Blog
IT outsourcing guides for IT leaders
What we learned delivering for French IT departments: choosing between embedded teams and fixed price, taking over undocumented estates, pricing an automation, and not getting caught by a day rate that is too low.
Contracts
Time and materials or fixed price: how to choose without regret
Both models are legitimate. The bad choice is taking an embedded team because the scope is fuzzy, then steering it like a fixed-price project. Here are the criteria that actually decide it.
· 4 min read
Maintenance
Application maintenance: what it really costs, and what you think you are paying for
Maintenance contracts are rarely compared on the right metric. Ticket counts say nothing; restoration time and accumulated debt say everything. How to read a maintenance proposal.
· 4 min read
Maintenance
Taking over an application nobody documented
The market's reflex is to propose a rewrite. It is the most profitable assignment to sell and the riskiest to buy. Here is the alternative, and what it actually costs.
· 4 min read
Procurement
What a specification must contain to get a firm price
Most fixed-price overruns come not from the code but from a scope never written precisely enough to be held. The six elements that are almost always missing.
· 4 min read
Procurement
What a day rate that is too low is hiding
A €180 day rate is not a bargain, it is a cost transfer. Where the difference goes, and how to compare two proposals that are not comparable.
· 4 min read
Procurement
Nearshore Morocco or far offshore: what the time zone actually changes
The rate gap between Morocco and South Asia is real. What it does not show is the cost of the time difference, the language, and the number of round trips needed to resolve an ambiguity.
· 4 min read
GDPR
Using a processor outside the EU without exposing your compliance
Transfers outside the European Union are not prohibited, they are governed. What you must obtain from your provider, and what your DPO will check in your register.
· 4 min read
GDPR
DPA: the nine mandatory items of Article 28, and the ones people forget
A processing agreement missing these items is non-compliant, however long it is. The list, plus the three additional clauses that actually protect you.
· 4 min read
GDPR
Standard contractual clauses: picking the right module and documenting the transfer
Signing the SCCs has not been enough since Schrems II. What to add, which module applies to your configuration, and what an acceptable transfer impact assessment looks like.
· 4 min read
GDPR
Answering a vendor security questionnaire without losing three weeks
Questionnaires all look alike. A base set of answers prepared once turns three weeks of work into two days, and it surfaces your gaps before the client finds them.
· 4 min read
Data
Migrating a Talend estate: what to decide before writing a line
A Talend migration rarely fails technically. It fails because nobody settled what should be migrated, what should be deleted, and who arbitrates the business rules.
· 4 min read
Data
Delete before you migrate: the cheapest gain in a data project
In a legacy estate a significant share of flows has no consumer left. Identifying them costs three weeks and mechanically reduces the cost of everything that follows.
· 4 min read
Data
Data lineage: why you cannot decide anything without it
Without lineage, every prioritisation decision is blind and every deletion is frightening. How to build it in three weeks, even on a heterogeneous estate.
· 4 min read
Data
Two dashboards, two numbers: it is almost never a bug
The most frequent cause is not technical. It is two different business definitions of the same metric, each correct in its original context. How to get the monthly reconciliation meeting off the agenda.
· 4 min read
AI
Why AI proofs of concept never reach production
It is not the model quality that blocks them. It is the move from laptop to server, a metric never settled, and data whose quality was not measured before starting.
· 4 min read
AI
The baseline before the model: the rule that saves time and buys credibility
A moving average, a simple business rule, last month's value. Any model must beat that reference to ship — and the reference regularly wins.
· 4 min read
AI
MLOps: what separates a model that works from a model that lasts
Moving from notebook to production often costs more than the model itself. The seven components that must exist before talking about go-live.
· 4 min read
AI
Model drift: the invisible cost of an unmonitored model
An unmonitored model degrades silently, often for months before anyone notices. What to measure, how often, and who should receive the alert.
· 4 min read
Automation
RPA: calculate the return before building, not after
A third of the processes submitted for automation fail the profitability threshold. How to time them, price them, and turn a project down with a written rationale.
· 4 min read
Automation
RPA or API: the deciding criterion, and why the obvious answer is often wrong
Where an API exists it is almost always preferable. But waiting for an API that will not arrive for eighteen months sometimes costs more than a robot.
· 4 min read
Automation
The processes you should not automate
A third of automation candidates should not be automated. The six profiles that systematically fail, and why identifying them early saves you money.
· 4 min read
Cloud
Cloud migration: why waves beat the big bang
A single-weekend migration mostly succeeds in sales decks. Wave-based migration costs slightly more and fails far less often.
· 4 min read
Cloud
Cloud bill out of control: the six items to check before negotiating
Cloud savings rarely come from price negotiation. They come from sizing, orphaned resources, and non-production environments left running overnight.
· 4 min read
Cloud
Do you really need Kubernetes?
Kubernetes earns its keep above a certain number of services and with a team able to operate it. Below that, a managed application service costs less and breaks less often.
· 4 min read
Cloud
Infrastructure as code: the one test that proves it actually exists
Many organisations have Terraform files and no infrastructure described as code. The real test fits in one sentence, and it is rarely passed.
· 4 min read
Cloud
The restore test nobody runs
A backup never restored is not a backup, it is a hypothesis. How to run the test, and why the number it produces often changes the continuity plan.
· 4 min read
Cloud
Observability: an alert that does not say what to do ends up ignored
The problem is almost never a lack of alerts, it is an excess of them. How to build an alerting system teams still read after six months.
· 4 min read
Maintenance
Your WinDev developer retires in three months
It is the most frequent and worst-handled situation. What to do while they are still there, and the mistake that costs the most.
· 4 min read
Maintenance
Should you leave WinDev? The four questions that decide
The criterion is not the age of the technology. A supported, documented WinDev application known to several people has no reason to be migrated.
· 4 min read
Maintenance
Opening a legacy application through an API rather than rewriting it
When the real need is integrating the application with the rest of the estate, exposing a thin API costs a fraction of a rebuild and leaves the business core intact.
· 4 min read
Delivery
Quantifying technical debt so an executive committee listens
'We need to refactor' never gets a budget. Debt expressed in days of delay on future deliveries does. How to make the conversion.
· 4 min read
Delivery
Onboarding an external team without collapsing your internal velocity
The most frequent mistake is starting at full headcount. A progressive ramp costs a few weeks and saves the assignment.
· 4 min read
Contracts
Reversibility: the clause you negotiate at signature, never on the way out
An exit clause negotiated at the moment you want to leave is no longer a negotiation. What it must contain, and the test that proves it works.
· 4 min read
Procurement
Comparing five proposals that are not comparable
Providers answer different scopes in different units. A six-column normalisation grid makes comparison possible in half a day.
· 4 min read
Delivery
Governing a remote team: what to require, and how often
The risk of remote delivery is not technical, it is informational. The minimum arrangement that restores the visibility you would have over an internal team.
· 4 min read
Twenty minutes is enough to know whether we are useful
No sales deck. You describe the need, we say whether it is in scope, at what price and on what timeline. If it is not for us, we say so during the call.