<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Stack Architect on Medium]]></title>
        <description><![CDATA[Stories by Stack Architect on Medium]]></description>
        <link>https://medium.com/@stackarchitect123?source=rss-fe981e26d7b8------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*PO544JSFdbRLFh7ZkwnlYA.png</url>
            <title>Stories by Stack Architect on Medium</title>
            <link>https://medium.com/@stackarchitect123?source=rss-fe981e26d7b8------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 22:26:48 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@stackarchitect123/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[My 62-page site sent Bing 55,000 URLs. I found out four months later.]]></title>
            <link>https://medium.com/@stackarchitect123/my-62-page-site-sent-bing-55-000-urls-i-found-out-four-months-later-5e9795356e12?source=rss-fe981e26d7b8------2</link>
            <guid isPermaLink="false">https://medium.com/p/5e9795356e12</guid>
            <category><![CDATA[web-development]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[seo]]></category>
            <category><![CDATA[cloudflare]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Stack Architect]]></dc:creator>
            <pubDate>Mon, 16 Nov 2026 17:06:35 GMT</pubDate>
            <atom:updated>2026-11-16T17:06:35.376Z</atom:updated>
            <content:encoded><![CDATA[<h3>An Astro IndexNow integration and a Cloudflare setting I’d forgotten about, four things that misled me about indexing, and the checks to run before you blame Google.</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*yg4EMqyviJSv9gX2.png" /></figure><p><strong>Two submitters I had forgotten about sent Bing about 55,000 IndexNow URLs for a 62-page site between April and September 2026: an Astro integration that re-sent every page on every build, because its change cache lived in a gitignored file, and Cloudflare Crawler Hints, which pings IndexNow whenever a deploy changes the HTML.</strong></p><p>I run <a href="https://stackarchitect.xyz/">Stack Architect</a>, a static Astro site on Cloudflare Pages with free Shopify automation guides. It has 38 pages in its sitemap today and had 62 at the time. Google has indexed one of them, the homepage. For months I treated that as a mystery to solve with one more technical fix.</p><p>This is what I found instead, in the order I found it. Three were real problems, two of them running for months. Four were tools and dashboards telling me something that wasn’t true. All of it is dated, and most of it you can check on your own site in under ten minutes.</p><h3>The number</h3><p>On 18 September 2026 I exported the IndexNow report from Bing Webmaster Tools. It showed about <strong>55,000 URLs submitted since 22 April</strong>, for a site with 62 pages in its sitemap.</p><p>The busiest days were 23 July (4,886 URLs), 22 July (2,508) and 21 July (2,392). Even on quiet days, 20 to 77 URLs were still going out. Over the same period Bing indexed <strong>33</strong> URLs, every one of them before 24 July, and then crawled nothing new for eight weeks.</p><p>I hadn’t written anything that sent 55,000 URLs. Two things I’d switched on and forgotten had.</p><h3>Culprit 1: an IndexNow integration that couldn’t remember anything</h3><p>On 28 April I added an Astro IndexNow integration. It’s the kind of package you install in a minute: on every build it collects the pages in dist/ and tells IndexNow about them, so Bing hears about new content straight away.</p><p>It also promised to send only pages that had changed. It did that by hashing each page and comparing the hashes with a cache file it kept in the project root. That file was in my .gitignore, which is exactly where a generated cache belongs.</p><p>Cloudflare Pages builds every deploy from a fresh clone of the repo. A fresh clone never has the cache file, so on every build the integration found an empty cache, decided every page was new, and submitted all of them.</p><p>Three things made it worse:</p><ol><li><strong>It ran on preview builds too.</strong> Every branch push built a preview, and every preview submitted the production URLs.</li><li><strong>I had an AI coding agent opening branches.</strong> Roughly 50 of the 355 successful builds between 28 April and 6 August were previews of the agent’s branches.</li><li><strong>It submitted everything in </strong><strong>dist/.</strong> That included noindex pages and pages that later became redirects.</li></ol><p>At 80 to 190 URLs a build, that adds up to the total. The worst days line up with the days I pushed most.</p><p>I removed the integration on 6 August. The submissions didn’t stop.</p><p><strong>The general lesson:</strong> any “only send what changed” logic that keeps its state on the build machine doesn’t work on ephemeral CI. It fails without an error, because from the tool’s point of view everything really is new every time.</p><h3>Culprit 2: a Cloudflare setting that turned itself back on</h3><p>After 6 August the only live submitter was <strong>Cloudflare Crawler Hints</strong>, a zone setting that lives in the dashboard, not the repo. It had been switched on at 10:44 UTC on 21 April. Bing’s export starts the next day.</p><p><a href="https://developers.cloudflare.com/cache/advanced-configuration/crawler-hints/">Crawler Hints</a> notifies IndexNow when Cloudflare sees content change. That’s sensible for a site whose pages change when their content changes. On a static site that redeploys several times a day, every deploy changes the HTML, so it fired constantly. That was the 20 to 77 a day.</p><p>I turned it off on 18 September.</p><p>On 19 October I opened Bing’s IndexNow page again and found Cloudflare submitting /_astro/*.woff font files and /diagrams/*.svg images about every 15 minutes. Crawler Hints was on again. Nothing in git had changed. I turned it off a second time and reloaded to confirm it stayed off.</p><p><strong>The general lesson:</strong> dashboard settings are configuration you can’t diff, review or roll back. If one matters, write down where it is and check it on a schedule. “I turned that off” is a statement about the past.</p><h3>Culprit 3: a redirect rule nobody could see</h3><p>While checking the zone I found a Cloudflare <strong>Redirect Rule</strong> called “enforce trailing slash”. It ran before my own middleware and did the same job, which would have been harmless except for two side effects:</p><ul><li>It added a 301 in front of every affiliate redirect, so each one took two hops instead of one.</li><li>It could match my API routes. A 301 on a POST makes most clients retry as a GET with no body, and a trailing-slash redirect of exactly that kind is how a form on the site had earlier lost every submission without an error.</li></ul><p>None of this was in the repo, so no code review could have caught it. The rule is now switched off, and the setup notes in the repo record that it exists.</p><h3>Four things that told me something false</h3><p>Fixing real problems is the easy part. Most of my wasted time went on things that looked like evidence.</p><h3>1. An AI fetcher’s cache</h3><p>On 9 October a summarising fetch tool showed a page with content I had replaced hours earlier, while the same URL with a throwaway query string showed the new version. I concluded Cloudflare was serving stale HTML to crawlers. I blamed a one-year Cloudflare-CDN-Cache-Control header, removed it, and told myself to purge the cache and resubmit.</p><p>Then I ran my own check, which fetches each page and reads the response headers. Every HTML response said cf-cache-status: DYNAMIC. Cloudflare wasn&#39;t caching any of them, because a Pages Function runs on every request. The stale copy was in the fetch tool&#39;s own cache.</p><h3>2. A fetcher that invented a status code</h3><p>In August a fetch tool reported that two subdomain redirects were 302s, which would have been worth fixing. They were 301s. The tool had followed the chain, reported the second hop’s destination, and made up a status for the first. One command settles it:</p><pre>curl -sI https://audit.stackarchitect.xyz/ | head -3</pre><p>curl -sI shows you the first response exactly as the server sent it. Anything that summarises is a different thing, however confident it sounds.</p><h3>3. My own crawler counts</h3><p>I log page requests per day with the claimed client in a small database table, without cookies, IP addresses or user agents stored. It showed 302 of 518 “Googlebot” requests ending in 404, which looked like Google hitting broken links.</p><p>Most of them were vulnerability scanners using a Googlebot User-Agent to probe paths like /.env and /.ssh/id_rsa. Cloudflare&#39;s security events showed every blocked &quot;Googlebot&quot; coming from Google Cloud addresses (35.242.x, 35.246.x), which anyone can rent. The one real Googlebot request that week came from 66.249.x, Google&#39;s own network (AS15169).</p><p>A User-Agent is a claim. Google’s own guidance on <a href="https://developers.google.com/search/docs/crawling-indexing/verifying-googlebot">verifying Googlebot</a> is to check the IP, by reverse and forward DNS lookup or against its published ranges. For how often Google really crawls you, Search Console’s Crawl stats report is the source: this site had 1,380 requests in 90 days, 86% answered 200, average response 111 ms. Healthy.</p><h3>4. My own notes</h3><p>In August my notes said “Bing indexes the site normally”, and I had built a theory on it: Bing was fine, so the problem must be specific to Google. Bing’s own export said 5 clicks and 129 impressions in 180 days, with 53 consecutive days of zero impressions at the end. The theory had nothing under it.</p><p>Writing a conclusion down makes it feel like a fact. Re-read your notes against the source data before you build on them.</p><h3>What the evidence actually says</h3><p>After all of that, here is where it stands, read directly from both webmaster tools.</p><p>Search Console’s URL Inspection for a typical page says <strong>Crawled — currently not indexed</strong>: fetched successfully as Googlebot smartphone, crawling and indexing allowed, my canonical accepted, no different canonical chosen by Google. Every page I’ve tested answers 200 to a browser, Googlebot, Bingbot and the main AI crawlers. Bing’s live URL test says the page can be indexed, with no SEO issues.</p><p>So Google can fetch every page and is choosing not to index them. That isn’t a fault I can fix with a header. It’s a judgement about the site: a young domain with few links, whose old URLs returned 404s for three months after a move in April, until I added redirects in July.</p><p>Knowing that changed what I work on. I set fifteen general guides that mostly restated vendor documentation to noindex, so Google judges the site on its original work. I add first-hand material, such as screenshots from my own accounts, instead of more words. One guide has gone back into the sitemap on that basis. Whether that works is a question of weeks, and I’ll report back either way.</p><h3>The IndexNow rules I use now</h3><p>I still use IndexNow, deliberately and by hand. The script that sends it follows five rules, and they apply to any site:</p><ol><li><strong>Only canonical URLs from the live sitemap.</strong> Never redirect sources, noindex pages or anything you wouldn’t want indexed.</li><li><strong>Only URLs whose </strong><strong>lastmod changed since the last send.</strong> Don&#39;t hash the HTML; a build changes it every time.</li><li><strong>Keep the record of what was sent in the repo</strong>, not on the build machine, and commit it after every run.</li><li><strong>Cap each run.</strong> Mine stops at 60.</li><li><strong>Dry run by default.</strong> Sending takes a flag, and it only happens after a production deploy, never from a build.</li></ol><p>Here is the core of it, small enough to adapt:</p><pre>// Send IndexNow only for sitemap URLs whose lastmod changed since the last run.<br>import { readFileSync, writeFileSync } from &#39;node:fs&#39;;</pre><pre>const SITE = &#39;https://example.com&#39;;<br>const KEY = process.env.INDEXNOW_KEY;          // key file served at /&lt;KEY&gt;.txt<br>const LOG = &#39;data/indexnow-sent.json&#39;;          // committed to the repo<br>const CAP = 60;<br>const submit = process.argv.includes(&#39;--submit&#39;);</pre><pre>const xml = await (await fetch(`${SITE}/sitemap-0.xml`)).text();<br>const live = [...xml.matchAll(/&lt;url&gt;([\s\S]*?)&lt;\/url&gt;/g)].map(([, u]) =&gt; ({<br>  url: u.match(/&lt;loc&gt;([^&lt;]+)/)[1].trim(),<br>  lastmod: u.match(/&lt;lastmod&gt;([^&lt;]+)/)?.[1].trim() ?? null,<br>}));</pre><pre>const sent = JSON.parse(readFileSync(LOG, &#39;utf8&#39;));<br>const changed = live.filter(({ url, lastmod }) =&gt; sent[url] !== lastmod).slice(0, CAP);<br>console.log(`${changed.length} changed of ${live.length}`);</pre><pre>if (submit &amp;&amp; changed.length) {<br>  const res = await fetch(&#39;https://api.indexnow.org/indexnow&#39;, {<br>    method: &#39;POST&#39;,<br>    headers: { &#39;Content-Type&#39;: &#39;application/json; charset=utf-8&#39; },<br>    body: JSON.stringify({<br>      host: new URL(SITE).host,<br>      key: KEY,<br>      keyLocation: `${SITE}/${KEY}.txt`,<br>      urlList: changed.map((c) =&gt; c.url),<br>    }),<br>  });<br>  if (!res.ok) throw new Error(`IndexNow answered ${res.status}`);<br>  for (const { url, lastmod } of changed) sent[url] = lastmod;<br>  writeFileSync(LOG, JSON.stringify(sent, null, 2) + &#39;\n&#39;);<br>}</pre><p>The request format is the one in the <a href="https://www.indexnow.org/documentation">IndexNow documentation</a>. The important part is the last two lines: the record of what was sent is written to a file you commit, so the next run, on any machine, knows what it already did.</p><h3>Check these before you blame Google</h3><p>If your pages aren’t being indexed, these take minutes and catch everything above:</p><ol><li><strong>Bing Webmaster Tools, IndexNow page.</strong> How many URLs a day, and from which source. If you didn’t expect a source, find it.</li><li><strong>Your CDN’s dashboard settings.</strong> In Cloudflare, Crawler Hints (Caching, Configuration) and every Redirect, Transform and Cache rule. Anything there is configuration your repo doesn’t show.</li><li><strong>Any build step that calls an outside service.</strong> Does it run on preview builds? Does its “changed” check keep state on the build machine?</li><li><strong>curl -sI on the URL</strong>, not a summarising tool, for the status code and headers of the first response.</li><li><strong>cf-cache-status or your CDN&#39;s equivalent</strong> before you believe anything about stale caches.</li><li><strong>The IP address behind any “Googlebot” in your logs</strong>, and Search Console’s Crawl stats for the real numbers.</li><li><strong>URL Inspection in Search Console.</strong> If it says the page was fetched, crawling and indexing are allowed and your canonical was accepted, stop looking for a technical fault. What’s left is whether Google thinks the page is worth indexing.</li></ol><p>The last one is the one I wish I’d accepted sooner.</p><p>If you’ve found a dashboard setting that switched itself back on, or a build step quietly calling an outside service on every preview, I’d like to hear about it in the comments. I’ll post an update on whether the indexing changes above work.</p><p><em>Originally published at </em><a href="https://stackarchitect.xyz/blog/indexnow-55000-urls-62-page-site/"><em>stackarchitect.xyz</em></a><em> by Luke Sandelands.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5e9795356e12" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Shopify Stocky’s export window can close on 29 November. Save these records before it does]]></title>
            <link>https://medium.com/@stackarchitect123/shopify-stockys-export-window-can-close-on-29-november-save-these-records-before-it-does-1f7124f07883?source=rss-fe981e26d7b8------2</link>
            <guid isPermaLink="false">https://medium.com/p/1f7124f07883</guid>
            <category><![CDATA[small-business]]></category>
            <category><![CDATA[retail]]></category>
            <category><![CDATA[shopify]]></category>
            <category><![CDATA[inventory-management]]></category>
            <category><![CDATA[ecommerce]]></category>
            <dc:creator><![CDATA[Stack Architect]]></dc:creator>
            <pubDate>Sat, 14 Nov 2026 19:28:19 GMT</pubDate>
            <atom:updated>2026-11-14T19:28:19.730Z</atom:updated>
            <content:encoded><![CDATA[<h3>Stocky’s read-only export access runs for at least 90 days after 31 August 2026, so it can close as early as 29 November. What to export, what you have to copy by hand, and how to check you have it.</h3><p>Shopify switched Stocky off on 31 August 2026. Its data didn’t disappear that day: Shopify says you keep read-only access to export it for at least 90 days after that date. It has not published an end date.</p><p>Ninety days from 31 August is <strong>29 November 2026</strong>. That’s the earliest the window can close, and it’s the only date Shopify has given, so treat it as the deadline. If your store used Stocky, here’s what to save, what you can’t download and have to copy by hand, and how to check you actually have it.</p><h3>What still works, and what doesn’t</h3><p>Stocky’s APIs stopped on 31 August with no grace period, so any tool wired into them stopped too. Nothing moved into Shopify Admin by itself. What’s left is read-only access to Stocky’s own reports.</p><p>Shopify’s migration guide is blunt about what replaces it: “You now need to manage inventory in the Shopify admin and Shopify POS.”</p><p>If you uninstalled Stocky after Shopify delisted it on 2 February 2026, you can’t reinstall it, and the export steps below won’t work for you. Skip to the last two sections.</p><h3>What you can export</h3><p>Shopify lists three kinds of Stocky record you can export through Stocky’s own reports. In your Shopify admin, go to <strong>Apps → Stocky → Reports</strong>.</p><ol><li><strong>Completed purchase orders, as CSV.</strong> Pick the purchase order report and set the date range to your full history. The default range cuts off older orders, and a partial archive looks complete right up until you need the missing year.</li><li><strong>Stocktake history</strong>, plus the inventory adjustment log beside it, from the same Reports menu with the same full date range. These are your records of what you counted and what you corrected. Nothing else in your stack holds them.</li><li><strong>Historical cost data.</strong> Before you archive it, check five SKUs against recent supplier invoices. A cost file you haven’t checked is worse than none, because you’ll trust it.</li></ol><h3>What you can’t export: copy it by hand</h3><p>Shopify states that suppliers can’t be exported from Stocky. There’s no hidden export and no API route. Reorder points and safety-stock settings don’t export either.</p><p>So while Stocky still opens, go through each supplier and copy these into a spreadsheet:</p><ul><li><strong>Supplier name and contact</strong>: Nobody else has the list in one place</li><li><strong>Lead time</strong>: Decides when you reorder</li><li><strong>Minimum order quantity</strong>: Decides how much you reorder</li><li><strong>Payment terms</strong>: Easy to forget, awkward to ask for again</li><li><strong>Reorder point and safety stock, per priority SKU</strong>: Judgement you built up over time, not data you can look up again</li></ul><p>This is the slow part. It’s also the only part nothing can recover for you after the window closes.</p><h3>Check that you actually have it</h3><ul><li>Put every file in one dated folder in cloud storage: outside Shopify, and not on a single laptop.</li><li>Open each CSV. The earliest date in the purchase order export should match when you started using Stocky.</li><li>Compare row counts with what you’d expect: roughly how many purchase orders and stocktakes you’ve done.</li></ul><p>An export you haven’t opened is an assumption.</p><h3>After the export: what replaces Stocky</h3><p>Shopify’s own answer is Shopify itself. Its migration guide lists transfers between locations, tracking and receiving purchase orders, adjusting quantities and reporting on inventory history as built into the admin, with Shopify Flow for low-stock alerts and Sidekick to draft purchase orders.</p><p>The gap many stores notice is the running record of what sold, in a sheet they can sort and share. If that’s what you miss, I publish a free <a href="https://stackarchitect.xyz/stocky-swap/">Make.com scenario that writes one Google Sheets row per order line item</a>: timestamp, SKU, quantity, order ID and product. It’s an order log, not stock management. It doesn’t see restocks, returns or manual adjustments. Since 11 November its Sheets step uses Make’s Retry error handler, so a failed write is retried and then kept for you to re-run, instead of being dropped without a word.</p><p>If you’re not sure how much of your history lives only in Stocky, the free <a href="https://stackarchitect.xyz/stocky-migration-risk-scorer/">Stocky migration risk scorer</a> asks about your purchase orders, suppliers and stocktakes and tells you what to move first.</p><h3>The short version</h3><ul><li><strong>Deadline:</strong> treat 29 November 2026 as the last day. Shopify has given a minimum, not a date.</li><li><strong>Export:</strong> purchase orders over your full history, stocktakes and adjustments, cost data.</li><li><strong>Copy by hand:</strong> suppliers, lead times, minimum order quantities, payment terms, reorder points.</li><li><strong>Check:</strong> open every file before you trust it.</li></ul><p>The full guide covers what Shopify Admin now does for free and what the paid replacements are worth: <a href="https://stackarchitect.xyz/stocky-alternative/">Replacing Shopify Stocky: data, deadline and alternatives</a>.</p><p><em>Originally published at </em><a href="https://stackarchitect.xyz/stocky-alternative/"><em>stackarchitect.xyz</em></a><em> by Luke Sandelands.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1f7124f07883" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I added Retry to every Make.com scenario I publish. Two of them still couldn’t retry anything.]]></title>
            <link>https://medium.com/@stackarchitect123/i-added-retry-to-every-make-com-scenario-i-publish-two-of-them-still-couldnt-retry-anything-47806b83da82?source=rss-fe981e26d7b8------2</link>
            <guid isPermaLink="false">https://medium.com/p/47806b83da82</guid>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[no-code]]></category>
            <category><![CDATA[shopify]]></category>
            <category><![CDATA[ecommerce]]></category>
            <dc:creator><![CDATA[Stack Architect]]></dc:creator>
            <pubDate>Sat, 14 Nov 2026 19:25:25 GMT</pubDate>
            <atom:updated>2026-11-14T19:25:25.844Z</atom:updated>
            <content:encoded><![CDATA[<h3>A Make.com error handler only fires if the step reports an error. One HTTP setting decides that, and in two of my public scenarios it was off.</h3><p>On 11 November I went through the free Make.com scenarios I publish for Shopify stores and replaced every <strong>Ignore</strong> error handler with <strong>Retry</strong>, and switched on <strong>Store incomplete executions</strong>. Failed calls would now be retried three times, fifteen minutes apart, then kept for re-running instead of being thrown away. Job done.</p><p>The next day I opened two of them again and found the fix did nothing for the failures that mattered most.</p><h3>The setting that decides whether an error exists</h3><p>Make’s HTTP module has an advanced setting called <strong>Return error if HTTP request fails</strong>. Make’s own description of it is short: “The module will treat 4xx and 5xx responses as errors.”</p><p>Read that the other way round. With the setting on <strong>No</strong>, a 400 or a 500 is not an error. The module receives the reply, the run carries on, and the scenario history shows a success. An error handler, any error handler, only runs when a module raises an error. So with this setting off, my new Retry handler had nothing to catch.</p><p>That’s exactly how an earlier failure stayed hidden. The Meta request in my CAPI Shield blueprint was written against a Graph API version that Meta retired in May 2026. From that day every request came back 400, and every run finished green, for about four months. Three settings hid it together: the HTTP module didn’t treat error replies as errors, the step’s handler was set to Ignore, and failed runs weren’t stored. I had fixed two of the three. On two scenarios, the Meta one and the TikTok one, the first was still off.</p><p>Both are now set to <strong>Yes</strong>. I found it by opening the module and reading the setting, not because anything told me.</p><h3>Why this one is easy to miss</h3><p>You won’t see it unless you go looking. It sits under <strong>Show advanced settings</strong> at the bottom of the HTTP module, a long scroll below the URL, headers and body you actually came to edit.</p><p>The error handler, by contrast, is right there on the canvas: an orange Retry circle hanging off the module. It looks finished. Nothing on the canvas tells you the module it’s attached to will never raise an error for it to handle.</p><h3>The five settings to check on every scenario that calls an outside API</h3><p>I now check these on every scenario before I publish or switch it on:</p><ol><li><strong>Every step that sends data out has an error handler, and it’s Retry, not Skip.</strong> Make renamed its handlers: <strong>Skip</strong> is what older guides call Ignore, and <strong>Retry</strong> is what they call Break. Skip discards the failed bundle and the run finishes as a success.</li><li><strong>Every HTTP module has Return error if HTTP request fails set to Yes.</strong> Without it, a 4xx or 5xx never reaches the handler.</li><li><strong>Scenario settings → Store incomplete executions is Yes.</strong> This is where Retry keeps a run that still fails after its attempts, so you can fix the cause and re-run it.</li><li><strong>A webhook scenario runs Immediately as data arrives.</strong> One of mine was set to run every 15 minutes, which delayed every order by up to a quarter of an hour.</li><li><strong>No module carries a Deprecated label.</strong> Make labels outdated modules on the canvas. When I opened my Google Ads scenario, its Upload a Click Conversion module said: “This module is deprecated and may not work. We recommend replacing it with an alternative.”</li></ol><h3>The gap even all five don’t close</h3><p>Some APIs report a rejected event with HTTP 200 and the error inside the reply. TikTok’s Events API can answer with a non-zero code in the body. To Make that&#39;s a 200, so it&#39;s a success, so it&#39;s never retried.</p><p>For those, check the reply, not the status. On a test run, click the HTTP step in the scenario’s history and read the reply: for TikTok, &quot;code&quot;: 0 means the event was accepted. If you want it enforced on every order, add a filter after the request that stops the run when code isn&#39;t 0, so the failure shows up instead of passing.</p><h3>The two-minute test that would have caught all of it</h3><p>Configuration you’ve read isn’t configuration you’ve tested. So test a failure on purpose:</p><ol><li>Clone the scenario, so the real one keeps running.</li><li>In the clone, change the access token to something wrong.</li><li>Click <strong>Run once</strong> and send it a sample order.</li><li>Open the run in <strong>History</strong>.</li></ol><p>If that run is a success, your error handling isn’t seeing errors, whatever the canvas looks like. With Retry working, Make marks the step as a warning and keeps the bundle for retry. Mine would have shown success on both scenarios until the day I changed the setting.</p><h3>Broken processes are loud. Automated ones aren’t.</h3><p>There’s a broader version of this that I gave in a research piece this week, about a Shopify profit report I automated before fixing the cost data underneath it. The automated report was “clean, confident” and wrong, and nobody noticed, because nothing about it looked wrong. My point there was that broken manual processes are loud: somebody complains, and that’s how they get fixed. “Automate one without fixing it and the complaints stop while the errors continue.”</p><p>An error handler that can never fire is the same thing in miniature. The scenario goes quiet, the history fills with green, and the one signal you had, a failed run, has been switched off without anyone deciding to.</p><p>So the job isn’t adding the handler. It’s proving the failure still reaches you.</p><h3>The short version</h3><ul><li>An error handler only runs if the module raises an error.</li><li>In Make’s HTTP module, <strong>Return error if HTTP request fails</strong> decides that. Set it to <strong>Yes</strong>.</li><li>Use <strong>Retry</strong>, not Skip, and set <strong>Store incomplete executions</strong> to <strong>Yes</strong>.</li><li>APIs that report errors inside a 200 need a filter on the reply.</li><li>Test a failure on purpose before you trust a green history.</li></ul><p>The full guide, with screenshots from my own Make account and the patterns for deduplicating orders and a daily count check that catches every kind of failure: <a href="https://stackarchitect.xyz/blog/the-complete-guide-to-reliable-shopify-automations/">Reliable Shopify automations: make every failure visible</a>.</p><p><em>Originally published at </em><a href="https://stackarchitect.xyz/blog/the-complete-guide-to-reliable-shopify-automations/"><em>stackarchitect.xyz</em></a><em> by Luke Sandelands.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=47806b83da82" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Before Black Friday, check these three Meta tracking settings on your Shopify store]]></title>
            <link>https://medium.com/@stackarchitect123/before-black-friday-check-these-three-meta-tracking-settings-on-your-shopify-store-209815f64e58?source=rss-fe981e26d7b8------2</link>
            <guid isPermaLink="false">https://medium.com/p/209815f64e58</guid>
            <category><![CDATA[shopify]]></category>
            <category><![CDATA[facebook-ads]]></category>
            <category><![CDATA[ecommerce]]></category>
            <dc:creator><![CDATA[Stack Architect]]></dc:creator>
            <pubDate>Wed, 11 Nov 2026 10:33:25 GMT</pubDate>
            <atom:updated>2026-11-11T10:33:25.846Z</atom:updated>
            <content:encoded><![CDATA[<h3>Shopify counts every order. Meta only counts what it can trace to an ad. Here is how to see the gap, close the part you can, and stop it from cutting campaigns that work.</h3><p>Shopify counts every order. Meta counts only the purchases it can trace back to an ad. When that trace breaks, Ads Manager reports a lower return on ad spend than you’re really getting, and Meta’s delivery system learns from fewer purchases than actually happened.</p><p>That’s an annoyance most of the year. In the weeks before Black Friday, when budgets are set from what Ads Manager says is working, it’s how a store ends up cutting the campaigns that are actually selling.</p><p>Here are three settings to check before your biggest week, and one number to use instead of Meta’s.</p><h3>First, measure your gap</h3><p>Take the last 30 complete days.</p><ol><li><strong>Shopify:</strong> Analytics → Reports → Sales over time. Note the order count.</li><li><strong>Meta Ads Manager:</strong> the Campaigns view, with the 7-day click, 1-day view attribution window. Note the purchases reported.</li><li><strong>The gap:</strong> (Shopify orders − Meta purchases) ÷ Shopify orders × 100.</li></ol><p>Not every Shopify order came from a Meta ad, so the gap is never zero, and part of it isn’t lost tracking at all. Nobody publishes a “normal” figure for it. What matters is the trend: a gap that widened when nothing else about your marketing changed is the signal.</p><h3>Setting 1: your Meta pixel’s data sharing</h3><p>On 13 January 2026, Shopify changed the default data sharing setting for app pixels from <strong>Always on</strong> to <strong>Optimized</strong>. Shopify’s changelog says it watches each marketing pixel, and when it sees zero signals from one over days or weeks, it pauses that pixel’s data sharing until signals return. A pixel that’s working keeps sharing. The risk is a pixel Shopify judges inactive, which then stops without telling you.</p><p><strong>The check:</strong> Settings → Customer events → App pixels → the Data column. Set the Facebook &amp; Instagram pixel to <strong>Always on</strong>.</p><h3>Setting 2: whether purchases also reach Meta from a server</h3><p>A browser pixel loses some purchases to things you can’t control. Since iOS 17, Apple’s Link Tracking Protection strips click identifiers from links opened in Mail, Messages and Safari Private Browsing, and Safari limits how long tracking cookies last. Ad blockers and shoppers declining tracking take more.</p><p>Meta’s Conversions API sends the purchase from a server instead, where none of that applies. For many Shopify stores it’s already one setting away: in the Facebook &amp; Instagram sales channel, Shopify’s help centre says the <strong>Enhanced</strong> and <strong>Maximum</strong> data sharing levels use Meta’s Conversions API as well as the pixel. <strong>Standard</strong> is the pixel only.</p><p><strong>The check:</strong> Sales channels → Facebook &amp; Instagram → Settings → Data sharing settings. If you’re on Standard, that’s the gap.</p><h3>Setting 3: that you’re not sending every purchase twice</h3><p>If you’ve also set up a webhook-based connection, through Make.com, a tag server or an app, check you aren’t running it <em>and</em> Shopify’s Conversions API for Meta. Two server routes can each send the same purchase.</p><p>Meta merges a browser event and a server event into one purchase only when the event name and the event ID match. If they don’t, each purchase can be counted twice, which flatters your numbers in exactly the week you most need them to be right.</p><p><strong>The check:</strong> pick one Meta server route. Then in Meta Events Manager, look at the Purchase event and confirm browser and server events are being deduplicated.</p><h3>The number to use instead: blended ROAS</h3><p>Meta’s reported ROAS is inflated by attribution overlap and deflated by tracking loss at the same time. Neither direction is reliable.</p><p><strong>Blended ROAS = total Shopify revenue ÷ total ad spend across every channel</strong>, for the same period and currency. If you spent 5,000 across all channels and Shopify recorded 25,000 in revenue, your blended ROAS is 5.</p><p>It uses Shopify’s own order data, which counts every sale whatever happened to the tracking. Calculate it before you pause anything this month. The most expensive mistake during a tracking problem is switching off a campaign that’s working because Meta can’t see its sales.</p><h3>What to expect after you fix it</h3><p>Your Shopify-to-Meta gap should narrow toward the part that was never Meta’s to begin with. Meta’s Event Match Quality score should reflect the customer details you now send; read your own score rather than chasing someone else’s target. Judge the result by comparing the same 30-day gap a month after the change.</p><p>The full guide covers all three causes, the six-step fix, the webhook route for stores that also advertise on TikTok, and how to check deduplication and match quality in Events Manager: <a href="https://stackarchitect.xyz/blog/shopify-meta-roas-dropped-2026-fix/">Shopify Meta ROAS dropped in 2026? Why, and the free fix</a>. To put your own figures in, there’s a free <a href="https://stackarchitect.xyz/shopify-vs-meta-attribution-gap-calculator/">Shopify vs Meta attribution gap calculator</a>.</p><p><em>Originally published at </em><a href="https://stackarchitect.xyz/blog/shopify-meta-roas-dropped-2026-fix/"><em>stackarchitect.xyz</em></a><em> by Luke Sandelands.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=209815f64e58" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[6 Reasons Your Meta Conversions API Events Don’t Match (and How to Check Yours)]]></title>
            <link>https://medium.com/@stackarchitect123/6-reasons-your-meta-conversions-api-events-dont-match-and-how-to-check-yours-2ce88c9e2c52?source=rss-fe981e26d7b8------2</link>
            <guid isPermaLink="false">https://medium.com/p/2ce88c9e2c52</guid>
            <category><![CDATA[web-development]]></category>
            <category><![CDATA[shopify]]></category>
            <category><![CDATA[api]]></category>
            <category><![CDATA[nodejs]]></category>
            <category><![CDATA[javascript]]></category>
            <dc:creator><![CDATA[Stack Architect]]></dc:creator>
            <pubDate>Tue, 29 Sep 2026 15:25:25 GMT</pubDate>
            <atom:updated>2026-09-29T15:25:25.469Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*QcNHtTz2NujrplSUxReNTA.jpeg" /><figcaption>6 Reasons Your Meta Conversion API Events Don’t Match and How to Check Yours</figcaption></figure><p>Server-side tracking is meant to fix the conversions a browser pixel loses. But a Conversions API event only helps if Meta can match it to a person, and a lot of setups send events that are accepted, counted as received, and then do much less than they should.</p><p>I build server-side tracking for Shopify stores, and the same six mistakes come up again and again. Every rule below comes from Meta’s own <a href="https://developers.facebook.com/documentation/ads-commerce/conversions-api/parameters">Conversions API parameter documentation</a>, read on 28 September 2026.</p><h3>1. event_time is in milliseconds</h3><p>Meta wants a Unix timestamp in <strong>seconds</strong>. JavaScript’s Date.now() returns <strong>milliseconds</strong>, so it&#39;s the single most common bug:</p><pre>event_time: Date.now()                    // wrong: 1727600000000<br>event_time: Math.floor(Date.now() / 1000) // right: 1727600000</pre><p>A quick check: a timestamp in seconds has 10 digits. Thirteen digits means milliseconds.</p><h3>2. Customer data is hashed before it’s formatted</h3><p>Meta requires email, phone, name, city, state, zip and country to be SHA-256 hashed. What catches people out is that Meta matches on the hash of the <strong>formatted</strong> value, and hashing doesn’t forgive formatting:</p><pre>sha256(&#39;Buyer@Example.com &#39;) !== sha256(&#39;buyer@example.com&#39;)</pre><p>Both are valid-looking 64-character hashes, so nothing flags the first one as wrong. Format first, then hash. Meta’s rules:</p><ul><li>em: Trim spaces, lowercase</li><li>ph: Digits only, no leading zeros, country code included</li><li>fn, ln: Lowercase, no punctuation</li><li>ct: Lowercase, no punctuation, special characters or spaces</li><li>st: Lowercase; US states as the 2-letter code</li><li>zp: Lowercase, no spaces or dashes; first 5 digits for US zips</li><li>country: Lowercase ISO 3166-1 alpha-2, such as gb</li></ul><p><strong>The UK phone trap.</strong> +44 (0)7700 900123 becomes 4407700900123 if you only strip symbols and leading zeros. The zero after the country code is a national prefix that isn&#39;t dialled internationally, so the number should be 447700900123. If your store sells in the UK, remove it before hashing.</p><h3>3. Fields that must stay raw get hashed too</h3><p>It’s tempting to hash everything “to be safe”. Meta says <strong>not</strong> to hash four fields:</p><ul><li>client_ip_address</li><li>client_user_agent</li><li>fbc (the click ID cookie)</li><li>fbp (the browser ID cookie)</li></ul><p>Meta uses these as they are. A hashed IP address is just a meaningless string.</p><h3>4. Website events are missing the page URL or user agent</h3><p>For events with action_source: &quot;website&quot;, Meta requires both event_source_url and user_data.client_user_agent. Server-side setups often drop them, because the server never sees the browser. The fix is to capture both in the browser at checkout and pass them to your server with the order.</p><p>action_source itself is required on every event, as is user_data.</p><h3>5. event_id is missing, or doesn&#39;t match the Pixel</h3><p>Meta marks event_id as optional, but recommends it for deduplication. If the browser Pixel and your server both send the same purchase, give them the same ID so Meta can treat them as one event:</p><pre>// Server<br>event_id: `order_${orderId}`</pre><pre>// Browser Pixel, same purchase<br>fbq(&#39;track&#39;, &#39;Purchase&#39;, { value, currency }, { eventID: `order_${orderId}` });</pre><p>Without a shared ID, you’re relying on Meta to work out that two events are the same purchase.</p><h3>6. Old events are retried in a batch</h3><p>If any event in a request is more than 7 days old, Meta returns an error for the <strong>whole request</strong> and processes none of its events. A retry queue that keeps resending one stale event can take good events down with it. Drop anything older than 7 days before it reaches the batch.</p><h3>A Purchase event that gets all six right</h3><pre>import { createHash } from &#39;node:crypto&#39;;</pre><pre>const sha256 = (v) =&gt; createHash(&#39;sha256&#39;).update(v).digest(&#39;hex&#39;);<br>const email = (v) =&gt; v.trim().toLowerCase();<br>const phone = (v) =&gt; v.replace(/[^0-9]/g, &#39;&#39;).replace(/^0+/, &#39;&#39;); // must already include the country code</pre><pre>const event = {<br>  event_name: &#39;Purchase&#39;,<br>  event_time: Math.floor(Date.now() / 1000),       // 1. seconds<br>  event_id: `order_${order.id}`,                    // 5. same ID as the Pixel<br>  action_source: &#39;website&#39;,<br>  event_source_url: pageUrl,                        // 4. captured in the browser<br>  user_data: {<br>    em: [sha256(email(order.email))],               // 2. formatted, then hashed<br>    ph: [sha256(phone(order.phone))],<br>    client_ip_address: clientIp,                    // 3. raw<br>    client_user_agent: userAgent,                   // 3. raw, and 4. required<br>    fbp,                                            // 3. raw<br>    fbc,                                            // 3. raw<br>  },<br>  custom_data: {<br>    value: Number(order.total),                     // required for Purchase<br>    currency: order.currency,                       // required for Purchase<br>  },<br>};</pre><h3>Check a payload in 10 seconds</h3><p>I turned these checks into two free tools:</p><ul><li><strong>In the browser:</strong> paste an event into the <a href="https://stackarchitect.xyz/meta-capi-payload-validator/">Meta Conversions API payload validator</a>. It runs locally, so nothing you paste is sent anywhere, and it has a formatter that applies Meta’s rules and hashes the result.</li><li><strong>In the terminal or CI:</strong> the open-source <a href="https://github.com/lsb11/shopify-capi-validator">shopify-capi-validator</a> package runs the same checks and exits with code 1 on any failure, so it can block a bad deploy.</li></ul><pre>npx shopify-capi-validator --payload ./event.json</pre><p>If you’re sending Shopify orders server-side for the first time, the <a href="https://stackarchitect.xyz/blog/shopify-server-side-tracking-complete-setup-guide/">server-side tracking setup guide</a> covers the whole flow, and the <a href="https://stackarchitect.xyz/meta-emq-score-estimator/">EMQ score estimator</a> shows what the customer data you send means for match quality.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2ce88c9e2c52" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Shopify Inventory in Google Sheets 2026 | Stack Architect]]></title>
            <link>https://medium.com/@stackarchitect123/shopify-inventory-in-google-sheets-2026-stack-architect-16e046d4ca8d?source=rss-fe981e26d7b8------2</link>
            <guid isPermaLink="false">https://medium.com/p/16e046d4ca8d</guid>
            <category><![CDATA[ecommerce]]></category>
            <category><![CDATA[shopify]]></category>
            <category><![CDATA[google-sheets]]></category>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[shopify-automation]]></category>
            <dc:creator><![CDATA[Stack Architect]]></dc:creator>
            <pubDate>Thu, 24 Sep 2026 00:00:54 GMT</pubDate>
            <atom:updated>2026-10-25T18:46:54.989Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*bDR9_kaIlJd1_4nw" /></figure><p>Most Shopify inventory management guides end with a recommendation to pay $29-$199/month for an app. This one doesn’t. Every system described here costs $0 and runs on tools you likely already have access to.</p><p>This guide covers the complete inventory management layer for Shopify stores, from basic order logging through supplier management, stock alerts, velocity analysis, and multi-location tracking. It also covers what to do now that <strong>Shopify Stocky has shut down. It stopped working on 31 August 2026.</strong></p><blockquote><strong><em>Don’t want to build the order log yourself?</em></strong><em> I’ve already built that part. It’s called Stocky Swap. It’s free, it runs on Make.com, and you can deploy it to your store in about 4 minutes. It logs every paid order to your sheet. The stock counts, reorder flags and alerts below are still yours to add.<br> 👉 </em><a href="https://stackarchitect.xyz/stocky-swap/"><em>Get Stocky Swap for free here</em></a><em>.</em></blockquote><h3>What Shopify’s Built-In Inventory Actually Covers</h3><p>Before building anything additional, understand what Shopify already provides:</p><p><strong>What Shopify tracks natively:</strong></p><ul><li>Current stock levels per variant per location</li><li>Stock deductions when orders are placed</li><li>Transfers between locations and adjustments to quantities</li><li>Purchase orders from suppliers, tracked and received in Shopify Admin</li><li>Low-stock alerts through Shopify Flow</li><li>Inventory history per product</li><li>CSV import and export for bulk updates</li><li>Multi-location inventory (on all plans except Starter)</li></ul><p>Shopify’s own migration guide for former Stocky users lists transfers, purchase orders, adjustments and Flow alerts as the native route. ( <a href="https://help.shopify.com/en/manual/products/inventory/transitioning-from-stocky">Shopify, “Transitioning from Stocky”</a>, read 10 September 2026.)</p><p><strong>What Shopify does not provide:</strong></p><ul><li>Automatic order logging to an external system you own</li><li>Real-time inventory ledger with running calculations</li><li>Sales velocity analysis (how fast products are selling)</li><li>Reorder point calculations based on lead time</li><li>Inventory valuation (COGS tracking per unit)</li><li>Automated low-stock notifications via Slack, email, or SMS</li></ul><p>The free Make.com and Google Sheets system fills part of that gap: the order ledger, velocity and reorder points. It does this without an app that could shut down, change pricing, or lose your historical data. It does not forecast demand, track bundles or handle barcode receiving.</p><h3>The Core System: Make.com + Google Sheets</h3><p>The foundation of free Shopify inventory management is a single Make.com scenario connected to a Google Sheets inventory ledger. Every time an order is paid in Shopify, Make.com captures the order data and appends a row to your sheet automatically.</p><p><strong>What this gives you:</strong></p><ul><li>A permanent, portable record of every order you own completely</li><li>Stock level calculations via spreadsheet formulas that you add (see Sheet 2 below)</li><li>Full order history that survives any app or platform change</li><li>The foundation for every advanced feature described below</li></ul><p><strong>Setup overview</strong> (full step-by-step at <a href="https://stackarchitect.xyz/stocky-swap/">Stocky Swap</a>):</p><ol><li>Create a free Make.com account at <a href="https://stackarchitect.xyz/make-com-shopify/">make.com</a> (free, 1,000 ops/month)</li><li>Create a new scenario with a Webhooks module as the trigger</li><li>Copy the webhook URL and add it to Shopify: <strong>Settings → Notifications → Webhooks → Order payment</strong></li><li>Add a Google Sheets module to append a row for each incoming order</li><li>Map the fields: order ID, date, product title, variant, SKU, quantity, price, fulfilment status, customer name</li><li>Activate the scenario</li></ol><p>From this point forward, every paid order in Shopify writes a new row to your sheet within seconds. That order log is what Stocky Swap does, and all it does. Stock counts, supplier tracking and alerts are the layers you build on top of it in the sections below. Stocky’s end of service was 31 August 2026.</p><h3>The Google Sheets Inventory Ledger Structure</h3><p>The sheet structure matters. A well-designed ledger supports every advanced feature below without rebuilding from scratch.</p><p><strong>Sheet 1: Orders Log (auto-populated by Make.com)Sheet 2: Product Inventory (manually maintained, formula-calculated)</strong></p><p>Column E calculates total units sold for each SKU by summing the Orders Log. Column F gives current stock, and Column H flags which products need reordering. Both update every time Make.com adds an order row.</p><p>Column F only knows about paid orders and the units you enter in Column D. Refunds, returns, damaged stock and manual adjustments in Shopify never reach the sheet, so the figure drifts from Shopify’s own count until you enter them by hand. Stocky Swap does not calculate stock on hand for you, and neither does this sheet on its own. Treat Shopify Admin as the count of record and reconcile against it.</p><p><strong>Sheet 3: Supplier Purchase Orders</strong></p><p>A simple table for tracking what you’ve ordered from suppliers, expected delivery dates, and quantities received. When stock arrives, manually update Column D (Units received) in Sheet 2: stock levels update automatically.</p><h3>Stock Alerts: Free Automated Notifications</h3><p>With the inventory ledger in place, you can trigger alerts when stock falls below reorder points without checking the sheet manually.</p><p><strong>Option 1: Email alert via Make.com (simplest)</strong></p><p>Add a second branch to your existing Make.com scenario: after logging the order to Google Sheets, check if the current stock level for that product is at or below the reorder point. If yes, send an email to your nominated address with the product name, current stock, and reorder point.</p><p><strong>Option 2: Scheduled daily check (more reliable)</strong></p><p>Create a separate Make.com scenario that runs every morning at 8am. It reads your Sheet 2, filters rows where Column H = “REORDER NOW”, and sends a consolidated email listing every product that needs reordering. This catches situations where stock dropped outside of order events (damaged goods, manual adjustments).</p><p><strong>Option 3: Google Sheets email alert (no Make.com needed)</strong></p><p>In Google Sheets, use <strong>Extensions → Apps Script</strong> to write a simple Apps Script function that sends an email when triggered. Set it to run daily via a time-based trigger. This works without Make.com if you prefer to keep everything in Google’s ecosystem, but be aware of the <a href="https://stackarchitect.xyz/blog/google-apps-script-quotas-explained-how-to-avoid-limits-and-scale-your-automations/">Google Apps Script quota limits</a> on consumer accounts.</p><h3>Sales Velocity Analysis</h3><p>Understanding how fast products sell tells you when to reorder before you run out, not after. With your Orders Log populated, add these calculations to Sheet 2:</p><p><strong>30-day velocity</strong> (units sold in last 30 days):</p><pre>=COUNTIFS(Orders!E:E, A2, Orders!B:B, &quot;&gt;=&quot;&amp;TODAY()-30)</pre><p><strong>Days of stock remaining</strong> (at current velocity):</p><pre>=IF(F2=0, &quot;OUT OF STOCK&quot;, IF(velocity=0, &quot;SLOW MOVER&quot;, ROUND(F2/velocity, 0)))</pre><p><strong>Recommended reorder date</strong> (based on lead time):</p><pre>=IF(days_remaining&lt;=lead_time, &quot;ORDER NOW&quot;, TODAY()+(days_remaining-lead_time))</pre><p>These formulas give you a live view of which products are running low based on actual sales rate, not arbitrary stock level thresholds.</p><h3>Multi-Location Inventory</h3><p>Shopify’s native multi-location tracking records stock per location, and for most stores it is the place to manage it. Stocky Swap’s ledger has no location column. The order payment webhook fires before an order is fulfilled, so an online order may not carry a fulfilment location at that point. Check a few real payloads in Make before you build on that field.</p><p>If your orders do carry a location, create a Sheet 4 with one column per location and SUMIF formulas that calculate sold quantities per SKU per location. Current stock per location = opening stock + received — sold, by location. The same caveat as Sheet 2 applies: returns and adjustments have to be entered by hand.</p><h3>After the Stocky Shutdown</h3><p>Shopify removed Stocky from the App Store on February 2, 2026. On August 31, 2026, Stocky stopped functioning entirely, and its APIs stopped the same day with no grace period. Purchase order history, stocktake records and inventory data held in Stocky are not deleted with it: Shopify says read-only access to export them continues for at least 90 days after that date, and it has published no deletion date. ( <a href="https://help.shopify.com/en/manual/products/inventory/transitioning-from-stocky">Shopify, “Transitioning from Stocky”</a>, read 10 September 2026.)</p><p><strong>If you used Stocky:</strong></p><ol><li>Export what you can now. Open Stocky in Shopify Admin and, if it still opens, export your purchase order history and stocktake records as CSV. Store the files in Google Drive. The read-only window has a floor of 90 days and no published end, so do not count on it lasting.</li><li>Move purchase orders, transfers and adjustments into Shopify Admin, which is where Shopify now handles them.</li><li>If you want your own copy of every order, deploy Stocky Swap. The <a href="https://stackarchitect.xyz/stocky-swap/">Stocky Swap setup</a> takes about 4 minutes. It logs forward from the moment you switch it on and cannot fill in orders placed before then.</li><li>Rebuild any supplier tracking you kept in Stocky using Sheet 3 above, or in Shopify’s purchase orders.</li></ol><p><strong>If you never used Stocky:</strong> Shopify Admin already tracks your stock. The Make.com and Google Sheets ledger is worth adding if you want an order record you own and the velocity and reorder views that Shopify does not calculate.</p><h3>Integrating Inventory with P&amp;L Reporting</h3><p>Once your inventory ledger is running, connecting it to profit and loss reporting costs nothing additional. It is a second branch on your existing Make.com scenario.</p><p>When each order arrives, Make.com writes to both your inventory sheet (quantity, fulfilment) and your P&amp;L sheet (revenue, an estimated processing fee, COGS from your own cost table, and gross profit). Ad spend and other overheads are not in it, so it is gross profit per order, not net profit. The <a href="https://stackarchitect.xyz/shopify-profit-loss-automation/">Shopify P&amp;L Automation</a> guide covers the full P&amp;L setup. If you deploy both together, you have an order ledger and a per-order profit sheet in one Google Sheets workbook.</p><h3>When to Consider a Paid Inventory App</h3><p>The free system described here handles most Shopify stores well. The genuine limitations appear at:</p><ul><li><strong>10+ locations</strong>: managing more than 3–4 locations via formula-based sheets becomes operationally complex</li><li><strong>1,000+ SKUs</strong>: SUMIF calculations across thousands of SKUs can slow sheet performance; a database-backed system handles this better</li><li><strong>Complex bundling</strong>: products that combine multiple SKUs (bundles, kits) require more sophisticated inventory logic than SUMIF can cleanly handle</li><li><strong>Barcode scanning for receiving</strong>: if your warehouse team scans items into stock, you need purpose-built receiving software</li></ul><p>Below these thresholds, Shopify Admin plus the Make.com and Google Sheets ledger covers what most small stores used Stocky for. It is not equivalent to a paid inventory app: it has no forecasting, and its stock figures need the manual reconciliation described above. It is free, owned by you, and exportable at any time.</p><h3>Deploying Your Free Inventory System Today</h3><p>The average store is paying anywhere from $29 to $500 a month for inventory management.</p><blockquote><strong><em>Don’t want to build the order log yourself?</em></strong><em> I’ve already built that part. It’s called Stocky Swap. It’s free, it runs on Make.com, and you can deploy it to your store in about 4 minutes. It logs every paid order to your sheet. The stock counts, reorder flags and alerts below are still yours to add.<br> 👉 </em><a href="https://stackarchitect.xyz/stocky-swap/"><em>Get Stocky Swap for free here</em></a><em>.</em></blockquote><ol><li><strong>Deploy Stocky Swap.</strong> The <a href="https://stackarchitect.xyz/stocky-swap/">setup</a> takes about 4 minutes and gets order logging running from the next paid order.</li><li><strong>Set up the sheet structure.</strong> Use the three-sheet layout above and copy the SUMIF formulas into Sheet 2.</li><li><strong>Add stock alert notifications.</strong> Build a second Make.com scenario or branch for email alerts on low stock, or use Shopify Flow.</li><li><strong>Export Stocky data</strong> if Stocky still opens in your admin. Its end of service was 31 August 2026. Shopify says read-only export access continues for at least 90 days after that date and has published no end date.</li><li><strong>Add P&amp;L tracking.</strong> A third branch on the same scenario, no additional setup cost.</li></ol><p>Every step above is free. The only ongoing cost is Make.com’s Core plan at $9/month if your order volume outgrows the free plan’s 1,000 credits a month.</p><h3>Get the Stocky Swap files</h3><p>The Complete Kit includes the Stocky Swap Make.com JSON blueprint. Import it and it logs one row per order line to your sheet from the next paid order. It does not record restocks, returns or adjustments; you add those as rows by hand. The kit also includes CAPI Shield (Meta purchase events), TikTok CAPI and P&amp;L Auto. $19.99 one-time.</p><p><a href="https://stackarchitect.xyz/pro/"><strong>Get the Complete Kit: $19.99 →</strong></a></p><p><em>Originally published at </em><a href="https://stackarchitect.xyz/blog/the-ultimate-guide-to-shopify-inventory-management/"><em>https://stackarchitect.xyz</em></a><em> on September 24, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=16e046d4ca8d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The four things that broke when I replaced $700/month of SaaS with a spreadsheet]]></title>
            <link>https://medium.com/@stackarchitect123/the-four-things-that-broke-when-i-replaced-700-month-of-saas-with-a-spreadsheet-c401c0770633?source=rss-fe981e26d7b8------2</link>
            <guid isPermaLink="false">https://medium.com/p/c401c0770633</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[javascript]]></category>
            <category><![CDATA[google-sheets]]></category>
            <category><![CDATA[saas]]></category>
            <category><![CDATA[automation]]></category>
            <dc:creator><![CDATA[Stack Architect]]></dc:creator>
            <pubDate>Mon, 14 Sep 2026 16:22:53 GMT</pubDate>
            <atom:updated>2026-10-25T18:49:18.322Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Stack Architect company logo: A green contrasting ziggernaut on a black background." src="https://cdn-images-1.medium.com/max/755/1*PO544JSFdbRLFh7ZkwnlYA.png" /></figure><p><em>Everything worked perfectly at ten events a day. At four hundred, it failed silently in four different ways.</em></p><p>A while back I looked at a stack of SaaS subscriptions that were, between them, doing something fairly unglamorous: moving rows of data from one system into another and doing arithmetic on them. The combined bill was around $700 a month.</p><p>So I rewrote the lot in Google Apps Script and a webhook layer.</p><p>It worked immediately, which turned out to be the problem. It worked at ten events a day and quietly stopped working at four hundred — in four separate ways, none of which produced an error message, an alert, or any symptom I noticed at the time.</p><p>That gap between “it works” and “it works at volume” is where the $700 actually lived. Here’s what I found in it.</p><h3>The first thing: I was counting the wrong resource</h3><p>My first version looked like every Apps Script tutorial you’ll find:</p><pre>events.forEach(e =&gt; {<br>  sheet.appendRow([e.id, e.timestamp, e.amount, e.source]);<br>});</pre><p>Fine for ten rows. Fatal for a thousand.</p><p>The mistake is conceptual rather than syntactic. I was thinking in rows, and Google charges in service calls. Four hundred appendRow calls are four hundred separate round trips to the Sheets backend, each with its own latency and each counting against a daily quota. You hit Service invoked too many times for one day long before you hit anything resembling real data volume.</p><p>The fix is to build the whole array in memory and write it once:</p><pre>const rows = events.map(e =&gt; [e.id, e.timestamp, e.amount, e.source]);<br>sheet.getRange(sheet.getLastRow() + 1, 1, rows.length, rows[0].length)<br>     .setValues(rows);</pre><p>Same output. One service call instead of four hundred. This single change took a script that died daily into one that hasn’t hit a quota in months.</p><p>The general rule I now apply without thinking: anything touching a Google service inside a loop is a bug waiting for volume.</p><h3>The second thing: webhooks lie about how many times they’ll call you</h3><p>Every webhook provider retries when your endpoint doesn’t respond quickly enough. Some retry on timeout even when your handler actually succeeded and just took too long to say so.</p><p>Which meant that while my endpoint was slow — because of the problem above — I was receiving the same event twice and writing it twice. Duplicate rows, inflated totals, no error anywhere.</p><p>The important realisation is that you cannot fix this at the transport layer. You don’t control how many times you get called. You only control what you do about it. So the deduplication has to live at the write layer:</p><pre>const cache = CacheService.getScriptCache();<br>if (cache.get(&#39;evt_&#39; + event.id)) return;   // seen it<br>cache.put(&#39;evt_&#39; + event.id, &#39;1&#39;, 21600);   // 6h TTL<br>writeEvent(event);</pre><p>CacheService rather than PropertiesService, because this expires on its own and you&#39;re not trying to maintain a permanent ledger — you&#39;re trying to survive a retry storm that lasts seconds.</p><p>One caveat worth internalising: the cache is best-effort and can evict early under memory pressure. It’s a retry guard, not a correctness guarantee. If duplicates are genuinely unacceptable, you also need a nightly job that reconciles against the sheet and dedupes on the ID column.</p><h3>The third thing: this one cost me weeks</h3><p>Two webhooks arrive 200 milliseconds apart. Both call getLastRow(). Both get 4,102. Both write to row 4,103.</p><p>One of them wins. The other event is simply gone.</p><p>No error. No log entry. No indication of any kind. I found it weeks later because a monthly total didn’t reconcile, and it took me another day to work out why.</p><p>The fix is a lock:</p><pre>const lock = LockService.getScriptLock();<br>try {<br>  lock.waitLock(30000);<br>  appendRows([event]);<br>} catch (e) {<br>  queueForRetry(event);      // don&#39;t silently drop it<br>} finally {<br>  lock.releaseLock();        // this line is not optional<br>}</pre><p>Use getScriptLock() rather than getUserLock() — the lock needs to span every execution of the script, not every execution by a given user. And note the finally block. An early return that skips releaseLock() will stall every subsequent execution for the full timeout, turning a race condition into a queue of them.</p><h3>The fourth thing: the six-minute wall</h3><p>Apps Script terminates any execution at six minutes. If you’re backfilling, migrating, or working through a queue that grew overnight, you will hit this. The execution dies mid-write, and there’s no transaction to roll back.</p><p>The answer is checkpointing: work until you’re near the limit, save your position, schedule a continuation, and exit cleanly.</p><pre>if (Date.now() - start &gt; MAX_RUNTIME_MS) {<br>  props.setProperty(&#39;cursor&#39;, String(cursor));   // save position<br>  scheduleContinuation();                        // pick up in 60s<br>  return;<br>}</pre><p>Two details that bite people. Leave more headroom than feels necessary — if each item takes eight seconds, a thirty-second buffer means dying mid-item. And time-based triggers are themselves a limited resource, capped at twenty per script, so delete each one after it fires or you’ll slowly accumulate orphans until trigger creation starts failing.</p><p>I wrote the fuller version of this pattern — batch sizing, resuming after a failed continuation, reconciling a partial run — <a href="https://stackarchitect.xyz/shopify-google-sheets-automation/">over here</a>.</p><h3>What the $700 was actually buying</h3><p>This is the part I’d want to tell myself at the start.</p><p>The SaaS tools I replaced were not charging $700 a month for the arithmetic. The arithmetic is trivial. They were charging for retry handling, deduplication, concurrency safety, and job continuation — four problems that are genuinely hard, entirely invisible at low volume, and which I had to rediscover one at a time by breaking things in production.</p><p>That doesn’t make the replacement a bad trade. It cost me maybe forty hours spread over a few months against roughly $8,400 a year, and I now own the thing outright and can read every line of it.</p><p>But it is a trade, not a free lunch. Anyone who tells you a spreadsheet and some webhooks “just replaces” a SaaS subscription has not yet run one at four hundred events a day.</p><p>If you’re building something similar: write these four patterns in <em>before</em> you have the volume that proves you needed them. The failures are silent, and by the time you notice, the missing data is already weeks old.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c401c0770633" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[When to Upgrade Free Make.com to Google Workspace]]></title>
            <link>https://medium.com/@stackarchitect123/when-to-upgrade-free-make-com-to-google-workspace-dcafea86927c?source=rss-fe981e26d7b8------2</link>
            <guid isPermaLink="false">https://medium.com/p/dcafea86927c</guid>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[shopify]]></category>
            <category><![CDATA[google]]></category>
            <category><![CDATA[apps]]></category>
            <category><![CDATA[google-workspace]]></category>
            <dc:creator><![CDATA[Stack Architect]]></dc:creator>
            <pubDate>Wed, 02 Sep 2026 00:00:52 GMT</pubDate>
            <atom:updated>2026-10-25T18:38:37.524Z</atom:updated>
            <content:encoded><![CDATA[<p>Upgrade from the free Shopify automation stack when you hit a real ceiling, not on principle. The three signals are Make.com credits exhausted before month end, Apps Script runs exceeding six minutes, and daily trigger runtime running out. What you buy is quota headroom, not features, so the decision follows the symptom rather than the price.</p><h3>Questions this guide answers</h3><h3>How many operations does Make.com give you for free?</h3><p>Make.com’s free plan gives up to 1,000 credits a month. How many orders that covers depends on your scenario: each module that runs typically uses one credit, so open one run in the scenario’s history, note its credit count, and divide 1,000 by it. A scenario using ten credits per order covers about 100 orders a month.</p><h3>What does upgrading to Google Workspace actually change?</h3><p>It raises Google Apps Script daily quotas, and that is the whole reason to do it for automation. The 6-minute per-execution ceiling is identical on both consumer and Workspace accounts: upgrading does not change it. What Workspace raises is the daily headroom: trigger runtime goes from 90 minutes to 6 hours, UrlFetch calls from 20,000 to 100,000, document creates from 250 to 1,500, and email recipients from 100 to 1,500. Workspace is priced per seat, so a second person is a second full seat rather than a marginal add-on.</p><h3>Do I need Google Workspace for Make.com to work?</h3><p>No. Make.com works with any Google account including free Gmail. Google Workspace is only needed when Google Apps Script quota limits are causing automation failures: typically scripts hitting the 6-minute consumer execution ceiling or the 90-minute daily trigger-runtime limit.</p><h3>Does upgrading to Google Workspace fix Apps Script timeouts?</h3><p>No. The 6-minute per-execution ceiling is identical on consumer and Workspace accounts, so ‘Exceeded maximum execution time’ will keep happening after you pay. The fixes are structural: batch reads and writes so each run does less work, split the job into chunks that each finish inside 6 minutes and save progress with PropertiesService between runs, or move the heavy work onto Make.com, which has no per-execution time limit. Upgrading raises daily quotas, not the per-run ceiling.</p><h3>When should I upgrade Make.com from free to paid?</h3><p>Upgrade Make.com when you consistently run out of the free plan’s 1,000 credits a month. Signs include scenarios stopping before the month ends, or switching off branches to stay inside the limit. Work out your own ceiling from the credits one order uses in your scenario history, rather than from a typical figure.</p><p><strong>The full guide, with every step, source and date:</strong> <a href="https://stackarchitect.xyz/blog/when-to-upgrade-free-make-google-workspace/">When to Upgrade Free Make.com to Google Workspace</a></p><p><em>Originally published on </em><a href="https://stackarchitect.xyz/"><em>Stack Architect</em></a><em> by Luke Sandelands. Figures there carry the date they were last checked; this copy does not, so prefer the original.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dcafea86927c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Shopify Agentic Storefronts — Eligibility & Limits 2026]]></title>
            <link>https://medium.com/@stackarchitect123/shopify-agentic-storefronts-eligibility-limits-2026-489d690f18ee?source=rss-fe981e26d7b8------2</link>
            <guid isPermaLink="false">https://medium.com/p/489d690f18ee</guid>
            <category><![CDATA[agentic-ai]]></category>
            <category><![CDATA[shopify]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[microsoft-copilot]]></category>
            <dc:creator><![CDATA[Stack Architect]]></dc:creator>
            <pubDate>Mon, 24 Aug 2026 20:07:20 GMT</pubDate>
            <atom:updated>2026-10-25T18:39:50.125Z</atom:updated>
            <content:encoded><![CDATA[<p>Shopify Agentic Storefronts connect your catalog to ChatGPT, Google AI Mode, Microsoft Copilot and Meta from one admin setting. Since 24 March 2026 the channels split: Google, Copilot and Meta complete checkout inside the AI surface, while ChatGPT redirects shoppers to your own Shopify checkout, which is why most limits apply to only some channels.</p><h3>Questions this guide answers</h3><h3>What are Shopify Agentic Storefronts?</h3><p>Shopify Agentic Storefronts connect your product catalog to AI assistants (ChatGPT, Google AI Mode and Gemini, Microsoft Copilot, and Meta surfaces such as Muse) through one setting in your Shopify admin, with no separate integration per platform. What each channel then does with a shopper differs: Google, Copilot and Meta complete the purchase inside their own interface, while ChatGPT sends the shopper to your own Shopify checkout. Announced in Shopify’s Winter ’26 Edition on 10 December 2025 and generally available to eligible merchants from 24 March 2026.</p><h3>Does ChatGPT still let customers buy without leaving the chat?</h3><p>No. OpenAI retired Instant Checkout on 24 March 2026, and Shopify’s documentation now describes ChatGPT as a discovery-focused referrer: shoppers complete the purchase on your own Shopify checkout, opened in the ChatGPT in-app browser or a new tab. This matters more than it sounds. Because the transaction happens on your checkout, the product-type restrictions, fulfilment restrictions and pixel restrictions that apply to direct checkout do not apply to ChatGPT traffic at all: your normal checkout rules and your client-side pixels work as they always have. Guides published before 24 March 2026 describe the opposite arrangement.</p><h3>Am I eligible for Shopify Agentic Storefronts?</h3><p>Eligibility is per channel, not one blanket rule, and it stacks two separate conditions: where your store is based, and where you sell. ChatGPT and Microsoft Copilot accept a store based in any country provided it sells to customers in the United States; Meta accepts one that sells to customers in the United States, Canada or Mexico. Google AI Mode and Gemini require the store itself to be based in the United States and to sell to customers there, and direct checkout renders only for US-based customers, with everyone else redirected to the store. So a UK-based store cannot use Google, but can use ChatGPT and Copilot if it sells to US customers, and Meta if it sells to customers in the US, Canada or Mexico. Google is still rolling out, and its product discovery runs on Google Merchant Center, which Shopify feeds through the Google &amp; YouTube sales channel. Agentic storefronts are active by default for eligible stores; the setting lives under Sales channels, then Agentic.</p><h3>How much does the Shopify Knowledge Base app cost?</h3><p>It is free. The Shopify Knowledge Base app is published by Shopify, launched on 16 May 2025, and its App Store listing shows no paid tier and no Plus-only pricing. Guides quoting a monthly fee for standard plans appear to be describing third-party knowledge-base apps instead. The app is not visible on your storefront: your FAQs serve as a data source that AI platforms read when answering questions about your products.</p><h3>Which products cannot be sold through Agentic Storefronts?</h3><p>Subscriptions, product bundles, customizable products and B2B-only products are not supported in agentic storefront direct checkouts. Local delivery, pickup in store and pickup points are not available as delivery methods, and custom checkout fields, checkout upsells, checkout validations, a subset of accelerated checkout options and requiring customer sign-in do not work either. Every one of those restrictions applies to the direct-checkout channels only: Google, Copilot and Meta. ChatGPT sends shoppers to your own checkout, where none of them apply.</p><h3>Do discount codes work in agentic checkout?</h3><p>Yes. Shopify’s documentation states that automatic discounts and discount codes you set up are both supported in agentic storefront direct checkouts. Several widely-shared guides claim the opposite; that claim does not match Shopify’s current documentation.</p><h3>Will Google Analytics and my Meta pixel track agentic orders?</h3><p>Not in direct checkout. Shopify states that Google Analytics and custom pixels will not fire there: only server-to-server pixels for the started and completed events. So an order placed inside Google AI Mode, Copilot or Meta reaches your Shopify admin but arrives in GA4 and Meta Ads Manager with no client-side event behind it. If you buy ads and measure return on ad spend from browser pixels, those orders look like they came from nowhere. Server-side tracking is the fix, and it is worth having in place before you enable the channels rather than after. ChatGPT is unaffected: that checkout is yours, so your pixels fire normally.</p><h3>Does Shopify charge a fee for selling through AI channels?</h3><p>No. Shopify’s documentation states there are no fees associated with selling in ChatGPT, in Google AI Mode and Gemini’s direct checkout, or in agentic storefront direct checkouts generally. You pay your standard payment processing fees and nothing further. Shopify also offers a separate free Agentic plan aimed at merchants who do not use Shopify as their primary storefront; it excludes the Online Store, gift cards, manual orders and Shop post-purchase offers.</p><h3>Is Perplexity a Shopify agentic channel?</h3><p>No, despite appearing in Shopify’s December 2025 launch announcement. In-chat buying on Perplexity runs on PayPal Instant Buy, those orders are managed through PayPal rather than your Shopify admin, and merchants apply to the Perplexity Merchant Program separately. Perplexity is no longer listed as an Agentic Storefronts channel in Shopify’s help documentation. Guides presenting it as a Shopify channel with redirect checkout are describing an arrangement that does not exist.</p><h3>My store is not eligible yet. Is any of this worth doing now?</h3><p>Yes, because discovery already happens without it. AI assistants crawl the web and read Shopify Catalog regardless of your agentic settings: what the feature adds is the checkout, not the visibility. Four things pay off immediately: install and populate the free Knowledge Base app, complete your product data so AI assistants can match products to conversational queries, add FAQPage structured data to your content, and get server-side tracking working so you have no attribution gap on the day the channels switch on.</p><p><strong>The full guide, with every step, source and date:</strong> <a href="https://stackarchitect.xyz/blog/shopify-agentic-storefronts-setup-guide-2026/">Shopify Agentic Storefronts — Eligibility &amp; Limits 2026</a></p><p><em>Originally published on </em><a href="https://stackarchitect.xyz/"><em>Stack Architect</em></a><em> by Luke Sandelands. Figures there carry the date they were last checked; this copy does not, so prefer the original.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=489d690f18ee" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Meta CAPI for Shopify — CAPI Shield, Free Server-Side Setup]]></title>
            <link>https://medium.com/@stackarchitect123/capi-shield-closing-the-conversion-tracking-gap-f1e4e422e85a?source=rss-fe981e26d7b8------2</link>
            <guid isPermaLink="false">https://medium.com/p/f1e4e422e85a</guid>
            <category><![CDATA[conversion-tracking]]></category>
            <category><![CDATA[meta]]></category>
            <category><![CDATA[ga4]]></category>
            <category><![CDATA[google]]></category>
            <category><![CDATA[serverside-tracking]]></category>
            <dc:creator><![CDATA[Stack Architect]]></dc:creator>
            <pubDate>Wed, 12 Aug 2026 18:32:56 GMT</pubDate>
            <atom:updated>2026-10-25T18:42:22.818Z</atom:updated>
            <content:encoded><![CDATA[<p>CAPI Shield is a free Make.com scenario that takes each paid Shopify order from a webhook and sends it to Meta’s Conversions API as a Purchase event, so iOS, Safari and ad blockers cannot stop it. This page is the setup. If your Meta results dropped and you do not yet know why, start with <a href="https://stackarchitect.xyz/blog/shopify-meta-roas-dropped-2026-fix/">the ROAS guide</a>.</p><h3>Questions this guide answers</h3><h3>What is the Shopify CAPI?</h3><p>The Shopify CAPI refers to server-side Conversions API implementation for Shopify: sending purchase events from Shopify’s servers to a platform’s server API, such as Meta’s Conversions API or TikTok’s Events API, instead of relying only on a browser pixel. Ad blockers and browser privacy limits cannot stop a server event from being sent. For Google Ads, the free route for most stores is enhanced conversions in Shopify’s Google &amp; YouTube app.</p><h3>What is Shopify server-side tracking?</h3><p>Server-side tracking sends Shopify conversion events directly from your server to ad platforms, bypassing the browser entirely. iOS restrictions, ad blockers, and cookie expiry cannot interfere. CAPI Shield implements this free via Make.com webhooks: the webhook fires from Shopify’s backend the moment an order is paid, completely independent of browser behaviour.</p><h3>How to set up Shopify server-side tracking free?</h3><p>Use CAPI Shield: a free Make.com scenario that routes Shopify order webhooks to Meta Conversions API. Sign up for Make.com free, create a scenario with a Webhooks trigger, add the Meta Conversions API request, create a Shopify Order Payment webhook pointing to your Make.com URL, and test with a sample order. no code required, free forever on Make.com’s free tier.</p><h3>What is Google Enhanced Conversions and how does it work with Shopify?</h3><p>Google Enhanced Conversions is Google’s server-side conversion tracking system. It supplements your standard Google Ads conversion tags by sending hashed first-party customer data (email, phone, name) alongside each conversion event. Google uses this to match conversions to Google accounts with greater accuracy, improving Smart Bidding and Performance Max optimisation even when cookies and pixels are blocked by iOS or ad blockers. CAPI Shield’s Google branch is not a working route to it as shipped: it uploads a click conversion keyed on a Google click ID, and Shopify’s order data has no field for one. Google is also moving this upload method, offline click conversion import, out of the Google Ads API and into its Data Manager API, and Make’s Google Ads module now warns that uploads may fail for developer tokens that have not recently sent click conversions. Its two consent fields must be set to match your own consent setup.</p><h3>What is Meta CAPI and how does it work with Shopify?</h3><p>Meta Conversions API (CAPI) is a server-to-server connection that sends Shopify purchase events directly to Meta without going through the browser. CAPI Shield sets this up free using Make.com, sending hashed email, phone, name, IP and user-agent with every event: the match-key set Meta scores Event Match Quality against, and a larger one than the browser pixel carries.</p><h3>How much conversion data am I losing without server-side tracking?</h3><p>We do not publish a number, because the honest answer is specific to your store: the loss is driven by how much of your traffic is mobile and iOS, and how many of those visitors decline tracking. Measure yours instead. The Shopify vs Meta Attribution Gap Calculator turns your Ads Manager conversion count and your Shopify order count into your own gap; the Meta EMQ Score Estimator scores the match quality of what you are currently sending; and the Shopify iOS Attribution Gap Benchmark publishes the sourced figures behind the mechanism. After you deploy CAPI Shield, place a test order and confirm the Purchase event arrives in the Test Events tab of Meta Events Manager.</p><h3>Does CAPI Shield replace my Meta Pixel?</h3><p>No. CAPI Shield works alongside your existing Meta Pixel, not instead of it. The pixel still captures upper-funnel events like ViewContent and AddToCart. CAPI Shield adds server-side Purchase events on top. Meta treats a pixel event and a server event as one purchase only when their event name and event ID match. CAPI Shield sends the order ID as the event ID, and Shopify does not document the ID its pixel sends, so check Test Events to see whether Meta is counting the pair once or twice.</p><h3>What is the difference between CAPI Shield and Elevar or Northbeam?</h3><p>Elevar’s Core plan is $225/month, and Northbeam lists its Starter plan at $1,500 with every plan quoted through a sales call. CAPI Shield does one part of what they do, for free: it sends each paid Shopify order to Meta’s Conversions API as a Purchase event. Elevar and Northbeam add attribution dashboards, more destinations and vendor support. If Purchase events reaching Meta is the only part you need, CAPI Shield covers it.</p><h3>Will CAPI Shield improve my ROAS?</h3><p>Possibly, and only indirectly. CAPI Shield sends Meta purchases the browser pixel missed, and Meta optimises delivery on the conversions it receives. We have not measured the effect on ROAS for any store, so we publish no figure or timescale: compare your own ROAS and Events Manager numbers for the weeks before and after.</p><h3>What is Event Match Quality and why does it matter?</h3><p>Event Match Quality (EMQ) is Meta’s 0–10 score for how well server-side events are matched to Meta user accounts. CAPI Shield sends hashed email, phone, name, IP and user-agent with every event. More match keys give Meta more to match on, which is what the score measures. We have not sampled EMQ across stores, so we do not publish a typical score: the EMQ estimator projects a range from your own parameter set.</p><h3>Is CAPI Shield GDPR compliant?</h3><p>The data flows through your own Make.com workspace. CAPI Shield hashes all personally identifiable information (email, phone, name) before transmission using SHA-256, as required by Meta’s Conversions API specification. Ensure your store’s privacy policy covers server-side tracking data sharing.</p><h3>How long does CAPI Shield setup take?</h3><p>We have not timed it on a fresh account, so we do not quote a time. The work is importing the Make.com scenario, pasting your Meta dataset ID and access token, adding one Shopify webhook, and placing a test order to check it in Meta’s Test Events. No developer, code or infrastructure is required.</p><h3>Why is Shopify’s built-in CAPI not enough?</h3><p>It may be enough. Shopify’s Facebook &amp; Instagram app sends Conversions API events at its Enhanced and Maximum data sharing levels, and for a store that only advertises on Meta that is the simplest free route. Shopify does not document exactly what triggers each server event. CAPI Shield is triggered by Shopify’s order webhook, which does not depend on the browser at all, and the same trigger can feed TikTok. Use one route for Meta, not both, unless you have confirmed in Events Manager that their event IDs match.</p><h3>Does this work with TikTok Ads as well?</h3><p>Yes. The same Make.com scenario can route events to Meta CAPI and TikTok Events API from one webhook. Most stores start with Meta, then add TikTok as a second branch if they advertise there: using the same webhook trigger, no additional Shopify setup required.</p><h3>Which Graph API version does CAPI Shield use?</h3><p>The Meta request names a Graph API version in its URL. Meta retires each version about two years after release, and requests to a retired version fail. Check the version in your scenario’s URL against Meta’s Graph API changelog and update it before it expires, then send one test order to confirm events still arrive.</p><h3>What happens if I want to remove CAPI Shield?</h3><p>Delete the Make.com scenario and the Shopify webhook. Nothing is installed on your Shopify store. Data stops flowing immediately. No residue, no lock-in, no cancellation fees.</p><p><strong>The full guide, with every step, source and date:</strong> <a href="https://stackarchitect.xyz/capi-shield/">Meta CAPI for Shopify — CAPI Shield, Free Server-Side Setup</a></p><p><em>Originally published on </em><a href="https://stackarchitect.xyz/"><em>Stack Architect</em></a><em> by Luke Sandelands. Figures there carry the date they were last checked; this copy does not, so prefer the original.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f1e4e422e85a" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>