Ask a mid-market commerce team what they monitor, and you will get a confident answer. Platform uptime. Payment gateway availability. Maybe page load time. Maybe Akamai or Cloudflare edge health.
Ask them what happens when the loyalty point accrual to their loyalty vendor stops syncing, and you will get silence. Not because they do not care. Because nobody set up an alert for it. Nobody even defined what “stopped syncing” looks like.
This is the observability gap that costs mid-market merchants real money. Not the loud failures. The quiet ones.
Loud failures vs. quiet failures
Some integrations fail loudly. If the payment processor goes down, customers cannot check out. You know immediately. If the platform itself is unreachable, the homepage is blank. Your phone rings.
These integrations get monitoring because the failure mode is impossible to miss. The alert is almost redundant. The failure is the alert.
Other integrations fail quietly. The loyalty platform stops receiving point accruals, but the checkout still works. Customers earn nothing for two weeks. They do not call. They just do not come back as often. The CDP event pipeline backs up, and personalization starts serving segments that are three days stale. Conversion dips 4%. Nobody connects the dip to the integration because the site is “up” and the dashboard shows green.
The tax calculation service starts returning fallback rates instead of real-time calculations because the API key expired and the error handling was written to degrade silently. You undercollect sales tax in three states for a month. The liability accrues invisibly until the finance team reconciles quarterly.
The inventory feed from the warehouse management system lags by six hours instead of the expected 15 minutes. Customers order items that are out of stock. Cancellation emails go out. Customer service absorbs the cost. Nobody traces it back to the feed lag because the feed is “running.”
Every one of these is a real pattern. I have seen each one multiple times.
Why the gap exists
Three reasons, all structural.
First, monitoring is usually configured during the build, by the build team. The build team’s mental model is “does it work?” Their definition of monitoring is: did the integration succeed or fail? Binary. Up or down. This catches loud failures. It does not catch degradation, staleness, drift, or silent fallbacks.
Second, the systems that fail quietly are usually owned by nobody. The platform is owned by the ecommerce team. The payment processor is owned by finance. But the CDP integration? The loyalty sync? The inventory feed? These live in a no-man’s land between ecommerce, marketing, operations, and IT. When nobody owns it, nobody monitors it.
Third, silent failures do not produce incidents. They produce slow, diffuse business impact that shows up in monthly metrics as a vague decline. Conversion down 3%. Repeat purchase rate softening. AOV flat when it should be growing. These get attributed to “market conditions” or “creative fatigue” or “seasonality.” The integration failure that caused them never gets identified because it never triggered a page.
What to monitor instead
The fix is to build your observability around a different question. Not “is it up?” but “is it current?”
For every integration in your commerce constellation, define three things:
Freshness SLA. How stale can the data be before it costs you money? Loyalty accruals need to sync within minutes, or customers see zero points at order confirmation and call support. Inventory needs to sync within the cycle time of your fastest-selling SKU, or you oversell. CDP segments need to refresh before the next personalization decision fires. Define the number. Write it down.
Silent failure signature. What does this integration look like when it is technically running but functionally broken? The API returns 200 OK but with empty payloads. The sync job executes on schedule but processes zero records. The event pipeline is flowing but dropping 15% of events due to schema validation errors that get swallowed. Each integration has a specific failure signature. You need to know what it is.
Business impact estimate. What does one day of silent failure cost? One week? One month? This does not need to be precise. It needs to be directional. If your loyalty program drives 30% of repeat purchases and the sync fails for two weeks, you can estimate the revenue impact. If your CDP serves personalized recommendations that lift conversion by 8% and the segments go stale, you can estimate the daily cost of serving generic content.
Once you have these three things for each integration, rank them. Not by technical complexity. Not by uptime history. By the cost of a silent failure multiplied by the expected time to detection.
I guarantee the ranking will surprise you. The integration you monitor most carefully will be near the bottom of the risk list. The one you have never alerted on will be near the top.
A practical starting point
You do not need a massive observability platform to close this gap. You need a simple audit.
List every integration in your commerce stack. For each one, answer two questions: What alert fires if this integration silently fails? How many hours or days would pass before a human notices?
If the answer to the first question is “no alert” and the answer to the second is “we would not know until someone complained,” you have found your highest-risk integration.
Build the alert. Define the freshness SLA. Assign an owner.
The loud failures take care of themselves. The quiet ones take care of your margin, slowly, while you are looking at an uptime dashboard that says 99.9%.