Ecommerce Replatforming: The Visibility Gap Nobody Plans For
.png)
Get the full episode here
Published by Noibu | The Ecommerce Toolbox: AI in Retail
Guest: Matt Small, Head of IT Service at Screwfix
Host: Kailin Noivo, Co-Founder at Noibu
🎧 Listen to the full conversation on Apple, Spotify, or YouTube
Ecommerce replatforming: the visibility gap nobody plans for
Ecommerce replatforming is the process of migrating your storefront to a new architecture — typically from a monolithic platform to a headless or composable stack of connected services. It almost always improves speed and flexibility. What teams rarely plan for is the side effect: as the architecture splits into more systems, the unified view of the customer journey splits with it, and errors, performance regressions, and checkout friction start slipping through the gaps.
That gap is exactly what Matt Small, who runs all "run services" at Screwfix, walked through on the latest episode of The Ecommerce Toolbox. Screwfix isn't a small operation — and the way they thought about visibility through a major replatform is a useful blueprint for anyone going composable.
Replatforming trades a known monolith for an unknown number of blind spots
A monolith is painful, but it's legible. One stack. One place things break. When you replatform to composable — stitching together best-of-breed services through integrations — you gain agility and lose that single seam.
Screwfix made exactly this move, shifting their websites across the UK, France, and Ireland off a single monolithic application onto linked SaaS services (commercetools). The flexibility was the point. But Matt was clear that visibility wasn't a nice-to-have alongside it — it was the thing that made the move safe.
"Moving away from the monolith on the web was really key. Visibility was actually really key for that as well."
— Matt Small, Head of IT Service, Screwfix
Here's why that matters for anyone evaluating a replatform: every new service is a new surface for a checkout error, a failed integration, or a Core Web Vitals regression — and none of it shows up in a single, tidy log anymore.
When your customers are your monitoring system, you're already too late
Before Noibu, Screwfix found out about problems the slowest way possible: a customer called in.
"Our connection to our customers was very disparate. We relied on customers phoning our call center and telling us about a problem — for us to then initiate a call and start looking from the customer's perspective. We had very little visibility about what the customer was actually seeing."
— Matt Small, Head of IT Service, Screwfix
This is the reactive trap, and it's nearly universal in ecommerce. The errors that hurt you most are the ones customers don't report. They don't open a ticket. They abandon the cart and they're gone. By the time enough people complain to register as a "problem," you've already lost the revenue.
The goal for any modern ecommerce team — and especially one mid-replatform — is to identify errors before customers or internal teams flag them. That's the shift from reactionary to proactive: instead of always chasing, you see the issue, scope its impact, and fix it on your own timeline.
Logs tell you what broke. They don't tell you what matters.
Screwfix already had monitoring on the back end — Datadog for logs and APM. That's the right tool for infrastructure, and Noibu integrates with it rather than replacing it. But back-end logs answer the wrong question for a revenue owner. They tell you what errored, not which errors are costing you conversions.
Matt put the problem bluntly:
A thousand 500s with no prioritization is noise. The "two or three that have huge impact" are revenue. Noibu's job is to do that ranking for you — grouping issues with AI and ordering them by conversion and revenue impact, with full session context attached, so the team triages what matters instead of what's loudest.
This is also where the back-end and customer-side layers complement each other. Datadog watches the infrastructure. Noibu watches the customer journey and tells you which technical issues are actually blocking conversions. Through a replatform, you want both — and you want them to agree on what's real.
Composable doesn't have to mean fragmented visibility
The fear with composable is that visibility shatters into as many dashboards as you have services. It doesn't have to. The point of a customer-side platform is that it sits over the top of the stack and reassembles the one thing your shoppers actually experience: the journey.
That means watching the funnel end to end — from the first click on a PLP to a completed checkout — regardless of how many services power it underneath. Checkout visibility is the part legacy and infra-only tools most often miss, and it's the part where a single broken integration quietly costs the most.
Here's the practical contrast a replatform forces you to confront:
That last row matters more during a replatform than at any other time. Replatforming is a release marathon — service after service shipping, any of which can break conversion. Release Monitoring connects every deployment to changes in stability, performance, and behaviour, so a regression gets caught at deploy, not two weeks later when the numbers dip.
Visibility shouldn't stop at the website
The most striking part of the Screwfix story is that they didn't stop at web. As an omnichannel retailer, they extended the same technology onto their in-store point-of-sale tills — so the platform now watches what store colleagues experience, not just online shoppers.
The mechanism is the same one that kills support guesswork online. When a till transaction fails, the old process was a colleague calling the help desk — and the help desk asking the unanswerable question:
"Can you remember what you did?" "No. I pressed some buttons and some stuff happened. Do you remember what the error was? Not really."
— Matt Small, Head of IT Service, Screwfix
You can't reproduce the problem from that. With session replay and sHelpCode™s, a colleague flags an issue and the team sees the exact session immediately — no reconstruction, no logging in as the user to recreate it. Matt's verdict on what that surfaced:
The real consolidation play: one tool over the top
Cost discipline is the through-line of every ecommerce conversation in 2026, and Screwfix is no exception. The instinct of the last few years — buy a specialized tool for every job — has flipped. The problem teams now describe is too many tools, not enough visibility, with each one calculating things differently.
Consolidation isn't just a line-item win. It's an accuracy win: when product, engineering, UX, and support all read the same view, they stop arguing about whose numbers are right and start fixing the thing.
There's a usability dimension too, and it's where AI earns its place. Matt — who describes himself as "not actually that technical" — wants issues surfaced in a way a non-engineer can understand and route, so engineers can go deep on the parts that need them. AI-powered grouping and prioritization does that: it translates a wall of errors into "here's what's costing you, here's the session, here's where it is in the journey." Decision-makers get clarity; engineers get the detail.
A visibility checklist for your ecommerce replatform
If you're migrating to headless or composable, put these in place before you flip the switch — not after the first revenue dip:
- Establish a customer-side baseline pre-migration. Capture current error rates, Core Web Vitals (LCP, INP, CLS), and funnel conversion so you can tell a regression from noise.
- Get checkout under 100% session capture. Sampling isn't good enough for checkout — the one place a single broken integration costs the most needs full coverage, not a sample.
- Prioritize by revenue, not volume. A thousand errors is a list. The two or three blocking conversion are the work. Make sure your tooling ranks for you.
- Tie every release to outcomes. A replatform ships constantly; connect each deploy to stability, performance, and behaviour so regressions surface at deploy time.
- Plan for reproduction, not reconstruction. Session replay means you watch what happened instead of trying to recreate it from logs.
- Extend the view to every channel that matters. If you're omnichannel, the in-store experience deserves the same visibility as the website.
Replatforming is the right move for plenty of retailers. Losing sight of your customer in the process isn't part of the deal — it's just the part nobody budgets for. The teams that come through a migration clean are the ones who can answer "what are shoppers actually experiencing right now?" on day one, not after the first dip.
If you're planning, mid-flight, or just live on a new stack, the fastest way to see your blind spots is to look for them.
Free website audit — scan your site for the errors, performance issues, and checkout friction your current tools aren't surfacing. → noibu.com/free-website-audit
Listen to the full episode now
.png)
.avif)
