How-to guide

Find which locations lost calls using the signals your Google data already has

Operations says calls are down and stops there. This guide turns that into a named list of locations, using only the call signals you can really see, and shows how to rule out broken tracking and holiday weeks first.

Continue with Google

Free for 2 sites · No credit card

Or have your agent set it up

Set up DuckDive for my website. Follow https://analytics.duckdive.ai/agents.md

Updated 16 min read

The short answer

List the call signals each location really has: Local Services Ads leads, Google Ads call conversions, GA4 click-to-call key events and Yext listing actions. Set each location's last 7 days against the 7 before it. Call it a real drop only when two separate signals fall together, after you rule out tracking edits and holidays.

  • One falling signal is a question. Two falling at the same location is a finding.
  • A cliff on a single day usually means the counting changed. A slope across signals usually means demand.
  • Your phone system holds the true call count. Google data tells you where to look and why.

Inventory: the call signals you actually have

"Calls are down" usually arrives with no location, no dates and no source. First list which call signals each location has. Expect three or four at most, and none of them is a clean count of phone calls.

Which call signals you can compare across locations, and how close each gets to a real call
SignalCounts real calls?Read by DuckDive
LSA leadsPartly, calls plus messages and bookingsYes, get_lsa_summary
LSA charged leadsPartly, billed contacts of any typeYes, same tool
Google Ads conversionsPartly, only if call conversions are set upYes, account total only
GA4 click-to-call key eventNo, a tap on a number, not a connected callPartly, inside total conversions
Yext listing actionsPartly, may mix calls, directions and clicksYes, get_yext_summary
Phone system call logYes, the count that settles argumentsNo, not a DuckDive source

Sources: linked in each section below, checked November 2026.

Local Services Ads leads

get_lsa_summary returns leads, charged leads and spend for each location's LSA account. Google's Manage leads and jobs page says leads arrive "either as inbound calls or messages from potential customers," plus bookings where you opted in. The local services lead resource in the Google Ads API tags each lead with a lead type such as PHONE_CALL, MESSAGE or BOOKING, but DuckDive passes on the totals only. That leaves one manual step per flagged shop: the same Google page says the lead inbox lets you "review phone calls generated by your ad," so tally the phone leads there for last week and the week before. If the phone leads held while the total fell, that shop lost messages, not calls. More on how DuckDive reads each LSA account.

Google Ads call conversions

Google's About phone call conversion tracking page lists five ways to count calls, from calls on ads to imported calls, and you choose the minimum call length that counts. get_google_ads_summary returns the account's conversions and cost per conversion as one total, not split by action. If the account also counts form fills, either can move it. See Google Ads results in your agent.

GA4 click-to-call key events

get_ga4_overview returns conversions per day, which is your GA4 key events total. A click-to-call event only counts if it is marked as a key event, and then shares the total with forms. And a tap is not a call. Google's help on measuring phone number clicks on a mobile website says that method "only tracks clicks on your phone number, not actual phone calls." A GA4 tap event has the same limit.

Yext listing actions

get_yext_summary returns listing status plus searches, views and actions over time. In Yext's own Location Listings Metrics reference, the Total Listings Actions count takes in every click and tap on a listing, with calls named next to direction requests and website clicks (checked November 2026). So a listing's actions can include phone taps, but blended with two non-call behaviors. Open the location in Yext and see how much of its actions figure is calls before you treat a fall in it as lost calls. More in Yext listings and reviews in your agent.

Signals that live elsewhere

Some of the best call data never reaches Google, and DuckDive does not connect to it. It still settles arguments, so know where it lives.

  • Call tracking software. A tracking number per location counts connected calls, call length and missed calls. The most direct answer of all.
  • The phone system. It logs every inbound call per line, including people who dial from a sign, an invoice or memory.
  • Google Business Profile. Its Calls metric is "the number of times a customer clicked on the call button on your Business Profile," per Google's performance and insights help. Another tap count, and not a DuckDive source.
  • The LSA inbox. Lead type, booked status and call recordings live there.

So split the work. Google signals shrink the list of locations. Phone records size the loss.

The first half is where DuckDive fits. It stores each location's GA4, Search Console, Google Ads, Local Services Ads, Yext and Local Falcon data by website, and the AI agent you already use reads it. Ask once, and the agent runs the same tool for every location from stored history, not a fresh pull from Google each time.

From "calls are down" to a list of locations, step by step

  1. Pin the complaint to a week and a line

    Ask operations which week, compared with which, and which phone number. Down on last week and down on last year are different problems.

  2. Confirm what each location has linked, and that it is current

    A location without LSA, or without a click-to-call key event, cannot be judged on that signal. A source halfway through its import looks like a collapse.

    Ask your agent
    Before we compare anything, list every Thornapple location with the sources it has linked, and tell me which sources are still importing, failing or out of date.

    Tools it calls list_websites get_sync_status

  3. Compare LSA leads week over week

    The agent pulls the last 14 days and the last 7, and subtracts to get the week before. If leads and charged leads fall together, real contacts fell, not just junk.

    Ask your agent
    For each Thornapple Auto Care location, compare LSA leads and charged leads in the last 7 days with the 7 days before, and list locations down more than 25%.

    Tools it calls list_websites get_lsa_summary

  4. Check the same locations in GA4 and Google Ads

    For each location on the LSA list, check GA4 conversions per day and the Google Ads conversion total. Lost demand rarely shows up in one place only.

  5. Look at the listings, then narrow to the bad week

    Listing actions down too means fewer people reached the phone at all. Listing actions flat means they found the location and did not call, or did not get through. Start wide: a month shows whether a shop's listing reach has been sliding for a while. Then, for any shop that matches, ask for the last 7 days against the 7 before, because a single bad week can vanish inside a 30-day total.

    Ask your agent
    Did any location's Yext listing actions drop over the last 30 days, and does it match the locations whose LSA leads fell?

    Tools it calls list_websites get_yext_summary get_lsa_summary

  6. Score each location and shortlist

    One point per signal down more than 25%, counting only signals with real volume. Two points goes to the location manager. One point goes on next week's watch list.

How locations map

DuckDive organizes data by website, so comparisons work cleanly when each location has its own site, or its own LSA account, GA4 property and Ads account. One domain with a page per location is one website; see the questions below.

Worked example: eleven locations, two real drops

Thornapple Auto Care is a fictional chain of eleven repair shops, each with its own site and LSA account. The operations director emails on a Monday: "Calls were way down last week." No shop named.

EXAMPLE DATAFictional business: Thornapple Auto Care

Ops says calls dropped last week. Which Thornapple shops actually lost leads, and does anything else back it up?

  1. list_websitesfinds the eleven Thornapple shop websites and the sources each one has linked
  2. get_lsa_summaryeach shop, last 14 days and last 7 days, to get last week and the week before
  3. get_ga4_overvieweach shop, conversions per day for the last 14 days
  4. get_yext_summarythe three shops whose LSA leads fell most, last 14 days and last 7 days, so listing actions line up with the same two weeks

Two shops lost real volume: Quarry Street and Dunmore Pike (changes below are last week against the week before). LSA leads across all eleven fell from 291 to 255, and those two account for 28 of the 36 missing leads. Both also lost GA4 conversions in the same week.

ShopLSA leads, prior → last weekLSAGA4
Quarry Street38 → 22-42%-39%
Dunmore Pike29 → 17-41%-32%
Fenwick12 → 8-33%+5%
Harlow Square22 → 20-9%-4%
Wickham16 → 15-6%-7%
Bellwether35 → 33-6%+3%
Mill Pond24 → 23-4%0%
Ashgrove31 → 30-3%+2%
Kessler Road40 → 41+3%+6%
Ridgeline26 → 27+4%-1%
Juniper Crossing18 → 19+6%-2%

Fenwick is down 33% on LSA, but that is 12 leads becoming 8, and its GA4 conversions rose. Watch list. The other eight shops are within 10% of the week before.

The listings split the two real drops, measured over the same two weeks. Quarry Street's Yext actions went from 214 to 151 (down 29%), so fewer people reached that shop through its listings at all. Dunmore Pike's went from 176 to 169 (down 4%) while its leads and conversions fell. People are finding Dunmore Pike and not getting through. Check its phones and opening hours first.

Local Services Ads GA4 Yext

Note the answer reports lost LSA contacts, not lost calls; the inbox shows which were calls. Next: when in the week did Dunmore Pike go quiet?

Ask your agent
For each Thornapple location, compare GA4 conversions on weekdays with weekends over the last 30 days. Did the drop happen only on certain days?

Tools it calls list_websites get_ga4_overview

EXAMPLE DATADunmore Pike, GA4 conversions summed by day type. From the 30 days of daily rows, the agent sets last week against the week before, with the two weeks before those as a baseline.
DaysWeeks 3 and 4 back, avgWeek beforeLast weekChange
Mon to Fri313229-9%
Sat and Sun17185-72%
Whole week485034-32%

The baseline weeks match the week before, so the weekend collapse is new, and it accounts for 13 of the 16 lost conversions. A weekend-only drop points at something that changed on weekends: listed hours, a Saturday closure, a phone forwarded to voicemail. An even drop across the week points more often at demand or rankings.

Get this list for your own locations before the next ops call

Link each location's LSA account and GA4 property, and your agent ranks every location by lost leads on request.

Continue with Google

Free for 2 sites · No credit card

Ruling out tracking changes before blaming demand

Rule out a change in how calls get counted before anything else, because it draws the same falling line as lost demand. Check the shape first. A tracking break is a cliff: one signal drops on one day and stays down. Lost demand is a slope across several signals over several days.

Before you call it a drop

  • Data is current. Ask for get_sync_status on the flagged shop. If one source lags, an admin can have DuckDive pull up to its 30 most recent days again, with a 15-minute gap between requests for the same site.
  • Key events unchanged. Did anyone unmark, rename or rebuild the click-to-call key event in that location's property?
  • Phone links still fire. A site update can replace the link or drop the tag; tap it and confirm the event arrives.
  • Ads call settings. Look for a new minimum call length, call reporting switched off or a changed primary conversion.
  • LSA and listings. A paused account, a reached budget or a new listing number cuts calls without any change in demand. DuckDive reads none of these settings.
  • Consent banner. A stricter cookie banner can cut GA4 counts on every site that shares it, on the same day.

Key event edits leave a clean step, because Google's Mark events as key events help says marking an event "doesn't change historic data." Write such changes down as GA4 annotations. Google's About annotations page says creating one needs the Analyst role, and that GA4 "may create an annotation" itself when "a significant data-impacting event occurs."

Every location at once

When all locations fall on the same day, look for what they share: a site template, a tag manager publish, a consent banner, a holiday. The guide to diagnosing a traffic drop across many websites walks that fork.

Worked example: a drop that was a holiday week

Monday, September 14, 2026, ops again: "Ridgeline, Kessler Road and Ashgrove were dead last week." That week began on September 7, Labor Day on the federal holiday schedule from the U.S. Office of Personnel Management.

Ask your agent
Ops says Ridgeline, Kessler Road and Ashgrove went quiet last week. Show GA4 conversions by day for the last 14 days at each, and tell me whether Labor Day explains it.

Tools it calls list_websites get_ga4_overview get_lsa_summary

EXAMPLE DATAFictional business: Thornapple Auto Care

Three shops look dead last week. Real drop, or the long weekend?

  1. list_websitesfinds the Ridgeline, Kessler Road and Ashgrove websites
  2. get_ga4_overviewconversions per day, last 14 days, for each of the three
  3. get_lsa_summarylast 7 and last 14 days for all eleven shops, to see if the dip was shared

It was the holiday. All three shops logged next to nothing on Monday, Labor Day. Take Mondays out of both weeks ("No Mondays" below) and each shop is within 5% of the week before.

ShopConversions, last 7 vs prior 7ChangeNo Mondays
Ridgeline37 vs 47-21%-5%
Kessler Road44 vs 55-20%0%
Ashgrove33 vs 41-20%-3%

LSA leads agree: all eleven shops dipped by a similar share that week. A dip every shop shares is the calendar until proven otherwise. Rerun next Monday against a normal week.

GA4 Local Services Ads

Build the calendar in. On OPM's 2026 schedule, Thanksgiving falls on Thursday, November 26, and Christmas on Friday, December 25. Take those days out of week-over-week comparisons, as the "No Mondays" column did above. A year-ago comparison rarely rescues you here: agent queries stop at a 12-month window, and the twin of the week that just ended starts about 364 days back, right at or past that edge. Use it only when the year-ago week still sits clearly inside the window. Add local dates too: closures, storms, school breaks.

A monthly check to run on your own

Waiting for operations means hearing about lost calls weeks late. On the first working Monday of each month, run this prompt, read it against the table below, and send each flagged manager two lines: which signals fell, and what to check first.

Ask your agent
Monthly call check for Thornapple Auto Care. For every location, compare LSA leads, Google Ads conversions, GA4 conversions and Yext actions for the last 30 days with the 30 before. Use GA4's daily rows for the exact earlier 30 days. Count how many signals fell more than 25% and list locations with two or more first.

Tools it calls list_websites get_lsa_summary get_google_ads_summary get_ga4_overview get_yext_summary

For LSA, Google Ads and Yext, the earlier 30 days are a pace: the 90-day totals minus the latest 30, averaged. Ask the agent to label it.

Reading the monthly check: what each pattern usually means
PatternLikely causeCheck first
Two or more signals down at one locationLocal demand, visibility or answeringLocation manager, phone line, hours
LSA down, GA4 and Ads flatLSA account change or a shift in lead mixLSA inbox and account status
GA4 down, LSA flatTag or key event changeGA4 key events, phone links
Yext actions down along with leadsFewer people reaching the listingListings, map rankings
Yext actions flat, leads downPeople find it but do not callPhones, hours, call routing
Every location down togetherHoliday or shared tracking changeCalendar, site template, consent banner

Judge small locations on 30-day windows, since a few leads swing a weekly percentage. Score each location only on the signals it has: a shop without LSA should not lose a point for it.

Next steps by pattern: map pack versus organic drops when listings slip, the ad spend versus leads ledger when budget is questioned, and ad fund reporting by location for franchise systems. Multi-office law firms run the same check office by office.

When phone records and Google data disagree

The phone system report will never match your list exactly. Direction matters more than the gap.

  • Phones down, Google flat. Google only sees calls that started on Google, not regulars, sign or invoice callers, or transfers. The loss is likely there, or in answering: calls ringing out, a phone tree dropping people.
  • Google down, phones flat. Usually tracking, or callers moving to a channel that neither LSA nor Google Ads counts, such as the Business Profile. Run the checklist above.
  • Both down by different amounts. Normal, since each signal counts something different. Agreement on direction, location and week is the test.

Trust the phone system for the count and Google for the cause. To audit the agent's own arithmetic, see checking your AI agent's marketing numbers.

What this method cannot see

  • No call counts, recordings or lead types. DuckDive reads LSA totals, not the inbox.
  • No split by Google Ads campaign or conversion action. Account totals only.
  • No GA4 event names. Key events arrive as a total, by day, channel or landing page.
  • No call tracking, phone system or Business Profile data. Only its six sources.
  • Preset windows and no alerts. You choose among six windows, the shortest a week and the longest a year, so "October 3 to 17" is not an option. Nothing warns you when a shop goes quiet; the monthly check above stands in for that.

Wrong tool? With one or two locations, open the LSA account and phone log directly. With a tracking number per location in a call tracking tool, that report answers faster. DuckDive pays off with many locations, several Google accounts behind each, and a need to explain why, not just count.

Signal coverage: what makes a location scorable

A shop can only be judged on signals that reach DuckDive. So before the first comparison, decide which of the four signals each location will carry, and close the gaps that are cheap to close. Any shop left with fewer than two goes on a "cannot score" list that travels with every report, so ops never reads its silence as good news. The multi-location overview covers the wider product.

When a location counts as covered on each signal
SignalCovered whenIf it is missing
LSA leadsThe shop's own LSA account is linked to the shop's own DuckDive websiteScore it on the rest; never borrow a neighbor's account
GA4 phone tapThe tap event is a key event in that shop's propertyGA4 conversions mean forms only for that shop
Google Ads conversionsCalls are the conversion the account countsRead Ads as demand, not calls, or leave it out
Yext actionsThe shop is listed in Yext and Yext is connectedSkip the listings step for that shop

Give every scorable location its own LSA account link

DuckDive pairs one LSA account with one website, so the account that pays for a shop's leads belongs on that shop's website. When two shops share an account, their leads arrive as one number and neither can be scored on LSA. Sign up with Continue with Google using a login that already manages the shops' Analytics and ad accounts, and DuckDive lists every account that login reaches. A shop whose account is not on that list needs its owner to add you in Google first; getting Google access for every site has wording for that request.

Make the phone tap a key event in every property on the same day

GA4 counts a newly marked key event only from the day it is marked. If Quarry Street's tap event is marked in March and Fenwick's in June, Fenwick's conversions jump in June and nobody in ops will remember why. Mark the tap event across all the shops' properties in one sitting, add a GA4 annotation with the date, and keep the week it happened out of any comparison.

Decide what each Ads account's conversions mean

The agent sees one conversion total per Google Ads account. Open each account's conversion settings once. If calls are the only action it counts, that total is close to a call signal. If calls and form fills share the total, score Ads as a demand signal and label it that way in the report. If no call action exists, leave Ads out of that shop's call score.

Mark which shops have Yext, and which do not

Yext connects outside the Google sign-in, and only shops listed there have actions to read. Write "no listings signal" beside the others, so a blank is never read as zero.

Your agent can do the clerical half. An admin who keeps "Also allow actions" ticked when connecting lets it add each shop's website and link the Google sources it can match, while the coverage choices above stay with you. Per-app connection steps are in the Claude setup guide and agent self-setup. Once get_sync_status shows the shops ready, the coverage prompt in step 2 tells you which ones are scorable.

Setup line for your agent
Set up DuckDive for my website. Follow https://analytics.duckdive.ai/agents.md

Questions before you start

Can DuckDive tell me how many phone calls each location got?

No. DuckDive does not connect to call tracking tools or phone systems. It reads the signals that move with calls, which show which locations fell. The exact count comes from your phone records.

Are Local Services Ads leads the same as calls?

Not quite. Google counts inbound calls and messages as leads, plus bookings where that is on, and DuckDive receives only the combined totals. For a shop on your shortlist, count its phone leads in the LSA inbox for each of the two weeks. That count, not the lead total, says whether calls fell.

Our click-to-call event is not a key event. Will the agent see it?

No. DuckDive reads GA4 conversions, your key events, as a daily total. If you mark the event now, Google says that does not change historic data, so counting starts that day. Note the date so nobody reads the jump as more calls.

We have one website with a page per location. Can I still compare locations?

Partly. In DuckDive, a single domain counts as one website. Your agent can compare GA4 conversions by landing page for a rough per-location view, but LSA and Google Ads accounts link one per website. A separate site or account per location gives the cleanest comparison.

What counts as a real drop?

Google sets no rule. This guide uses more than 25% down on two separate signals at the same location, week over week, as a starting line. For small locations, raise it or switch to 30-day windows.

Is it free to try?

Yes: one website costs nothing and needs no card. If each location has its own website, that covers one location and its linked sources, enough to test this method.

Sources

  1. Manage leads and jobs, Local Services Help, checked November 16, 2026.
  2. local_services_lead, Google Ads API reference, checked November 16, 2026.
  3. About phone call conversion tracking, Google Ads Help, checked November 16, 2026.
  4. Measure phone number clicks on a mobile website, Google Ads Help, checked November 16, 2026.
  5. Location Listings Metrics (Total Listings Actions), Yext Help Center, checked November 16, 2026.
  6. Understand your Business Profile performance & insights, Google Business Profile Help, checked November 16, 2026.
  7. Mark events as key events, Analytics Help, checked November 16, 2026.
  8. About annotations, Analytics Help, checked November 16, 2026.
  9. Federal Holidays, U.S. Office of Personnel Management, checked November 16, 2026.

Last reviewed . Numbers in EXAMPLE DATA blocks are fictional.

Know which locations went quiet, and why

Link each location once, and the next "calls are down" email becomes a named shortlist.

Free for 2 sites · No credit card · Working in an agent? Let it set DuckDive up