Marketing Agency Software Problems: The 7 That Cost Real Money, and How to Avoid Them

Why does the scope get written as a task manager instead of a contract model?

Agency briefs almost always start with workflow. Projects, tasks, assignees, a board. It is the most visible part of the day, so it is what gets described, and it produces an estimate for something you already own in Asana or ClickUp.

The object your business actually turns on is the retainer, and it exists nowhere except a PDF and an account director's memory. A contract carries contracted hours per role per month, a rate card, rollover rules with expiry, a deliverable list, revision round counts and targets. Time entries have to join to that, not to a project name. Build tasks first and you get a second task manager plus a spreadsheet, which is where you started.

This is specific to agencies because your contracts are the product boundary. A software team's scope creep shows up as a slipped date. Yours shows up as work you did and never invoiced, which never appears as a line item anywhere, so nobody fights for it. The margin leaks silently by design.

The fix is to hand a developer your ugliest statement of work, the one with role level allowances, rollover with a 60 day expiry and a quarterly true up, and ask them to model it on a whiteboard before you sign anything. If they reach for a contract entity with role level allowances, they have built this. If they start drawing tasks and projects, you will spend the engagement teaching them your business. Write the scope from the ops director's month end process backwards, because that process is the requirement.

What goes wrong when you migrate Harvest time entries and Asana projects?

Extraction is not the problem. Both expose interfaces that make pulling records straightforward, and teams therefore assume migration is a scripting afternoon. The real work is mapping historical projects to contract records, and your old project names rarely map cleanly to statements of work.

The specific mess is predictable. One project called Client X 2026 spans two contract versions with different rate cards. A rebrand created three projects for one account. Someone logged six months of a retainer against a project named after a campaign that was actually a separate fixed fee job. Time entries carry a task but no role, so a burn model that needs strategy hours separated from paid media hours has nothing to split on. And a chunk of entries are logged to a catch all called Internal that turns out to contain client work.

The fix is to migrate 12 to 24 months of time entries deliberately, with one person from your operations team owning the reconciliation for a couple of weeks. Map projects to contracts explicitly, assign a role to every historical entry using the person rather than the task where the task is silent, and flag entries you cannot classify rather than guessing. Then validate by rebuilding three months you already reported on and checking the numbers match. If they do not, find out why before go live rather than after.

Why do the ad platform and finance integrations break after launch?

The reporting layer is the part clients see, and it is the part that dies quietly. Ad platform interfaces have rate limits, token expiry and account hierarchy behaviour that naive builds discover in month two. A client with three ad accounts under two business managers breaks a connector written against a single account assumption. Tokens expire on a Saturday and nobody finds out until a Monday client call.

Finance breaks differently. A read only sync to your accounting system is comparatively cheap. Bidirectional sync with classes, multi entity structure and media pass through is where estimates double, and it fails on the edge cases your finance lead already knows about: a client billed from two entities, a credit note that has to reverse a marked up pass through line, and a freelancer invoice that arrives after the client was invoiced.

The fix is to specify the failure behaviour before the happy path. Ask what happens when a token expires at the weekend, and expect an answer that includes retry, alerting to a named person and a visible stale data indicator on the report rather than a silently old number. Ask specifically about a client with multiple ad accounts under multiple business managers, because that client exists in your portfolio. And decide the depth of the finance sync at scoping, since it is the single largest swing in this category.

What happens when approval chain of custody is not covered?

Proofing tools give you annotations and versions. What they cannot give you is the join between a proof round, a contracted round count, a rate card and an invoice line. So round four opens on a three round deliverable and nobody notices, because the rounds are scattered across proofs, email threads and a video link, and nobody wants to be the person raising it with a client mid campaign.

The second half of the problem is evidence. A client marketing lead approves a headline in a direct message to a junior copywriter on a Friday. Six weeks later their legal team disputes it, and your defence is a screenshot.

The fix is an approval record that carries the deliverable identifier, the round number, the approver identity, a timestamp, an immutable snapshot of exactly what was approved, and the contract clause it sits under. When a round opens beyond the contracted count, the system drafts a change order priced from the rate card and routes it to the account director while the work is still fresh and the client still remembers asking. Inbound client email should parse back into the record through a monitored mailbox so clients never have to log into anything, because any process that requires client behaviour change will fail. Classification of inbound messages into feedback, approval or new request is a narrow and genuinely useful automation here, with the triggering sentence quoted back to the account director rather than a bare label.

Should you build custom or configure what you already own?

Plenty of agencies should configure and stop. Under roughly 25 concurrent retainers with fairly uniform contract shapes, Productive.io or Scoro set up properly will beat anything you commission, and the money belongs in people. Kantata is the right shape for larger professional services operations with more conventional contract structures. If your retainers are broadly the same and your differentiator is craft rather than process, buying is the correct answer and building is a distraction dressed as an investment.

Build when the exceptions are the business. The signals: you employ someone whose real job is joining tools together and you have quietly hired a second; your contracts carry role level allowances, rollover with expiry or quarterly true ups that the tools flatten; you can name a margin number lost to unbilled work but cannot prove it account by account; reporting labour has crossed a meaningful monthly total and none of it is billable; or your process is something you sell against, in which case putting it inside someone else's product caps it at their roadmap.

Hitting three of those, the build usually pays back on recovered unbilled scope alone. Hitting one, it does not.

How do hidden costs get into the quote?

Connector count is the first. Every ad and analytics platform is its own integration with its own quirks, and quotes commonly price two and assume the rest are similar. Ask for a per connector rate plus an ongoing maintenance rate, because platforms change and someone has to keep up.

Contract modelling is the second and the most underestimated. If every statement of work is bespoke prose, somebody has to read them and encode the exceptions, and that is interpretation work by a person who understands your commercial terms. It looks like configuration and it is not. If your quote has no line for it, the assumption is that your ops director will do it in evenings.

Third is finance depth, already covered, which can double an estimate on its own. Fourth is migration, which is reconciliation rather than export. Fifth is structure: multiple offices, multiple currencies and multiple billing entities each add real work to every money touching screen.

Sixth is compliance. You hold client ad data, customer lists and sometimes personal data across regions, so you need data residency options, per client access boundaries so a junior on one account cannot query another, audit logs on approvals and exports, and a data processing agreement your clients' procurement teams will sign. Retrofitting tenant boundaries into a system that assumed everyone sees everything is expensive, and enterprise clients will audit you on it.

The fix is to demand the estimate split into engineering, connectors, contract modelling, migration and compliance, each with an owner and a date.

What separates a build that works from one that fails here?

Timesheet compliance decides everything downstream. Burn data, profitability, capacity and change orders all rest on entries being filed, and if people file three days late the alerting is useless. Any build that treats time capture as a form has already lost. The version that works assembles a draft from calendar events, task activity and ad platform change logs and offers it at the end of the day as pre filled entries a person confirms in one tap. That shift is the single largest determinant of whether the rest of the system is worth anything.

The second determinant is where the alert lands. A pacing rule that fires at day 10 when media hours are running ahead of schedule is only useful if it reaches the account director in the place they already work, naming the account, the delta and the specific tasks driving it. A notification inside a system nobody has open is a log entry.

Third is restraint. A partner worth hiring will tell you to keep Slack, keep Google Drive, keep your proofing tool, and spend the budget only on the join those tools cannot make. Anyone quoting a full replacement of your stack is selling scope rather than judgement, and the resulting system gets routed around because it is worse at the things the specialists do well.