The tracking page is your busiest page
More people open a tracking page than open anything else you run, usually on a phone, usually more than once. When it takes seven seconds, the call comes to your support team instead.
Every slow second is a phone call
A customer checks a shipment three times. If the page is slow they call, and the call costs more than the page ever did. Meanwhile the operation runs on integrations that were written once, by somebody who left, against a vendor system with no API.
Five signs this page is for you
Tick the ones you recognise. Nothing is sent anywhere and there is no score: the count is for you, not for us.
Read them again. If none of them land, this page is not for you, and the services page is the shorter route.
Our services, written for your work
The same ones we sell to everybody. What changes is which part of your job each one takes off your desk.
Web development
The tracking experience itself, held to a p95 rather than to a design review, because this page is measured by the calls it prevents.
Read the serviceAI transformation
The integration layer you own, sitting between the vendor system and everything else, and the case work that follows a rule taken off a person.
Read the servicePerformance marketing
Demand for the commercial side, measured to the account that ships rather than to the enquiry.
Read the serviceGrowth transformation
A managed platform team for the peak you cannot hire for, and a fractional CTO where the decisions outrun the seniority in the room.
Read the serviceThe terms that matter before the price does
Measured in p95, not in averages
An average hides the customer who waited nine seconds. The baseline and the target are both p95.
Never stop the line
Modernisation runs module by module. There is no cut over weekend where the operation is down.
The rollback is designed first
If reconciliation fails at hour six, the way back exists before the way forward is attempted.
Owned by your team
Integration code lives in your repository from the first commit.
What it looked like on a real engagement
A tracking portal that stopped queueing behind itself
Peak hour timeouts on a shipment tracker, caused by a vendor system nobody was allowed to touch.
- p95 response
- 6.8s710ms
- Throughput
- 1.0x3.9x
- Tickets
- 100%46%
Measured over 90 days post launch against the prior 90 days. Source: Datadog APM, April 2025
When we are the wrong call
- The vendor contract forbids integration and cannot be renegotiated
- There is no measurable baseline and nobody wants to establish one
- The deadline is a season away and the parallel run alone takes longer
Asked on the first call
Our core system has no API. Is that a blocker?
No, it is the normal starting condition. A vendor system with no API is one of the constraints this firm exists for, and there is a published engagement of exactly that shape.
How fast can a tracking page get?
On the last engagement of this shape, from 6.8 seconds to 710 milliseconds at p95, read from Datadog over ninety days. The window and the source are on the case.
Can you work during peak?
We would rather not cut anything over during peak, and we will say so. Stabilising before peak and modernising after it is the honest sequence.
Do you replace our WMS?
Rarely. Putting an integration layer you own in front of it is cheaper, faster and reversible.
One email a month, on what actually shipped
What we built that month, what it cost, and what we would do differently. No campaign, no offer, and one click to leave.
Bring the brief you are not sure you can deliver
Forty five minutes with the engineer who would run it. We will tell you on that call if the date is not reachable.