A Migration Year Through a FinOps Lens: A Fictional Team Story

Rajesh Kumar

Rajesh Kumar is a leading expert in DevOps, SRE, DevSecOps, and MLOps, providing comprehensive services through his platform, www.rajeshkumar.xyz. With a proven track record in consulting, training, freelancing, and enterprise support, he empowers organizations to adopt modern operational practices and achieve scalable, secure, and efficient IT infrastructures. Rajesh is renowned for his ability to deliver tailored solutions and hands-on expertise across these critical domains.

The following story is fictional. The team, the company and every number in the narrative are invented for illustration, and no real organisation is meant. The FinOps ideas threaded through it are drawn from the entry on this subject in the xOps School Atlas, and wherever a real figure appears it is attributed to that entry.

Chapter one: the quarter that looked fine

A mid-sized delivery-tracking company decided to move its platform from two rented datacentre racks to a public cloud over eighteen months. The platform team had six engineers, one product manager and a finance partner who attended the monthly review mostly to listen. For the first quarter the bill sat close to the forecast, and everyone relaxed.

The first sign of trouble was a line in the finance partner’s notes: the forecast had been built from last year’s trend. It did not include a customer campaign planned for the following month or the fact that two environments would run in parallel until the old racks could be switched off. The Atlas lists exactly these blind spots among common forecasting anti-patterns, trend lines that miss campaigns and launches, and it describes dual-running costs as the defining feature of a long migration.

Chapter two: the monthly meeting that changed

In month four the product manager asked a plain question: “Did the bill go up because we have more parcels, or because things got more expensive?” Nobody could answer. The finance partner suggested splitting the change into pieces. How much came from volume, how much from price per unit, and how much from new things the team had switched on? The Atlas calls this a price-volume variance analysis, with separate volume, rate and scope effects, and suggests keeping back a reserve roughly as large as the typical forecast error.

Splitting the numbers took one afternoon and changed the conversation. About half of the rise was volume, which the product manager happily owned. A smaller part was scope, since a new tracking feature had gone live. The rest was a rate effect from a service that had been left on its default pricing. For the first time the room argued about causes rather than blame.

Chapter three: a purchase nearly goes wrong

Around month six a well-meaning engineer proposed buying a three-year commitment for the whole expected footprint. The saving on the price list looked enormous. The finance partner remembered a warning from the Atlas: the commitment is evaluated hour by hour, an unused hourly amount is lost, and the sizing should follow the lowest steady level of eligible usage, not the busy peak. The team rechecked the usage graph and saw that nights ran at less than half the daytime level.

They made a smaller purchase, matched to the night-time floor, and subtracted the racks’ workload that was still scheduled to move. They also held back the flexible layer for later and agreed to buy the next tranche only after reviewing utilisation daily for the first few weeks. Months afterwards, the engineer admitted that the first plan would have locked the company into paying for capacity that the migration was about to make unnecessary, a trap the Atlas describes as buying for a shape the estate no longer has after optimisation.

The culture side of that change interested the wider organisation. A reader curious about how habits shift in a team may like the practical guide on managing culture change in delivery teams.

Chapter four: the argument about who owns the leftover

By month nine the team had labelled most resources with an owning squad. Yet one report stubbornly showed a block of spend with no owner. It consisted of support charges, a shared monitoring platform and some taxes. Labels, the finance partner pointed out, would never fix this, because none of those charges belonged to one resource. The Atlas makes the same point: tagging does not resolve allocation by itself, and a shared-cost strategy is needed in addition.

The team agreed on a simple split of the shared pool, in proportion to each squad’s own direct spend. It was not perfect, and they wrote down that it was a policy choice. Within a month the unowned share fell from a nervous double-digit figure to a small remainder, and the squads stopped treating the shared pool as free.

Chapter five: the right-sizing that was done carefully

In month twelve the engineers saw several machines running at low utilisation and proposed shrinking them. The product manager worried about customers. The team followed the sequence from the Atlas’s right-sizing playbook: choose from data that includes the peaks, agree on service-level guardrails before touching anything, change a single dimension at a time, watch through a whole business cycle, and roll back if a service level slips. They also notified the finance partner, because some of those machines were covered by a commitment, and shrinking a covered machine could leave part of the purchase unused.

One change was rolled back after a quiet-looking weekend revealed a batch job that needed the memory. Nothing broke for customers, and the rollback itself counted as a success of the process.

Chapter six: what the team learned

At month eighteen, the old racks were gone and the bill was higher in absolute terms than at the start, because the business carried more parcels. What had changed was the unit cost, which fell, and the number of surprises, which almost vanished. The team kept three habits: a daily look for unusual spend, a weekly squad-level review, and a monthly session on forecast and allocation. The finance partner no longer sat quietly.

Looking back, the team listed the following lessons, which match the Atlas’s guidance closely:

  • Judge the bill by cost per unit of work, not by its size.
  • Size commitments to the quiet floor and buy in stages.
  • Tag what you can, and agree a rule for what you cannot.
  • Change one thing at a time and keep a rollback ready.
  • Treat the practice as shared decision rights, not a task for one department.

A final note on risk. Any programme that touches production needs boundaries, and the Atlas states that service levels are never to be sacrificed for savings. The piece on weighing risk in reliable, secure delivery expands on how to set those boundaries before the first purchase or resize.

Remember that the company in this story does not exist. What does exist is the pattern: a forecast that missed a parallel-running period, a purchase sized to the wrong level, and a pool of costs with no obvious owner. Most teams will recognise at least one of them.

Leave a Reply