Stack Surgery: A Keep-Cut-Trial Audit for Your Dimensional Ad Toolkit

Every studio I've walked through this year has the same artifact buried in its billing history: thirty line items, and no single person who can name the owner of all of them.

A texture library nobody has logged into since spring. Two seats of a real-time renderer for a team of one. Cloud credits still burning on a freelance motion designer who wrapped in March. A CDN contract sized for a campaign that ended before Labor Day.

None of that is a crisis on its own. It's entropy, and entropy has a monthly price. When I asked a dozen US shops — Brooklyn to Austin to Portland — to guess their 3D tooling spend before pulling the actual invoice, nearly everyone underestimated. The waste is rarely dramatic. It's the long tail of small renewals nobody questions.

So here's the audit I'd actually run, instead of another "best tools" listicle. It takes about a week, needs three people in the room, and usually frees up enough budget to put a real colorist or a second animator on the next job.

Four lenses, one scorecard

Score every tool from one to five on four axes. Not list price — the axes that actually predict whether a tool pays for itself.

Cost. Effective cost per delivered asset, not per seat. Divide annual spend by the number of finished deliverables that touched the tool. A cheap tool used once is expensive. An expensive tool used daily is a bargain.

Speed. Count the handoffs it eliminates. Time a real project step rather than trusting the vendor's benchmark reel — a 40-second preview export that saves an hour of ping-pong with a client is worth more than a renderer that's theoretically faster on a machine nobody owns.

Accessibility. Can it export a static fallback frame, captions, or alt text without a workaround? Does its web output respect a visitor's motion preferences? That lens used to be optional. It isn't anymore, and it's cheaper to solve at the tool layer than in a panic two days before launch.

Privacy readiness. Where does the data live, who can see uploaded client assets, and can you produce a data processing agreement on request? US state privacy laws keep widening, and your client's legal team will eventually ask. Having an answer is a sales advantage, not just a compliance chore.

Anything scoring below twelve out of twenty earns a conversation. Nothing gets cut on score alone — but nothing gets renewed by default either.

What earns a keep

A tool survives the audit when most of these are true:

• Two or more people open it in an average week.
• Removing it would add at least a day to a typical project.
• It owns a pipeline step nothing else covers — a genuine gap, not a preference.
• It exports in formats your client's ad ops team accepts without conversion.
• Somebody will defend it in writing, by name, with a reason.

That last one matters more than it sounds. The odd, ugly tool that saves your topology on every hard-surface job deserves protection. Protection just has to be stated out loud, once a quarter, or it becomes the first thing cut by someone who wasn't there when it mattered.

What gets cut

• Single-seat logins with no activity in 90 days.
• Tools that overlap 80% with something you already pay for. You do not need three UV painters.
• Licenses bought for a pitch that was never used in delivery.
• Contracts priced against a campaign that has already shipped.

Cut in that order. Cancellations early in the week, before anyone gets sentimental about a renewal date.

What's worth a trial in 2026

Three categories are pulling their weight right now, and each deserves a 30-day test with a written kill date:

Browser-based compositing and preview. Getting a client a reviewable link instead of a 400MB file has cut entire rounds of feedback for teams I've talked to. Test it on one live project, not a sandbox.

Real-time path tracing with clean web export. The pitch is speed; the real payoff is iteration count. Pick one product shot and see whether you get three extra lighting passes in the same budget.

Privacy-respecting measurement for interactive units. If your dimensional banner reports engagement through a third-party script that collects everything, an alternative is worth a look — especially for clients in regulated categories who ask pointed questions about what the unit collects in the background.

Structure every trial the same way: one project, one owner, one metric, one date when it dies if the metric isn't met. Trials that drift become subscriptions.

The one-week sprint

Monday: export twelve months of billing and tag an owner to every line.

Tuesday: pull usage data — logins, exports, minutes rendered.

Wednesday: score everything against the four lenses. This is the longest day and the most useful one.

Thursday: fifteen-minute conversations with the people who actually touch each tool. Ask what they'd miss and what they quietly route around. The workarounds tell you more than the feature lists.

Friday: decisions. Renewals, cancellations, trial scoping, and a short memo that lives somewhere findable so you're not re-litigating this in February.

The quiet wins

The spreadsheet is satisfying, but the side effects are the real prize. A consolidated stack means one place where accessibility export options live, one vendor risk surface to review, and one set of file conventions the whole team can memorize. Fewer tools, less translation loss, better work.

Do it once a quarter. It's less painful than a surprise renewal, and considerably cheaper.