Data Consultancies: Build Projects and Retainers Are Different Books

Say a firm delivers a six-month data platform build at a healthy margin, then rolls the client onto a monthly retainer to maintain and extend the dashboards. The build project had a clear scope and a team sized specifically for it. The retainer, priced early and rarely revisited, can quietly become unprofitable as the client asks for more dashboards, more data sources and more ad hoc analysis than the original retainer fee ever accounted for. A controller who tracks actual hours delivered against every retainer, not just the fee collected, catches that scope creep before a year has passed and the retainer has been running at a loss the whole time.

Data platform work often runs real cloud compute and storage costs, and third-party data sources sometimes carry their own licensing fees, both of which should be tracked by client or by engagement rather than lumped into one general infrastructure expense. A client running heavy queries against a large dataset can cost meaningfully more to serve than one with a lighter footprint, even at the same retainer price, and that difference is invisible until someone ties the cost to the client that generated it.

Who owns a custom model, pipeline or dashboard built during an engagement, the firm or the client, is a contract negotiation question that should be settled before the work starts, not an accounting question to resolve after the fact. What finance can do is flag when a contract's IP terms are unclear or missing before work begins, since an ambiguous ownership clause can create a genuinely expensive dispute later, well beyond anything a clean set of books could prevent on its own.

Once build and retainer margins are visible separately, and cloud costs are tracked by client, a fractional CFO can answer the harder questions: should the firm push more clients toward retainers for revenue predictability even if build projects carry better margins today, how should a retainer be repriced once scope creep is documented, and how much of the firm's capacity should go toward proprietary tooling that could improve margin on future builds versus billable client work right now.

The national median salary for accountants and auditors is $83,680 a year, a real cost for a smaller consultancy weighing a dedicated controller¹. A firm running mostly one-off build projects with a handful of simple retainers can often cover the need with a fractional CFO and outsourced bookkeeping. Once the firm is running several retainers with real variance in cloud cost and scope, the daily tracking discipline that catches drift early becomes worth a dedicated controller's fixed cost.

Say a controller's monthly report shows a specific retainer client's actual hours running 40% above what the retainer fee was priced to cover, and has been for the past four months. That data makes the repricing conversation with the client concrete and defensible: the firm can point to exactly what expanded, new dashboards, additional data sources, more frequent ad hoc requests, rather than making a vague case that the account has gotten busier. Firms without that tracking either eat the extra cost indefinitely to avoid an uncomfortable conversation, or raise prices without being able to justify the increase clearly, and both outcomes are worse than having the data in hand.

A brand-new client's retainer is priced off an estimate, not a track record, which means the first two or three months of actual delivery are the best early warning system a firm has for whether that estimate was right. A controller who reviews new retainer accounts specifically at the ninety-day mark, comparing actual hours against the original pricing assumption, catches a mispriced deal while it's still small and easy to renegotiate, rather than letting it run for a year on the same flawed assumption that started it, which is a far more painful conversation to have later.

A controller can catch retainer problems early by making these checks routine:

• Report build project margin and retainer margin separately, so the two businesses under one roof are never blended into one number.
• Compare actual delivery hours against the hours each retainer fee was priced to cover, and flag clients whose requests keep expanding.
• Review every new retainer at the ninety day mark, when real delivery data first shows whether the original estimate was right.
• Track cloud compute and data licensing costs by client or engagement instead of lumping them into one infrastructure expense.
• Flag contracts with unclear or missing intellectual property terms before work begins, since ownership is a contract question.