Home › Services › Cloud cost optimisation
Your cloud bill grew and nobody can say why
Finance wants an answer this quarter, and the line item that did it is not obvious from the console. We find the number, fix the causes, and, if you would rather, take our fee out of the saving instead of out of your budget.
- Fee
- Fixed fee, or a share of what we save you + GST · 2 weeks
- Runs for
- 2 weeks
- You provide
- Billing + read access
- Clouds
- Public cloud and hybrid
For scale: an estate spending ₹10 lakh a month, at the 20% we aim for, keeps about ₹24 lakh a year. The fee is under two months of that.
Almost none of it is dramatic.
In the Mumbai region a NAT gateway costs about ₹3,100 a month before a single byte crosses it, and the recommended one-per-zone across three zones makes that ₹9,400 a month, idle.
It's gp2 volumes nobody switched to gp3, which is around 20% cheaper for the same disk with no downtime. It's 500 GB a day replicating across availability zones at roughly ₹28,600 a month, because a subnet was picked in a hurry two years ago. It's snapshot policies with no expiry, quietly compounding.
Above roughly ₹1 lakh a month of spend, data transfer alone is usually between 10% and 30% of the bill. It almost never appears on anyone's dashboard, because the console groups it under headings nobody opens.
There are three places savings come from, and doing them out of order is expensive:
- Waste: idle instances, orphaned volumes, oversized databases, unattached public IPv4, snapshot sprawl. Typically 10–20%, with no architecture change and no downtime.
- Commitments: Savings Plans, committed-use discounts, reserved capacity. The next slice, but only worth buying against a baseline that has already been cleaned up.
- Architecture: transfer paths, storage classes, scheduling, managed-service choices. The rest, and the part that takes real design work.
If a provider quotes you 30% in the first week, ask what they are committing you to. Locking three-year commitments onto an un-rightsized estate saves money this quarter and costs you for the next twelve. We would rather show you a smaller honest number and then beat it.
The default is a fixed fee, agreed before we start. You own the output, you can hand it to your own team, and nothing about our incentive changes if the number turns out to be small.
The alternative is outcome-linked: we implement the changes and take an agreed share of the saving we can demonstrate, measured against the baseline we took before we touched anything. No saving, no fee. Most people who ask for this ask for it because they have been sold a report before and would rather buy the result.
It comes with a warning we would rather give you now than argue about later. A share-of-savings model quietly rewards us for finding big numbers.
The honest version of this work sometimes means telling you that your estate is already in reasonable shape.
So the baseline is agreed and frozen in writing before any change is made, commitments bought during the engagement are measured on what they drew, not on list price, and the arrangement runs for a fixed window, and never becomes a permanent tax on your bill. If a provider offers you share-of-savings without all three, the model is working for them.
- Read-only access to your billing data and, ideally, a read role in the accounts.
- Every meaningful line item traced to the resource and the team that owns it.
- A costed list of changes, each with the saving, the effort, the risk, and whether it needs a maintenance window.
- A separate list of things we would not change, and why.
- A one-hour walkthrough with whoever has to approve the spend.
You can hand the output to your own team and never speak to us again. It's written to be actionable by someone else. If you'd rather we implemented it, that's a separate fixed quote once the list exists.
Sometimes the assessment concludes that a workload was never suited to public cloud, and the honest recommendation is to leave it, or bring it back. Steady, predictable, always-on compute with heavy egress is frequently cheaper on hardware you own, and a licence bound to a physical core count can make the cloud version arithmetic indefensible. Data residency or a plant system with a latency budget can settle it before cost even enters the argument.
So we also design and build the hybrid case: what stays on-premise, what runs in the cloud, the interconnect between them, one identity across both, and backup and DR that treats the two halves as one estate, not two contracts. It is a design decision with a cost model attached, not a philosophical position, and we have no reseller margin riding on which way it goes.
Keeping it from creeping back
Savings decay. Left alone, most estates drift back inside a year.
Talk to us about thisSavings decay. Left alone, most estates are back where they started inside a year. If you want that handled, it folds into an AMS contract: anomaly detection that flags a spend change the week it happens, a live cost dashboard instead of a monthly PDF, monthly forecast-versus-actual, and right-sizing raised as a change, never applied silently.
If you would rather work through it yourself first, the 26-point checklist we use on a first pass is published in full, with the rupee figures and nothing held back.
What you walk away with
Yours to keep, and to hand to anyone else, including a provider that isn't us.
Talk to us about thisEvery change with the rupee saving, the effort, the risk, and whether it needs a maintenance window.
The things we would leave alone, and why. Usually the more useful half.
What to buy, how much, and when, after the rightsizing, not before it.
One hour, with the person who has to approve the spend, not just the person who runs it.
Asked before you ask
The ones that come up on nearly every first call.
Talk to us about thisDo you take a percentage of the savings?
We can, and some buyers prefer it: we implement the changes and take an agreed share of the saving we can demonstrate against a frozen baseline.
It is not the default, because the model rewards us for finding big numbers instead of the right ones. If you choose it, we agree the baseline in writing first and the share runs for a fixed window, not forever. A fixed fee remains the cleaner option if you mainly want to know the truth about your estate.
What if you don't find much?
We'll tell you in week one instead of spending your fee proving a point. It happens, usually with estates that have been through a serious FinOps effort already. On a fixed fee you'd stop there and pay for the week, not the fortnight; on the outcome-linked model there is simply nothing to invoice.
Can you assess a hybrid or on-premise estate too?
Yes, and sometimes that is where the answer is. We look at over-provisioned hardware and idle capacity the same way on either side, and the recommendation is occasionally that a workload comes back out of the cloud. We have no reseller margin either way.
Do you need production access?
Read-only is enough for the review. Billing export access plus a read role gets us 95% of the way; the rest comes from a couple of conversations with whoever runs the workloads.
We're on a cloud you may not work with.
Then we may be the wrong people for the cost review, and we will say so on the first call. We work across the major cloud OEMs, and where yours is not one we run every day we would rather tell you than learn it on your bill.
Not sure this is the one you need?
Tell us the symptom, not the service, and we'll say which of these applies, or that none of them do. A reply within one working day, from someone technical.
The rest of what we do
Most engagements start with one of these and grow into another.