23 June 20267 min read
The cheapest quote in a tender is not cheap by accident. A rate is the output of an arithmetic: what the people are paid, what proportion of their time is billable, what the provider spends on management, quality and recruitment, and what margin is left at the end. To arrive at the lowest number in the pile, at least one of those had to be smaller.
The useful question is therefore never whether the cheap quote is suspicious. It is which term was reduced, because each one produces a different and quite predictable problem.
The five things that get removed
- Seniority: the mix shifts junior. Under a blended rate this is invisible until the work arrives.
- Attention: people are shared across several clients, so the day you bought is a fraction of a day of thought.
- Quality layer: no code review, no QA, no testing standard. Acceptance is entirely your problem, performed by your senior people.
- Coordination: no project management, no written specification support. That work still exists, and it moves onto your team.
- Selection: thinner pre-vetting, faster placement, a wider variance in who arrives.
Notice that four of the five are not removed at all. They are relocated, onto the buyer, where they are performed by more expensive people who had other work to do. The saving is real on the invoice and partly fictional in the accounts.
The rework multiplier
The mechanism that turns a cheap rate into an expensive year is rework, and it compounds in a way that intuition does not track.
When work is not accepted first time, you pay for the second attempt, you pay for the second review, and you pay for the delay to everything queued behind it. Your reviewer, who is senior and scarce, spends the week reviewing instead of building. And the feedback loop lengthens: the longer a piece of work takes to be accepted, the later its lessons reach the next piece.
This is why the only rate comparison that survives contact with reality is cost per accepted unit of work rather than cost per day. A provider whose work is accepted first time at a higher rate can be the cheaper one by a wide margin, and no rate card will ever show it.
The second-order costs
There is a cost above rework that finance never sees. When output is unreliable, your own team stops relying on it. They stop putting anything on the critical path with the provider's name on it. They start reviewing defensively, then start rewriting rather than reviewing, and eventually route the interesting work back in-house.
At that point you are paying for capacity that is not being used, and the engagement has failed while every invoice still looks correct. Nobody escalates it, because there is no single incident to escalate. It is worth naming because it is the most common way a cheap engagement actually ends.
An engagement rarely dies from a disaster. It dies when your own team quietly stops putting anything important on the other side of it.
Underpricing and the overrun
An underpriced engagement is also the one most likely to run past its budget, and the mechanism is structural rather than malicious. If the price was won by assuming an optimistic pace, an optimistic seniority mix or an optimistic scope reading, then the moment reality intrudes there are only two levers: reduce what is delivered, or bill more days.
So budget for the engagement running longer than planned, always, on any provider, and be specific about it. Decide in advance who has authority to approve an extension, what evidence they need to see, and at what point the answer becomes no. An overrun that has been anticipated is a decision. An overrun that has not is an emergency, and emergencies get approved because stopping halfway is worse.
When cheap is genuinely right
This is not an argument that price does not matter or that expensive is safe. Plenty of expensive providers relocate exactly the same costs onto you while charging for a coordination layer that does very little.
There is a real category of work where the lowest defensible rate is the correct choice: work that can be fully specified up front, that is loosely coupled to the rest of the system, where acceptance can be tested rather than judged, and where being wrong is cheap to correct. Queue-shaped, high-volume, well-bounded work. For that, buy on price, define acceptance tightly and do not pay for a coordination layer you will not use.
The mistake is applying that model to work that is coupled, ambiguous and judged rather than tested, which describes most of what teams actually want help with.
A selection rule you can defend
Rank the quotes on total cost of ownership over a realistic horizon, not on rate. Eliminate any provider that will not commit to a seniority mix. Then, among the survivors, choose the cheapest, and put the difference between it and the cheapest quote overall into the trial period and the definition of done.
That is a rule you can put in front of a CFO. It is not a rule that says buy expensive things. It says buy the cheapest one whose price you can explain, which is a different and much more defensible position.