Frequently asked questions

An assessment. We source candidates, test them properly, and give you two or three people with our reasoning and our reservations. Forwarding CVs is not a service and it is what most of this market sells.

Most agencies are paid on placement and optimise for volume through the funnel, which is why your engineers end up interviewing ten people. We do the technical filtering so they interview two or three, and we tell you what is wrong with each of them.

Yes, when that is our honest view. If delivery is slow because nobody can review, or because the architecture needs attention, more developers makes it worse. That conversation costs us an engagement and it is the only way this works long term.

A running environment on day one, a named reviewer with genuine capacity, access to the real system, and an agreed overlap window. Almost every engagement that goes badly is missing at least two of those.

Not by comparing a salary against an hourly rate, which is the comparison almost everyone makes and it is wrong. The fair version puts the fully loaded employer cost of a local seat, including payroll taxes, benefits, recruitment and the months it stands empty, against the total cost of the alternative including coordination overhead.

The published median annual wage for software developers is $135,980. On top of that sit employer payroll taxes, benefits, equipment and recruitment cost, none of which appear in a salary figure. Every location page on this site shows the employer payroll tax arithmetic for that market using the statutory rates.

Because a number without the context of the role, the level and the market is exactly the kind of unsourced figure this site exists to avoid. Terms are agreed in writing before anything starts.

The US Bureau of Labor Statistics Occupational Employment and Wage Statistics, May 2025 release, read on 25 September 2026. Every page cites it. The figures are metropolitan-area estimates covering all employers in that occupation, not rates we are quoting.

Because the Bureau classifies by occupation rather than by technology, so no official figure exists for React or Python specifically. Anyone quoting one has modelled it or made it up. We publish the occupation band instead and say which occupation each technology belongs to.

As a direction, not as a growth rate. The Bureau does not design these releases to be read as a time series, because the survey panels overlap and the methodology is revised between releases. We publish the caveat alongside the table on every page that shows one.

When the Bureau publishes, which is annually. Every page states the release and the date we read it, so you can tell whether what you are reading is current.

Outside the US, working to a committed overlap with your hours. Where a role genuinely requires presence in a specific US city, we will tell you rather than pretend otherwise.

Agreed before the engagement starts, based on which hours genuinely need covering. The pattern that works is a fixed window of a few hours used for review and decisions rather than for status.

Not within an engagement. Split attention produces the outcomes you would expect and we do not do it.

We handle the replacement and the handover, and the terms are in the agreement. What we will not do is present a replacement who has not been assessed to the same standard just to fill the seat quickly.