Ask a pest control owner where last month's new customers came from and you get a confident answer. LSA is carrying us. Paid search is hit or miss. Lots of referrals.
Then ask a narrower question: how many of those customers can you tie to a specific call or form, with a record on both ends? Not what the CSR remembers. A phone number on a tracked call that matches the phone number on a customer account.
That count is almost always lower than the owner expects. In the example below it is barely half. And the half you cannot prove is where most marketing budget decisions quietly get made.
Every new customer belongs in exactly one of three piles
Take one closed month and sort every new recurring customer into one of these:
Measured. A record links the customer to a source. The phone number on their account matches a call to a tracking number, or their phone or email matches a web form that carried its source with it. Exact match, both ends on file, no judgment involved.
Estimated. The source is inferred. The CSR asked "how did you hear about us" and picked an answer from a dropdown, or someone made a reasonable guess. Useful, and a different kind of fact.
Unmatched. No record and no answer. The customer exists in your field software and nothing says how they got there.
The three piles have to add up to your new-customer count, every time. A source report that does not sum to the customers in your CRM is describing whatever the software happened to see, and you should read it that way.
A worked month
An illustrative shop signs 120 new recurring customers in May. It runs a separate tracking number on each paid surface (Local Services Ads, paid search, its Google Business Profile) and swaps numbers dynamically on the website. The main office number, the one on the trucks, the invoices, the yard signs and every directory listing since 2014, is untracked.
Match the month's customers to calls and forms by phone and email, and 63 come back measured:
- Local Services Ads: 24
- Google Business Profile tracking line: 18
- Paid search, calls and forms: 14
- Organic website visits, calls and forms: 7
Another 38 called the main line or were signed in person by a tech, with a source typed in by the CSR: "Google" 21, neighbor or referral 10, saw the truck 5, mailer 2. Those are estimated.
That leaves 19 unmatched. 63 plus 38 plus 19 is 120. The match rate, the share of new customers with a measured source, is 52.5%.
This shop believes it knows where its customers come from. It knows for about half of them.
"Google" is a search box, not a channel
The biggest single answer in the estimated pile is the least useful one. When a customer says they found you on Google, they mean the search box. That covers an LSA listing, a paid ad, the map pack, an organic result, and the homeowner who saw your truck, forgot the name, and searched for the pest company with the green trucks.
The dropdown records all five as one. Any report that hands those 21 customers to paid search, or splits them across your Google spend by formula, is guessing in a column labeled data.
Keep the source field anyway. It is the best evidence you will ever have on referrals and trucks, which no tracking line captures. Keep it in its own column, labeled estimated, and never add it to the measured count.
The main office line is the biggest hole, and it moves the credit
Most shops put tracking numbers on their ads and leave the main number alone, for good reason: it is printed on every truck, invoice and fridge magnet from the last decade. So the main line becomes the untracked pile by construction.
Here is what follows from that. If every paid surface shows its own tracking number, a call that rings the main line did not arrive through a paid click. The ad may have planted your name a week earlier. It did not deliver the call. That caller dialed a number they got from a truck, a neighbor, a directory, an old invoice, or a search for your company by name.
Which means the default way reports fill the gap, spreading unproven customers across channels in proportion to the tracked ones, overcredits paid every time. Run it on the worked month.
Say paid search spent $6,000 in May. On measured customers alone that is 14 customers, or $429 per customer.
Spread all 57 unproven customers by measured share and paid search picks up another 12.7, for 26.7 customers and $225 per customer. Nearly half the cost, on the same spend in the same month.
Now apply what you know about the main line. The 38 estimated customers came in on a number no ad displays, or were signed at the door, so none of them belong to paid search. Spread only the 19 truly unmatched, and paid search picks up 4.2, for 18.2 customers and $329 per customer.
Those three numbers are the difference between scaling a channel and cutting it. At a $115 quarterly ticket and a 55% gross margin, the top of the healthy range for a pest operation per NPMA industry data and PCO Bookkeepers guidance, a recurring account produces $253 of gross profit a year, or $63.25 a visit. At $225, paid search pays back on the fourth visit, inside a year. At $329 it takes six visits, about fifteen months. At $429 it takes seven, about a year and a half. A blended report shows you the $225 and invites you to scale it.
The working rule: decide on the measured number when it is good enough, and if a channel only works on the estimated number, it does not work yet. The estimate is allowed to tell you where to look. It is not allowed to sign the budget. More on why the booked customer is the right yardstick in grading ad spend on booked customers, and on what a given cost per customer can afford in the acquisition cost post.
One fix worth making this month: if your Google Business Profile lists the main line, give it a tracking number as the primary phone and keep the main number as the additional one. Otherwise your map-pack calls blend into the untracked pile, and a surface that can carry a large share of your phone volume ends up with almost none of the provable credit.
Match rate is the number to raise
Arguing about how to split the unproven pile is a loop with no exit. Shrinking the pile is not. The match rate is the one number in this subject you can move directly, and every point you add narrows the range you are deciding inside.
Lift the worked shop from 52.5% to 80%, which is 96 measured customers out of 120, and suppose paid search goes from 14 measured customers to 18. Its cost per customer now reads somewhere between $267 and $333, instead of $225 to $429. That is a range a budget decision can live in.
Most of the lift is clerical:
- Get a phone number on every customer record, in one format. The phone number is the join key. A CRM that stores "(919) 555-0142" on one account and "9195550142 cell" on another will not match against anything.
- Record the number they called from. The CSR who books the job under the spouse's cell or the landline breaks the match on a customer who was tracked perfectly.
- One tracking number per surface, and swap numbers on the website. A single "marketing number" shared by every ad tells you a call was marketing and nothing about which dollars.
- Make phone a required field on web forms. A form with only an email often matches nothing, because plenty of customers never give the office an email at all.
- Count customers, not calls. A homeowner who called three times before booking is one customer. A report that counts calls shows three conversions from one sale.
Some of the pile never matches. The walk-in, the neighbor a tech signs on the spot, the customer who called from a work phone. That is fine. The goal is to match as many real sales to real sources as the records allow, and to label the rest for what it is.
There is no benchmark match rate worth having
I won't give you an industry match rate, and the figures that circulate do not survive a close look.
Your rate depends almost entirely on your own setup: how many numbers you publish, how much of your book arrives by referral and trucks (which no tracking line catches), and how clean the phone field in your CRM is. Two well-run shops can sit at 45% and 80% for reasons that have nothing to do with marketing skill. A referral-heavy shop in a dense suburb should match less than a young shop buying most of its growth, and that is a sign of health, not a tracking failure.
The statistics that do get quoted about untracked calls mostly come from call-tracking vendors measuring their own customers: a population that already installed tracking and leans toward the shops that care about it. They describe those accounts, not yours.
The rate worth watching is your own, each month against the same month last year. A match rate that falls while spend rises means new money is going somewhere you cannot see.
How to run it on one closed month
- Export new recurring customers for one finished month from your field software, with phone, email and start date.
- Export every call and form from your tracking platform for that month and the 30 days before it, with caller number and source. People often call weeks before they book.
- Match on phone and email. Strip phone numbers to ten digits and lowercase the emails. Exact matches only. Each match is measured.
- Pull the source field for everyone who did not match. Those are estimated.
- Count what is left. That is unmatched.
- Confirm the three piles sum to the customer export. If they do not, find the gap before you read anything else.
- Compute cost per customer by channel twice: once on measured customers only, and once with the unmatched pile spread by measured share. The truth sits between the two, and the width of that range is exactly how much you do not know yet.
The first pass takes an afternoon. After that it is a monthly habit, or a report someone else runs for you.
That second option is what the Almanac does. It reads the customer report your CRM already emails, matches customers to calls and forms deterministically, and shows measured, estimated and unmatched side by side, summing to your actual customer count. It never blends the two, and it never claims to know which click drove a sale, because at local volume no one can. It also never touches your bids and never connects to your bank.