Closing the Loop on LSA and Direct Mail: One Ledger, Three Channels, Three Different Answers - Enterprise Digital Marketing

Closed-loop attribution applied to the two channels that never carry a click ID: Google Local Services Ads and direct mail. At one location, LSA leads phone-matched to the point-of-sale returned 4.4x gross profit (floor 2.1x). The mail vendor's matchback claimed 53 percent of the location's entire year; the front desk's own tags said 7.8 percent. Same ledger, same months. The method for each channel, and why the answers diverge.

August 20, 2026 · 7 min read · By Nick Martinelli

Two of the three paid channels a typical local service business buys produce no click identifier under any configuration. Google Local Services Ads deliver a phone call or a message, not a website session. Direct mail delivers a piece of paper. If closed-loop attribution depended on the click ID, neither channel could ever be measured against revenue. Both can be, because both leave a durable identity behind: a phone number on the lead, and a household address on the mail file.

This paper takes one single-location business in the book and closes the loop on both channels against the same point-of-sale ledger, then shows why the answers come out so differently from what each vendor reports. The companion paper on front-desk tagging covers the Google Ads leg at a different operator; this one is about the channels where the tag and the phone number are the only joins available.

The ledger

The location completed 2,384 tickets worth $1,823,423 in 2026 through mid-August. Every figure below is a share of that ledger, which is the only honest denominator. Vendor reports almost never state theirs.

Channel 1: Local Services Ads, closed by phone match

LSA is the cleanest closed loop a location can run that does not involve a developer, because every lead has a phone number and the platform charges per lead, so spend and lead count reconcile to the invoice.

Inputs. The LSA lead inbox export (249 leads in the window), the LSA billing report pulled monthly (May through mid-August: $6,530 charged for 188 billable leads, $34.74 per lead), and two fresh POS exports: the per-ticket file and the customer list.

The join. Normalize every lead phone to ten digits. Match to the customer list on phone; where phone fails, fall back to an exact, unique full-name match. Then attach every completed ticket for that customer within 90 days of the lead.

Results.

The 33 percent match rate is a floor, not a conversion rate. Leads that called from a different number than the one on file, leads that booked under a spouse's name, and leads that converted after 90 days are all unmatched. The owner also disputes junk leads actively (71 of 249 were uncharged or credited), so the denominator in the billing report is already cleaner than the inbox.

Two numbers deserve emphasis. First, the location's desk tag says "Google" without distinguishing LSA from paid search, which is why the phone match had to do the work the tag could not. Second, the strict floor of 2.1x counts only customers who had never been in the system before the lead, and it still clears the cost of the channel twice over on gross profit. That is the number to use when someone asks whether the channel is incremental.

Channel 2: direct mail, closed two ways that disagree

Direct mail has its own closed-loop claim built in: the vendor's matchback report, which joins the mail file to the location's customer list by address and counts every ticket from a mailed household as a response. The location's front desk runs a second loop in parallel by tagging the source on each ticket. The two loops were measured over the same ledger.

A channel that was 53 percent of the location would be the business. It is not. The 6.8x gap between the two loops is not a rounding disagreement; it is two different definitions of "attributed." The desk tag records who said they came because of the mailer. The matchback records who received a mailer and then bought anything, for any reason, during the window.

Why the matchback runs high. The vendor mails the location's own customer list. Its two "targeted" campaigns went to roughly 800 existing customers and claimed 26 to 34 percent response rates at $130 to $221 returned per dollar. Those customers were going to come back; the mail arrived in between. The vendor's own suppression files, which remove customers serviced in the prior eleven months from each drop, confirm they know which households are active. Their fall penetration report nonetheless counted 1,113 of the 1,255 suppressed households, representing $2.1 million in lifetime spend, inside the matchback base. The report takes credit for the customers the mail was designed to skip.

What the desk tag shows instead. Run as a prospecting channel, the mail spent about $41,000 in 2026 to produce 136 new-customer tickets at a $268 average ticket. As acquisition, it is under water. Its actual role is retention, with a 52 percent repeat rate among mail-tagged customers, and that is exactly the component the matchback counts several times over.

Four lenses on the same spend. Because the two loops disagree so badly, we report the mail return four ways and let the reader pick the definition that matches their question.

The LSA channel, measured the same strict net-new way, sits at 2.1x against mail's 0.53x. That comparison is only possible because both were closed against the same ledger with the same 90-day rule.

Why the three channels produce three kinds of answer

Laid side by side, the location's paid channels close their loops through different joins, and the reliability of each answer tracks the join, not the channel.

The pattern generalizes. A join on a durable personal identifier (phone, email) with a tight time window produces a floor you can defend. A join on a household address with a loose window produces a ceiling the vendor will call a result. The desk tag sits in between and is only as good as the discipline behind it.

The protocol for a multi-location operator

Every location should be able to produce the LSA table above from its own exports in an afternoon. The steps are the same at any scale:

1. Export the lead list with phone numbers from each lead channel, and the billing report for the same window.
2. Export the POS customer list and per-ticket file fresh, the same day.
3. Match on normalized phone first, exact unique name second. Record the match rate and publish it as a floor.
4. Attach completed tickets within 90 days. Report new-customer revenue separately from all revenue.
5. Apply the location's actual gross margin before comparing to spend.
6. For any vendor-supplied matchback, demand two numbers before accepting it: the prior-customer share of claimed responders, and the response rate on a holdout set of routes that were not mailed. A vendor who cannot produce either is reporting the ledger back to you.

The desk tag remains worth maintaining for every channel, because it is the only loop the front desk can see in real time. But it is the cross-check, not the source of truth. The phone match is.

Methodology notes

• LSA spend from the Local Services API account report, pulled in monthly windows (the full-range call times out). Lead inbox from the platform's CSV export.
• POS exports: per-ticket master (16,020 rows all-time) and customer list (14,243 rows). Mail-source figures from the Marketing Source column of the per-ticket export.
• Mail vendor figures from the vendor's own ROI export, penetration report, and 12 suppression files (2023 through 2026), each spanning 333 to 335 days of service history.
• Gross margin 59.1 percent is the location's reported blended margin for the period.
• All figures anonymized. The mail analysis is covered in more depth in the matchback audit.