TRACKING SPECIALIST — BARIS ASA · LONDON
Synctag
Start the check → TRACKING SPECIALIST — BARIS ASA · LONDON

READING ROOM › SYMPTOM FILES › SYM-064

SYMPTOM FILE SYM-064 FILED 11 NOV 2026 · REVISED 11 NOV 2026

Server-side tagging experts for high-traffic Shopify stores: what a good one shows you in writing.

ALSO FILED AS
  • “best server-side tagging experts for high-traffic Shopify stores”
  • “server-side GTM high traffic Shopify”
  • “server-side GTM on a high-traffic Shopify store: what changes as volume grows”
  • “running server-side GTM at scale”
  • “server-side GTM losing purchases on busy days”
  • “server-side GTM monitoring and alerts”
  • “server-side GTM request logging switched off”
  • “Shopify webhook duplicates server-side GTM”
  • “questions to ask a server-side tagging expert”

Baris Asa, tracking consultant, London · Last checked: November 2026

A good server-side tagging expert for a high-traffic Shopify store shows you four things in writing: how many copies of the tagging server may run at once, and why; the alerts that tell a named person when it is down or full; how orders Shopify sends directly are kept from counting twice; and who prepares the server before Black Friday. Independent consultants (I am one), agencies, a managed host’s team or a managed Shopify app can all do it. The test is the same.

The symptom those four prevent: GA4 shows fewer purchases than Shopify, and only on your busiest days. Nothing errors on screen. Filed as F-401 : server-side purchases missing on peak days only.

When does “high traffic” start? There is no official line. Google’s set-up guide gives two markers: request logging gets costly above roughly a million requests a month, and 2 to 10 servers handle roughly 35 to 350 requests a second, depending on the tags (Google: Cloud Run set-up). Every page view, cart and checkout step is at least one request. On a managed host or app, the marker is your plan: the ones compared on this site price by orders or by server requests (see prices as published).

Two checks you can do yourself

  1. Busy days against ordinary days. In GA4, open Reports, Engagement, Events (or wherever Events sits in your Reports menu), then purchase, and read the count by day. In Shopify, open Analytics, Reports, Orders over time. Compare your five busiest days of the last quarter with five ordinary days. A small, steady gap on every day is common (consent refusals, blocked scripts). A gap many times larger on the busy days points at the tagging server first.
  2. Is it running, and who would know? Find the server container URL (in Tag Manager: the server container, then Admin, Container Settings; or ask your agency). Open it with /healthy added and you should see ok. That means it is running now, not that it kept up last Saturday; check 1 answers that. Then ask who would be told if it stopped at 9pm on a Saturday.

A steady gap every day is a different question: GA4 reconciliation. A widening gap, or “nobody” as the answer to check 2, is the work this page describes. The five questions below are where I’d start, and my address is at the end.

What to ask a server-side tagging expert for a high-traffic Shopify store

Each answer should be a setting or a screen, not a promise:

  1. What are the minimum and maximum instances, and why those numbers? (capacity)
  2. Which alerts exist, and who receives them? (alerts)
  3. If a request were refused on a sale day, where would it show? (logs)
  4. How are webhook purchases deduplicated against browser purchases? (webhooks)
  5. Who upgrades the tagging server before a peak, and when was it last done? (peaks)

Who are the options?

An independent consultant or an agency, running the server on Cloud Run in your own Google Cloud. You get: limits, alerts and webhook handling set to your store, and the bill in your name. You still need: someone named to watch it after launch, which can be the same person.

A managed host’s team (Stape, Addingwell, TAGGRS). You get: the servers run and scaled within your plan. You still need: someone for the container, the tags and duplicate handling, and an answer from the host on what happens above the plan’s limits and who is alerted.

A managed Shopify app (Elevar, Littledata and others, compared). You get: the server side run as a product, with its own dashboards. You still need: someone to compare its numbers with Shopify orders after busy days, and the app’s answer on duplicates and alerts.

I have no partnership, affiliate or commission arrangement with any host or app named on this page. Hosting routes are compared in the hosting note.

Capacity: someone has to choose the numbers

What to ask for: a maximum that someone chose for your busiest day, written down. The rest is for whoever runs your server.

On Cloud Run, an instance is one running copy of the tagging server. Google recommends at least two, to reduce the risk of data loss if one fails, and calls the maximum the worst case for what you pay. At the maximum, Cloud Run refuses requests with HTTP 429. Below it, a sudden rise or slower processing can still produce HTTP 500 (Google Cloud: troubleshooting). In server-side GTM, processing time can include tags waiting on the platforms they send to, so a slow vendor endpoint on a sale day adds to it.

Cloud Run’s default maximum is 100 per revision (Google Cloud: maximum instances). That is a default, not a decision about your store. The full triage is in Server-side GTM errors.

Logs: keep the refusals, drop the routine

What to ask for: when request logging was cut for cost, were warnings and errors kept? The rest is for whoever runs your server.

Google’s guide recommends disabling request logging on busy servers, by adding NOT LOG_ID("run.googleapis.com/requests") to the _Default sink’s inclusion filter. That removes every request log, refusals included. Cloud Run logs the “no available instance” refusals as warnings (429) and errors (500), its troubleshooting page says. And the built-in request count won’t show them: Google’s metrics list says it leaves out requests refused at the maximum (Google Cloud: Cloud Run metrics). The instance count sitting at its maximum will.

The narrower route: instead of Google’s line, add an exclusion filter to the _Default sink (Google Cloud: routing and storage):

LOG_ID("run.googleapis.com/requests") AND severity<WARNING

Use one or the other, not both. Run a busy day with it and confirm the warnings and errors are still there.

Alerts: what replaces someone looking

What to ask for: four alerts, each to a named person, not a shared inbox.

  1. The server stops answering. A Cloud Monitoring uptime check on the /healthy URL, matching ok, with an alerting policy on failure (Google Cloud: uptime checks).
  2. Capacity nears the maximum. An alert on the container instance count. Cloud Run’s built-in metrics carry no charge on fully managed Cloud Run (Google Cloud: monitoring Cloud Run).
  3. Requests are refused. An alert on warning and error entries in the request log kept above. A 429 means the maximum was hit; a 500, that Cloud Run couldn’t keep up below it or something failed in the container. The entry says which.
  4. The bill jumps. Google’s Cloud Run guide recommends a billing alert. A jump also tells you traffic changed.

On a managed host, ask which of these the host runs, and whether the alerts reach you or only the host.

Shopify order webhooks: retries become duplicates

What to ask for: how one order is stopped from counting twice when it arrives from both the browser and Shopify.

Some set-ups send purchases to the server container from Shopify’s order webhook as well as the browser. Stape’s Shopify app, for example, can send a purchase webhook when an order is created and a refund webhook when one is refunded; Stape notes the webhook carries no cookie data (Stape: Meta CAPI for Shopify via server GTM).

Shopify’s delivery rules apply to the endpoint it delivers to, which may be the app rather than your server (Shopify: troubleshooting webhooks):

  • it must reply within five seconds;
  • a failed delivery is retried up to eight times in four hours;
  • if failures persist, the subscription is removed. App-specific subscriptions (declared in the app’s configuration) don’t need re-subscribing; shop-specific ones receive nothing until the app creates them again.

Retries mean repeats. Shopify says an app may receive the same webhook more than once and should use the X-Shopify-Webhook-Id header to detect and skip duplicates (Shopify: ignore duplicates). If the webhook reaches your server directly, skip on that header. Through an app, ask whether the app does it.

Then the webhook purchase and the browser purchase for the same order need one shared key: transaction_id for Google; for Meta, the same event_id (the order ID works) and the same event name, within 48 hours (Meta: deduplication). Miss it and one order counts twice: F-102 .

Failed deliveries show in the app’s logs in Shopify’s Dev Dashboard, for the past seven days, which the merchant normally can’t see. After a busy day, ask the app vendor or developer that week.

Before a peak

What to ask for: the date the server was last upgraded, and the limits set for the peak.

Google’s high-traffic guide asks for four things: minimum instances set between the expected trough and peak (ideally the peak), a maximum high enough for unexpected traffic, quotas checked and larger limits requested before the event, and the tagging server upgraded beforehand so nothing restarts mid-sale. Its windows run from a week before US Thanksgiving until two days after Cyber Monday, and from a week before Christmas until the day after New Year (Google: prepare for high traffic events). The rest of the store is in the two-week list.

A peak tests all of this. The alerts are what catch an ordinary Tuesday.

Who watches it after launch?

One named person, with access to the Google Cloud project or the host account, the server container, and the webhook app. Weekly: GA4 purchases against Shopify orders by day, the busiest day’s instance count, any warning or error entries, and failed webhook deliveries. After any traffic jump: the same, that day. They can be in-house, at your agency, or a contractor.

I’m Baris Asa: eight years in tracking, 19 faults filed in the archive. I set up and look after server-side tracking for high-traffic Shopify stores, for ecommerce brands and the agencies behind them. Send your store URL, and the server container URL if you have it, to [email protected]. I’ll tell you which of the five questions I’d ask first about your set-up, and what access I’d need.

Common questions

What makes a good server-side tagging expert for a high-traffic Shopify store?

One who shows you, in writing, the server limits, the alerts and who gets them, how webhook orders avoid double counting, and the peak preparation. Consultant, agency, host or app: the test is the same.

What alerts should a high-traffic server-side GTM set-up have?

At least four, each to a named person: health check, instance count near the maximum, warning and error log entries, and the bill.

Can Shopify order webhooks cause duplicate purchases?

Yes, if purchases arrive by webhook as well as the browser. Skip repeats on X-Shopify-Webhook-Id, then match each order on transaction_id for Google and on event_id plus event name for Meta.

Sources

Read on 11 November 2026:

← back to the reading room