I Priced 14 Dedicated Python Teams in Dubai — the 7-Step Teardown That Cut Our Budget 31%

Fourteen proposals for the same brief — a dedicated Python team of five for a data-heavy platform — and a spread between the cheapest and the most expensive of more than two to one. That spread is not what it looks like. Almost none of it is quality. Here is the method we now use to find out what a quote actually costs.

Why the headline rate tells you nothing

Every rate card in this market presents the same way: a table of roles, a band of seniority, a day rate in dirhams or dollars. It looks comparable. It is not.

A day rate is a price for attendance. What you are buying is output, and the ratio between the two varies more between vendors than the rate itself does. Two proposals with identical headline rates can differ by a third once you account for who pays for public holidays, whether onboarding weeks are billed at full rate, how much coordination overhead lands on your invoice, and how many of the people in the team are engineers rather than managers.

The exercise below took about ten working days across fourteen vendors. It reduced our best-value option by 31 per cent against its first quote, and — more importantly — it changed which vendor we chose. The cheapest headline rate finished sixth.

Step 1 — Convert every quote to a cost per productive engineer-day

One number, computed identically for every vendor. Take the total annual cost of the proposal, including every fee, and divide it by the number of engineer-days you will actually receive.

Building the denominator is the whole exercise. Start from 260 weekdays, then subtract, for each named person:

• Public holidays in the delivery location — and check whether they are billed. Several vendors bill them; several do not; none volunteer the information.
• Annual leave, and whether cover is provided at no extra cost or the seat simply goes empty.
• Onboarding time at the start of the engagement, billed at full rate in most proposals we received. Two to four weeks per engineer is normal, which on a five-person team is a material number.
• Ceremony and coordination time, if the proposal attributes it to your budget rather than absorbing it.

Do this once and the shortlist reorders itself. In our case, the vendor with the second-highest headline rate produced the lowest cost per productive day, by a wide margin, because it billed no onboarding and provided leave cover.

Step 2 — Separate the seniority ladder from the billing bands

Vendors publish billing bands: junior, mid, senior, lead. These bands describe what you are charged. They do not necessarily describe who turns up.

The specific pattern to look for is a proposal where senior engineers are heavily present in months one to three and thin out afterwards. It prices attractively over a year and delivers poorly, because the people who convinced you in the pre-sales meetings rotate off once the contract is signed.

Two questions cut through this immediately, and both should be answered in writing:

• Which named individuals hold the senior seats, and for how long are they committed? A vendor that cannot name them at proposal stage does not have them free.
• What happens to the rate if a senior is replaced by a mid-level engineer? If the answer is “nothing”, the band is a billing construct rather than a staffing commitment.

We now insist on named continuity for one senior seat for the full contract term, and we accept a higher rate for that seat specifically. It is the single clause that most improved delivery quality, and it costs less than one additional mid-level engineer.

Step 3 — Price the roles you did not ask for

We asked for five engineers. Eleven of fourteen proposals contained six to eight people. The additions were always presented as free value: a delivery manager, a scrum master, a quality lead, an account director.

None of them are free. They appear either as billable line items or as a margin uplift distributed across the engineer rates. Either way you pay for them, and the question is whether you need them.

Our rule: one coordination role per team of six or more, none below that. Below six people, a competent senior engineer covers coordination and your own product owner covers priorities. Above six, a dedicated coordinator earns their cost within a month.

Removing the roles we did not need accounted for roughly a third of our total reduction, and no vendor objected. They add them by default because most clients do not ask.

Step 4 — Ask who pays for the bench

A dedicated team means the vendor reserves people for you. Reserved capacity has a real cost, and someone pays it. There is nothing wrong with that; there is a lot wrong with not knowing.

Three questions, asked plainly:

• If our backlog empties for three weeks, is that time billable?
• May you assign our reserved engineers to another client during that period?
• If we reduce the team by one person, what notice do you require and what do we pay during it?

The combination of answers tells you what you are actually buying. “Billable, and we may reassign” means you are funding a shared pool while carrying the cost of a dedicated one. “Billable, and exclusively yours” is a genuine dedicated model and is legitimately more expensive. “Not billable” usually means the cost is already inside the rate.

Step 5 — Cap the annual escalation in writing

Almost every proposal we received contained an annual increase, expressed vaguely: adjusted in line with market conditions, reviewed annually, subject to inflation. Vague escalation is the most expensive clause in a multi-year engagement, and it is almost never negotiated because it does not affect year one.

Compounded over three years, a difference between a capped 3 per cent and an uncapped “market adjustment” is easily the largest single variable in the contract — larger than the rate you spent two weeks arguing about.

What we now require: a numeric cap, a named index or a fixed percentage, and a notice period of at least ninety days before any increase takes effect. Every vendor agreed to some version of this. None had offered it.

Step 6 — Find the three clauses that move the total by 20%

Across all fourteen proposals, the same three clauses accounted for most of the difference between the quoted price and the price we would actually have paid over three years.

Billed onboarding. Two to four weeks per engineer at full rate, repeated every time someone is replaced. On a team of five with normal turnover, this is several weeks of pure cost per year. Ask for it to be unbilled on replacements at minimum — a vendor replacing its own leaver should not bill you for the ramp-up.

Uncapped escalation. Covered above, and it is the biggest single lever in the contract.

The notice period. Ninety days fully billable is standard and is worth negotiating hard, because it is the clause that determines what a mistake costs you. It also has a variant worth asking for: the right to reduce the team by one or two people on shorter notice, rather than an all-or-nothing exit.

None of these three appears in the rate table. All three are in the terms document that nobody reads until the second year.

Step 7 — Negotiate the exit before the start date

The last step is the one most teams skip, and it is the one that protects everything else.

Before signature, agree in writing what happens when the engagement ends: who owns the code and the infrastructure accounts, what documentation is delivered and in what state, how much handover time is included at no extra cost, and how quickly access is revoked. Put a number on the handover — ten working days is reasonable for a team of five — and make it a deliverable, not a courtesy.

The reason is simple: after signature your leverage drops to zero, and the moment you need these terms is precisely the moment the relationship is worst. Vendors that agree readily are telling you something reassuring about how they operate. Vendors that resist are telling you something too.

This is the same principle we apply to knowledge transfer on longer engagements, and our colleagues at HireDeveloper.sg reach the same conclusion from a different market: the handover clause predicts the quality of the whole engagement better than any reference call. In Tokyo, where the constraint is supply rather than price, the team at JapanDev finds the equivalent lever is continuity of named engineers rather than commercial terms.

Three mistakes we made first

Running a rate-only negotiation. We spent two weeks pushing on day rates and won four per cent, while leaving the escalation clause untouched. That was the wrong fight, and we would have lost more over three years than we saved in year one.

Accepting the proposed team shape. Vendors size teams to their own bench, not to your problem. Specify the shape yourself — how many engineers, what seniority, which coordination roles — and ask them to price it. The proposals become comparable for the first time.

Skipping reference calls about the second year. Everyone asks references about delivery quality. Almost nobody asks what happened at the first escalation, the first replacement, or the first scope reduction. Those three moments predict the relationship far better. The same principle applies when you are sizing the underlying build: our breakdown of what it costs to build a real estate platform in Dubai shows how much of a budget sits in decisions made before anyone writes code.

Frequently asked questions