What it is
From January to June 2026 I worked on Dynatrace observability for Helloworld Travel’s estate: about 100 hosts and 11,000+ monitored entities across two production platforms. Reported
What it does
- A configuration audit. In my first week it cut the annual Dynatrace cost by 16 %. Measured
- An assessment for the CTO. I assessed the full estate’s observability, with eight prioritised findings and a 0–180-day remediation plan. Reported
- Incident analysis. With the development and operations teams I investigated incidents through distributed tracing. I traced about 17,000 recurring booking-integration errors to their root cause, and a SiteMinder booking-delivery failure. Reported
- Monitoring as configuration. I automated monitoring configuration through the Dynatrace API and Settings 2.0 and integrated it with CI/CD, so coverage scales with the estate.
How it is built
I set up monitoring for applications, services and infrastructure, configured dashboards, alerts, custom metrics and SLOs, and instrumented code for tracing and performance. I analysed metrics, logs and traces with Davis AI to find bottlenecks and anomalies. The instrument below compares the annual cost before and after the audit, in relative units.
Stack. Dynatrace, Davis AI, Dynatrace API, Settings 2.0
Where it stands
My role at Helloworld ended in June 2026. Of this page’s figures, only the 16 % is shown, in the instrument below. Measured
My part
I ran the configuration audit and wrote the assessment, as Technical Business Analyst and Product Engineer.
Measured in my first week
Dated update. This instrument carries method until employer-approved figures can be shown.
Claim
In my first week at Helloworld I reduced the annual Dynatrace platform cost by 16 %. Measured
The figure below is in relative units: the bar starts at 100 and lands at 84, and the categories it decomposes into are unlabelled because their names are the employer’s.
Also at Helloworld, by kind
Demand-forecasting, pricing-optimisation and customer-segmentation models in scikit-learn and PyTorch, each with a handoff package; A/B tests, causal-inference studies and statistical models on production telemetry. Method
A strategic financial analysis of ASX:HLO (DCF, segment revenue) for board presentation, and GA4 analytics pipelines with automated executive reports. Reported
FX-rate automation, estimated inside the proposal at about 200 person-hours a year saved. Proposed
Also in the role: GDS integrations (Amadeus, Sabre, Travelport), Tourplan, SiteMinder, REST and SOAP/XML services, and MS SQL.
Method
Source: Method, stated here. Date: Jan 2026, first week. Scope: The four levers of the configuration audit, what each means, and how the figure was measured. Not shown: Rule names, thresholds, host counts and the absolute cost.
Four levers, applied in the first week:
- Performance audits. Read what the platform was ingesting and from where, host by host and rule by rule, before changing anything.
- Monitoring-rule optimisation. Rules reviewed for overlap with other rules and for whether anything acted on them.
- Redundant instrumentation removed. Instruments that measured the same thing as another on the same host, or that had no consumer.
- Licence right-sizing. The licence tier matched to the ingestion that remained.
How it was measured. The baseline was the annual platform cost before the changes; the comparison was the same figure after them. The difference is the 16 %. It is a cost figure, so it is a measured outcome, not an estimate; it is shown here in relative units because the absolute figure is the employer’s.
What the bar does. The 100 separates into the instrumentation categories the cost was made of, drawn as unlabelled segments; the redundant segments slide to the margin with their reason; the bar lands at 84. Only the two ends of the bar are figures: 100 and 84.
Annual Dynatrace platform cost, in relative units
84The bar starts at 100 and lands at 84.
- Redundant instrumentation: two instruments measuring the same thing on the same host.
- Monitoring rules that fired on the same condition as another rule.
- Rules and instruments with no consumer: no dashboard, no alert, no reader.
- Licence tier right-sized to the ingestion that remained.
Relative units; the 16 % is the measured figure (Helloworld Travel, first week, 2026). Segment sizes are illustrative: only the two ends of the bar, 100 and 84, are figures.
Raw
Source: Method, stated in general form. Date: Jan 2026. Scope: The rules behind the four levers, with the employer’s rule names, thresholds and host counts left out. Not shown: Rule names, thresholds, host counts and the absolute cost.
The decision rules
- Count every ingestion source before touching any. A cut that cannot be traced to a named source is not made.
- A rule that fires on the same condition as another rule is one rule. Keep the one with a consumer.
- A rule with no consumer, no dashboard, no alert and no one who reads it, is retired, and the retirement is written down with its date.
- Two instruments measuring the same thing on the same host are one instrument.
- An instrument with no consumer is removed the same way as a rule with none.
- The licence tier follows the ingestion that remains, never the other way round.
- Measure the annual cost before and after on the same basis, and publish the relative figure only.
