps remain missing | Do not infer an appointment decline from no-response records | Do not call the segment weak until records mature | Do not use one large job to tell the story |
The Stages
Don’t jump from inquiry to close.
A workable pipeline normally has four to seven open stages, enough to separate qualification, discovery or site inspection, proposal, and decision without creating labels nobody can apply. Each needs an observable entry and exit condition. “Proposal sent” is evidence. “Seems interested” is an employee’s weather report.
Keep three clocks separate: receipt to ownership, ownership to first attempted contact, and receipt to the first two-way exchange. An automated acknowledgement stops none of them.
In the cited 2021 InsideSales vendor study of inbound leads, attempts in the first 5 minutes converted at over eight times the rate of attempts made between five minutes and twenty-four hours. That’s evidence that response is a control variable. It isn’t a licence to install a universal five-minute promise in a field business.
The corpus’s conditional starting bands are 5–15 minutes during staffed hours for explicit callback requests, 1–2 business hours for ordinary quotation inquiries, and next business day for low-intent leads.
Your bands have to match your actual coverage. If you’re tied up on a site visit for 90 minutes, you don’t possess a five-minute response standard — you need a route, a shared phone, an answering service, or an alternate owner. Otherwise the metric measures wishful staffing.
Preserve the activity timestamp and the CRM-entry timestamp separately. A field representative can have a real site conversation and enter it later.
Replace the missing activity time with the last-modified time and you’ve turned data-entry delay into customer-response delay. It makes a territory look slow for the convenience of the software. Very modern.
Read the Decline
Read the shape of the decline before you change commission or scripts.
First-attempt timing worsens in every catchment and source → intake or routing is broken
Look for inquiries with no owner, no first-attempt timestamp, duplicate contact, or reassignment delay. A lead black hole is a control failure before it’s a motivation problem.
Attempt rate is stable, but two-way contact falls in one catchment → access or customer availability is broken
Compare phone validity, inbound timing, after-hours handling, travel-related eligibility, and source quality. A completed task proves that someone clicked something. It doesn’t prove the buyer was reached.
Two-way contact is stable, but appointments fall → offer fit or local referral flow is broken after first contact
Compare the stated reason, property or job type, appointment availability, travel condition, and originating referrer. A property-manager stream can carry a different kind of work from an untagged directory stream inside the same postal code.
Appointments hold, but quote acceptance falls in one catchment → competitive conditions or proposal control is broken
Compare accepted scope, quote version, buyer-confirmed competitor, delivery terms, and the local alternative’s ability to serve that address. A remote supplier counts only where it can meet your customer on acceptable terms.
One territory loses leads while another still converts → geography or source data is broken, or a local market changed
Check whether service location, source, and travel boundary were retained. Without them you’ve blended catchments with different competitive conditions into a single failure rate.
Loss reasons suddenly say “competitor,” but the notes are empty → evidence quality is broken
Record a controlled primary reason, a secondary reason where needed, who supplied it, and a short verbatim note. Keep buyer confirmation, seller inference, and management recoding separate. Three losses are not a pricing study.
The Local Signals
Competition follows substitution, not city limits.
U.S. merger guidance treats distance, transport cost, regulation, custom, reputation, and local service availability as possible restrictions on geographic substitution. That’s useful operator logic. A rival is a real alternative only when your buyer can use it on acceptable terms at that address.
Referral flow is a source system, not a compliment counter. A referral can narrow search, carry a judgment about fit, and put the referrer’s reputation behind the introduction.
A cited randomized study of supplier and client firms found that subsidized introductions led to transactions and partly displaced existing partners, while information-only introductions did not.
Track the estate agent, insurer, building manager, trade counter, former customer, or platform that originated the lead. Then compare accepted work and gross margin by referrer and catchment.
That also stops you drawing a bad conclusion from good top-line demand. Dense markets bring more customers and more substitutes at the same time, and a new local competitor can change the competitive field without your salespeople suddenly forgetting the job.
Google’s local-results guidance names relevance, distance, and prominence as the main factors. The operating takeaway is less glamorous: keep locations, categories, hours, service areas, links, reviews, and location-specific proof accurate.
A single national landing page doesn’t contain a blocked bridge, a building-access rule, or a trusted local referrer. Sadly, the website doesn’t get a parking pass.
The Review
Run the comparison weekly while the change is live, but don’t overwrite its evidence. Preserve the cohort definition, source rules, catchment map, stage history, and report date.
Then check your CRM data quality as five separate tests: completeness, timeliness, validity, consistency, and accuracy, following the UK Government Data Quality Framework distinction the corpus uses. A full field is not necessarily a true field.
Before you act, ask three questions:
- Is the loss concentrated at response, appointment, proposal, or close?
- Is it concentrated in a catchment, source, offering, value band, or seller?
- Is the reason buyer-confirmed, and does the underlying record support the CRM label?
A blended close rate is a reporting result. A segmented stage break is an operating fact. A sales script can change behaviour. A catchment record can reveal the market.
Split your last 90 days by catchment and source. Find the first stage that moved. Change commission or scripts only after the record earns the conclusion.