Skip to main content

v2026.08 — August 2026

· 7 min read
Product team
Product team
@product-team

The August 2026 release — a full Tax & Compliance engine, affiliate and referral operations platforms, a rebuilt SEO hub with eleven new documented surfaces, a dedicated social hub, enterprise SSO and SCIM, and a set of reliability fixes for problems that were making dashboards confidently wrong.

New features​

  • Tax & Compliance (Pro+). An independent auditor of the tax Shopify collected at checkout: recomputation across six regimes (India GST, EU VAT, UK VAT, US sales tax, Canada, APAC GST), a reconciliation worklist with cause clustering and materiality policy, registrations with real format validation, economic-nexus monitoring, and statutory filing exports in Excel, PDF, CSV, JSON and portal formats. It never sets checkout tax. See Tax & Compliance.
  • Direct GSTN e-filing (Pro+, opt-in beta). File GSTR-1 and GSTR-3B without leaving the app, through a licensed GST Suvidha Provider — maker, checker and authorised signatory as three distinct roles, the statutory per-head input-tax credit set-off, PMT-06 challans, EVC or DSC signing, and crash recovery that reconciles against the portal rather than blind-retrying. See Direct GSTN e-filing.
  • Affiliate operations (Starter+). A partner CRM with maker-checker approval, click-and-coupon attribution that credits a partner even when the code is not used, an idempotent commission ledger with refund clawback, payout batches gated on separation of duties, a five-detector fraud queue, tax and FTC compliance, a creative library and AI-assisted recruitment. See Affiliate program.
  • Referral growth platform (Starter+). Customer referrals as a governed programme: launch approval, kill switch, rolling budget, copy reconciled against the configured reward, invite campaigns to your existing base, and copy A/B tests measured against a real exposure denominator. See Referral program.
  • Social hub (Starter to generate, Pro to publish). Nine views — compose, calendar, queue, analytics, approvals, A/B tests, hashtags, connections and settings — replacing the split between a generate-only tab and a publish-only page. Engagement analytics now tie back to revenue. See Social publishing.
  • Enterprise SSO & SCIM (Enterprise). OIDC login with group-to-role mapping and just-in-time provisioning, plus SCIM 2.0 user and group provisioning from Okta, Microsoft Entra and Google Workspace. RBAC audit rows are HMAC-signed for tamper evidence. See SSO & SCIM.
  • Scoped API keys (Agency+). Public API keys now carry scopes, so a reporting integration can read usage without being able to spend the merchant's AI budget. Existing keys keep working unchanged. See API keys & scopes.
  • Pixel ops console (Starter+). The storefront event pipeline is now observable: installed versus delivering separated, priority-aware back-pressure that never sheds money events, a replayable failure queue, and anomaly detection for forged traffic. See Pixel & event pipeline.

SEO suite​

The SEO hub is now documented as the six working groups it actually is, with eleven surfaces getting their own pages for the first time:

Fixes​

Most of this release's fixes share one theme: a surface that reported a confident number when it should have reported that it did not know.

  • Sitemap: auto-regeneration ran every 48 hours, not daily. A freshness window equal to the cron interval always ratchets to double it. Fixed with a grace period.
  • Sitemap: expiry deleted the live file. Housekeeping removed a store's only served sitemap seven days after generation, making the public URL 404 for Googlebot. Expiry now means "stale, rebuild" and only supersession deletes.
  • Technical: every indexability bucket was pinned at exactly 100. Counts were read from row arrays trimmed for display. A 504-page canonicalised bucket rendered "100", and the share bar normalised against a fabricated denominator.
  • Technical: every crawl warning was computed and discarded. Wrong-host crawl, stale crawl and no-Search-Console-run warnings all existed and never reached the page — which is why "the dashboard is wrong and nothing says why".
  • Risk register reported an all-clear over a real backlog. It queried lazily-created work items rather than source rows, so a store with thousands of open crawl issues and broken links read as fully resolved.
  • Schema: a failed drift sweep rendered as a green "all in sync". A failed run persists a zero count as its initial value; the card read the count without the status. Failed runs now have their own state.
  • Schema: disabling or deleting markup left it live for Google. The row changed; the storefront metafield did not.
  • Outreach: link-rot monitoring had never run once. An invalid filter on a non-nullable column threw, and the error was swallowed — so "links lost" was permanently zero.
  • Scheduled crawls had never fired. Neither scheduling path had a caller. One store had an active weekly schedule 29 days overdue with a null last-run. Both systems are now swept hourly.
  • Queue monitoring covered 52 of 98 queues. The daily drift warning meant 46 queues — including the dead-letter queue, campaign sending, all ten product-pipeline stages and all four tax queues — had no depth alert and no metrics. A test now fails the build on drift.
  • Webhook delivery. Deliveries are now sent to the sub-path /webhooks/events. The reverse proxy's location /webhooks/ block has a trailing slash, so the bare /webhooks path received an implicit redirect — which Shopify does not follow, and after which it prunes the subscription. See Shopify webhooks.
  • Competitor registry was never auto-seeded, so stores that had been tracking competitors for months saw "no competitors tracked" and nothing was ever recrawled. Opening the registry now backfills from existing tracking surfaces.
  • Competitor crawls stopped at page one on any site that redirects apex to www. The whole comparison was computed from a single page and presented as fact.
  • Rank tracker: the "Rank Keywords" figure read a table with no writer, so it was zero for every store since launch. Alert configuration, CTR curve monotonicity and the Enterprise keyword cap were fixed in the same pass.

Scope change​

read_themes has been added so theme publishes and live-theme edits can be correlated with SEO crawl regressions. Existing installs must re-authorise before those webhooks deliver. Until they do, the regression incident drawer reports theme changes as an unmonitored blind spot rather than implying the site was untouched. See Authentication & scopes.