Backed by Y CombinatorThree documents and three sheets that gate every new self-serve component on real request volume and template fit, not a guess.
Free download · No account needed
What you get
Pack of 6 files: 3 documents, 3 spreadsheets
Marketing Design System TemplateTemplate Register · six-month request audit
| Asset type | 6-month requests | Meets volume | Gate verdict |
|---|---|---|---|
| Local event flyer | 51 | Yes | Fails template-fit |
| Instagram class-schedule graphic | 46 | Yes | Promoted |
Volume alone never promotes a component. A type only earns one once its own past requests converge on a shape one template can actually cover.
Hawthorne Fitness's two-person marketing team logs every request from its 42 gym locations: a schedule graphic for one location, a referral card for another, a trainer spotlight, a flyer for a one-off community event. A six-month audit sorted 220 logged requests into eleven specific asset types. The highest-volume type by far, the local event flyer at 51 requests, still did not earn a self-serve component, because its requests never converged on one shareable layout.
Four other types cleared both conditions a component actually needs: at least fifteen requests logged in six months, and a structure consistent enough that one template with fill-in fields covers most of them. Instagram class-schedule graphics, referral cards, trainer spotlights, and new class-launch announcements, 127 of the 220 requests combined, each earned a full specification in Component Guidance. Three months after launch, those four types generated 64 real requests; 58 were self-served directly by location managers, and only six still reached the design team, a 90.6 percent drop from what the prior six-month rate predicted.
Xanadu's own analysis of enterprise design queues puts the templatable share of a typical queue at 60 to 80 percent once a team actually audits it by request type, close to Hawthorne's own 58 percent. The gap between that advice and this pack is what happens after the audit. Every candidate here gets gated on volume and fit together, on a fixed schedule rather than once, and anything that fails either half keeps escalating to a real creative brief instead of a template forced onto it.
Template RegisterHawthorne Fitness, six-month audit, 220 requests across 42 gym locations sorted into 11 asset types.
| Asset type | Requests | Meets volume | Gate verdict |
|---|---|---|---|
| Local event flyer | 51 | Yes | Fails fit, escalates |
| Instagram class-schedule graphic | 46 | Yes | Promoted |
| Referral card | 34 | Yes | Promoted |
| Trainer spotlight social post | 28 | Yes | Promoted |
| New class-launch announcement | 19 | Yes | Promoted |
| Email header banner | 12 | No | Monitor |
| Custom co-branded partner asset | 9 | No | Always escalates |
| New-member welcome one-pager | 8 | No | Monitor |
| Facility maintenance notice flyer | 6 | No | Monitor |
| Price increase notice flyer | 4 | No | Always escalates |
| Grand-opening signage | 3 | No | Monitor |
The highest-volume type in the whole audit still fails the gate: its 51 requests never converged on one shareable layout, so it escalates every time regardless of volume.
Component GuidanceOne full specification of four active components. Instagram class-schedule graphic shown.
| Field | Specification |
|---|---|
| Canvas | 1080 x 1080px, square |
| Logo | Primary mark, bottom-right corner, 32px clear space |
| Background | One of two approved brand gradients only |
| Class table | Name, instructor, and time; six classes maximum |
| Overflow | A seventh class escalates rather than shrinking the type |
A location manager can tell from this alone whether a request fits, without submitting it first to find out the hard way.
Volume AvoidedThree months after the four components launched.
| Component | Self-served | Still to designer | Drop vs. baseline |
|---|---|---|---|
| Instagram class-schedule graphic | 21 | 2 | 91.3% |
| Referral card | 16 | 1 | 94.1% |
| Trainer spotlight social post | 13 | 2 | 85.7% |
| New class-launch announcement | 8 | 1 | 89.5% |
| Total | 58 | 6 | 90.6% |
64 real requests came in for these four types; 58 were self-served and only 6 still reached the design team.
Every logged request classified by specific asset type over a six-month window, checked against a fixed volume threshold before template fit even gets considered.
Full specification for each active component: canvas size, fixed versus editable fields, and what it explicitly does not cover, so a brand guide's locked elements stay locked.
Real mistakes from the first weeks of self-serve use next to the correct version, for every active component.
The specific triggers, pricing, legal claims, co-branding, or a request outside the register, that route straight to the design team regardless of volume.
How often each active component actually gets used, by month, so one sitting unused shows up as clearly as one in heavy rotation.
Baseline demand against what actually reached the design team, filed once finished in an asset library, so the drop is a measured number, not a felt impression.
A ticket queue export, a shared inbox, or a rough list from memory of what location or department managers ask for most, over at least three months.
Each one is classified into a specific asset type, then checked against real six-month volume and whether its instances actually share a structure.
A type that passes both conditions gets a full Component Guidance specification; anything that fails either one, including a high-volume type with no shared shape, escalates by name.
Usage Tracking and Request Volume Avoided keep measuring each active component against the real designer request volume it was actually built to remove.
Yes, the whole pack downloads free, six files total, with no signup or card needed to see it. Edit with AI installs it as a private workspace and swaps Hawthorne's worked example for your own request history. More packs live in the template library.
Three Word documents and three CSV sheets in one zip. The documents open in Word, Pages, or Google Docs, and the sheets open as real spreadsheets in Excel, Numbers, or Google Sheets, so request counts and gate verdicts stay usable data instead of a static page.
Those platforms solve turning an approved template into a locked, fillable form, and they solve it well. None of them decide which asset type earns a template in the first place. This pack automates that decision from a team's own request history instead of leaving it to whoever's judgment happens to be in the room.
Send whatever exists, a ticket export, a shared inbox, even a rough list from memory of what gets asked for most. If nothing is logged at all, this pack starts tracking every request by type going forward, so the next audit has real data instead of a guess.
Only when it clears two conditions together: enough recurring volume to be worth templatizing, checked against a fixed threshold, and a structure consistent enough across those requests that one component with fill-in fields actually covers most of them. The highest-volume type in the worked example still fails on the second condition.
Every active component's specification assumes the current identity: logo file, color values, approved gradients. Once a rebrand rolls out across every surface, each one needs its Component Guidance re-specified against the new identity before location managers keep using it, the same audit-and-gate process applied to what already exists.
Written by
Co-founder & CEO, River
Chandler has spent the last six years working in AI, building tools for millions of users and Fortune 100 companies. In 2025, he co-founded River to build a new kind of operating system for AI agents.
Send your request history. River audits it by asset type, gates each one on real volume and template fit, and tracks how much designer time the ones that clear it actually save.
Audit my requests