Three productised services with a fixed scope and a fixed price. Each one runs on its own site, so you can read the whole argument before you talk to us.
25 findings from real accounts. Client names are never published, so the stores are described rather than named: the figures are exactly as we found them.
"Meta was receiving 114 of 189 orders. From the first full day after our fix it received 219 of 219, with the match score up from 6.1 to 8.2."
Every order reaches MetaMeta Conversions API · D2C eyewear brand
Every order reaches Meta
94% of orders invisible
32,887 junk conversions a day
675 dead URLs
56 feed problems
100% and still wrong
Every order reaches Meta
94% of orders invisible
32,887 junk conversions a day
675 dead URLs
56 feed problems
100% and still wrong
Every order reaches Meta
94% of orders invisible
32,887 junk conversions a day
675 dead URLs
56 feed problems
100% and still wrong
Every order reaches Meta
94% of orders invisible
32,887 junk conversions a day
675 dead URLs
56 feed problems
100% and still wrong
0.0% eligible
30 down to 2
22.9 GB reclaimed
Crashing every 40 min
281x slower
23 days of dead cron
0.0% eligible
30 down to 2
22.9 GB reclaimed
Crashing every 40 min
281x slower
23 days of dead cron
0.0% eligible
30 down to 2
22.9 GB reclaimed
Crashing every 40 min
281x slower
23 days of dead cron
0.0% eligible
30 down to 2
22.9 GB reclaimed
Crashing every 40 min
281x slower
23 days of dead cron
4.5x slower from a CDN
Every visitor blocked
Barcodes full of junk
95 of 97 pages empty
618 offers, 18 real issues
46.8% was an artefact
4.5x slower from a CDN
Every visitor blocked
Barcodes full of junk
95 of 97 pages empty
618 offers, 18 real issues
46.8% was an artefact
4.5x slower from a CDN
Every visitor blocked
Barcodes full of junk
95 of 97 pages empty
618 offers, 18 real issues
46.8% was an artefact
4.5x slower from a CDN
Every visitor blocked
Barcodes full of junk
95 of 97 pages empty
618 offers, 18 real issues
46.8% was an artefact
3,222 SKUs
A whole dashboard inert
8 things we refused
Two agencies, same bug
733 products rebuilt
99.3% of dead settings gone
3,222 SKUs
A whole dashboard inert
8 things we refused
Two agencies, same bug
733 products rebuilt
99.3% of dead settings gone
3,222 SKUs
A whole dashboard inert
8 things we refused
Two agencies, same bug
733 products rebuilt
99.3% of dead settings gone
3,222 SKUs
A whole dashboard inert
8 things we refused
Two agencies, same bug
733 products rebuilt
99.3% of dead settings gone
The findings, drawn to scale
Five numbers from real accounts. Three are things we moved, two are what we found before anyone had moved anything: the distinction matters, so we keep it. Client names are never published, so each is described rather than named, and we can show the working on any of them.
Meta Conversions API
Shopify orders reaching Meta, in percent
An Indian D2C eyewear brand
60% → 100%
Purchases reached Meta only from the thank you page, so cash on delivery buyers who left early were lost and the server events were in test mode. Now every order goes to Meta from the store, once. On the first full day Meta received 219 of 219 orders, and its match score for Purchase rose from 6.1 to 8.2.
Server-side tracking on Stape
Orders that never passed through the browser
An Indian ethnic womenswear label
94%
A third-party checkout took payment off the store's pages, so no page tag could count the sale. We set up server-side Tag Manager on Stape and sent each order's click id to Google server to server, and over the first three clean days Google recorded 105% of the value of Google-clicked orders. The same account had an auto-created action logging up to 32,887 junk conversions a day.
Google Merchant Center
Product disapprovals, before and after
A multi-brand camera store
30 → 2
Fixed product categories and description wording. No appeals were filed and not one page on the website changed.
WordPress SEO
Product pages with no SEO title or description
A saree and handloom label
95 of 97
Almost the entire catalogue was invisible to search. Two pages had been written in-house and became the exemplars for the rest.
WooCommerce → Shopify
Search clicks landing on dead product URLs
A camera and photo retailer
67%
A platform migration left 675 product URLs unredirected. They were still earning two thirds of the site's search clicks.
Shopify
Products shipped with no product type set
A 2,700-product multi-brand catalogue
2,686 of 2,686
Shopify's own coverage query reported all 2,686 as healthy: the presence filter it uses counts an empty string as a value. Four hand-checked samples disproved it.
Object cache
Cache backend response time against memory
A WooCommerce store on managed hosting
281x
Pages were instant when cached and forty seconds when not. The plugin's own connection test reported the backend healthy throughout.
Cloudflare
Page load with the CDN proxy on and off
An India-hosted business site
4.5x
A free-plan CDN was serving an India-hosted site out of Singapore. Turning the proxy off made the site four and a half times faster.
WordPress database
Settings rows loaded on every page view
A camera and pro video retailer
99.3%
Old settings, expired transients and licence data from plugins removed long ago. Exported with a rollback key and removed in batches while the site stayed under load.
Built in house
We ship our own Shopify apps
These apps come from the same team that fixes client stores, which is why the fixes are practical. We build tools for the problems we keep meeting, instead of only configuring other people’s.
Awaiting Shopify review
Flying Star GST Capture
Collect the buyer's GSTIN in the cart and keep it on the order.
Works on every Shopify plan, including Basic
Checks format, state and check digit as they type
Tags the order GST B2B and exports a GSTR-1 B2B CSV