How to Improve Margins in a Software Development Agency: Estimation, Change Control, Bench, Rates and Support
Software agency profit margins rarely collapse in a single bad month, because they leak slowly through five places that every founder recognises once they are named: estimates that were optimistic before the contract was signed, change that was absorbed rather than priced, bench time that nobody planned for, rate cards that have not moved in years, and support tails that were sold cheaply and never revisited. Improving margin means closing those leaks in roughly that order, and keeping them closed.
None of this requires a transformation programme. It requires knowing which of the five is costing you most, which is usually not the one that feels most painful, and then changing the habit behind it so that the fix holds after the founder stops watching it.
Start by measuring where the margin actually goes
Most agencies know their overall gross margin, but far fewer know it by project, by client and by type of work, which is where the useful truth sits. A blended figure can look acceptable while two fixed-price builds quietly subsidise themselves out of a profitable retained account, and nobody notices until the retained client leaves.
If you have not yet done this properly, a project profitability analysis is the first step, because it turns a vague sense that margins are slipping into a ranked list of the engagements responsible. It is also worth knowing what a healthy figure looks like for your delivery model, which our piece on what a good gross margin looks like for a consultancy covers, so that you are fixing towards a sensible target rather than an arbitrary one.
Rule of thumb: if you cannot say which three projects lost you the most money last year, you are not yet ready to change your pricing, because you do not know what you are pricing against.
Estimation discipline on fixed-price builds
Fixed-price work is where software agency margins most often erode, and the cause is rarely incompetence in the estimate itself but the conditions under which it was produced: a salesperson or founder under pressure to win, a senior developer asked for a number on a Friday afternoon, and no record afterwards of how the estimate compared with what actually happened.
A disciplined estimation process has a few features that are simple to describe and surprisingly rare in practice:
- Estimates are made by the people who will deliver, or at least reviewed by them, so that the number reflects the team you actually have rather than the team you wish you had.
- Assumptions are written down and sent to the client, because an assumption that lives only in a developer's head cannot be used to justify a change request later.
- Contingency is explicit, held as a named line that the delivery lead manages, rather than quietly padded into every task where it is spent without anyone deciding to spend it.
- Actuals are compared with estimates at the end of every project, so that the business learns which kinds of work it consistently underestimates and adjusts accordingly.
That last point matters most, because without it the same optimism repeats indefinitely. A common pattern is that integrations, data migration and testing are underestimated on almost every build, while the visible feature work is estimated reasonably well, and the only way to see that is to look.
It is also worth asking whether some of your fixed-price work should be fixed price at all, which our guide to time and materials versus fixed price works through engagement by engagement.
Change control that the client can live with
The second leak is change that gets absorbed, usually because the project manager does not want a difficult conversation and the founder, when asked, says yes to keep the client happy. Each small concession seems reasonable on its own, but together they can consume the whole of a project's margin without a single decision having been made to give it away.
Good change control is less about saying no and more about making change visible, so that every request is logged, sized and either priced, traded against something else in scope, or consciously given away as goodwill by someone with the authority to do so. Clients tend to accept this far more readily than founders expect, provided the process was explained at the start rather than introduced halfway through a dispute.
Bench time and how it hides
Bench time is the cost of people who are paid but not billable, and in a software agency it tends to hide inside projects rather than show up on its own. When a developer finishes early or a project stalls, they are often attached to another build where their hours inflate the cost without adding much value, so the bench disappears from the utilisation report and reappears as poor project margin.
The fix is partly forecasting, matching people to pipeline far enough ahead that a lumpy quarter does not leave a team idle, and partly honesty in timesheets, so that unallocated time is recorded as exactly that. Once bench time is visible it can be managed, whether by bringing forward internal work, adjusting hiring plans or making a deliberate choice to carry capacity ahead of a large win.
Rate cards that keep pace with cost
Many agencies set their day rates when the business was smaller, salaries were lower and the team was more junior, and then left them largely untouched while every cost underneath them rose. The result is a rate card that no longer reflects either the cost of delivery or the value the client receives.
Reviewing rates is uncomfortable but straightforward: work out the fully loaded cost of each role, including the non-billable time that role realistically carries, set a target margin above it, and compare that with what you actually charge. Where the gap is large, existing clients usually move on a planned schedule with notice, while new clients go straight onto the current card.
A rate card that has not moved in three years is not a pricing decision; it is the absence of one.
Support contracts and the long tail
The final leak is the support tail, which is often sold at the end of a build as an afterthought, priced low to close the deal, and then left running for years while the system ages, the original developers leave and the work becomes more expensive to deliver. Support can be some of the most valuable revenue an agency has, because it is predictable, but only if it is scoped and priced as a product rather than a favour.
A sound support contract defines what is included, sets response expectations the team can actually meet, caps or prices work that falls outside the core service, and includes an annual review of both scope and price. Contracts that lack those features are worth renegotiating at renewal, starting with the clients whose systems consume the most time for the least fee.
Where Vitori fits
Every one of these fixes can be made by a capable founder or operations lead, whether with Vitori or anyone else, and for some agencies the honest answer is simply to find the time and do it. Where the difficulty is that the founder is still the person estimating, approving changes and setting rates, the problem is less about margin than about how the business is run, and that is the work Vitori does.
Vitori works inside the leadership team, uses its Operational Scale Framework to assess Growth, Delivery and Operations, implements the changes itself and stays accountable until they hold, with engagements scoped by agreeing outcomes first and then the timescale. If that sounds useful, you can get in touch with Vitori to talk it through, with the aim of a business that runs, and scales, without the founder in every decision.
Common questions about software agency margins
Should a software agency stop doing fixed-price projects to protect margin?
A software agency does not need to abandon fixed-price projects to protect margin, but it does need to reserve them for work it understands well and can estimate from evidence. Fixed price suits well-defined builds with stable requirements, while discovery-heavy or uncertain work is usually safer on time and materials, or on a fixed-price discovery phase followed by a priced build.
How often should an agency review its day rates?
An agency should review its day rates at least once a year, ideally alongside salary reviews, so that rates move in step with the cost of delivery. The review should compare the fully loaded cost of each role with the rate charged, and existing clients should be given clear notice before any increase takes effect on their work.
Is low utilisation always a margin problem?
Low utilisation is not always a margin problem, because some non-billable time is deliberate, such as sales support, training or carrying capacity before a large project starts. Utilisation becomes a margin problem when the non-billable time is unplanned, unrecorded or hidden inside client projects, where it inflates delivery cost without anyone having chosen to spend it.
