The content you paid for but never published

The draft was cheap. The abandoned campaign was not.

AI makes it easy to create another version, another channel adaptation, another subject line, another landing page. The request itself can cost pennies. The source work, review queue, correction loop, and missed launch date cost people.

The content you never publish still consumed a budget. If your report counts drafts and ignores abandoned work, it rewards the process that creates the most leftovers.

The useful question is not “How many pieces did AI generate?” It is “How much accepted work did the process produce, and what did the rest cost?”

The leaky content funnel#

Track the same assets through each stage. Here is an illustrative month:

The process generated 100 drafts and published 36 assets. A “cost per draft” report will make that month look productive. A “cost per published asset” report will force the team to ask what happened to the other 64.

Where the money goes when an asset dies#

An asset can fail at several points:

Do not collapse these into one rejection rate. The repair for a canceled campaign is different from the repair for a stale product source. One is planning. The other is a content-quality failure.

Measure waste against accepted output#

Use two measures together:

unused asset rate
= rejected + abandoned assets ÷ submitted assets

waste carried by each published asset
= production cost for rejected and abandoned work
  ÷ published assets

The first shows how often work dies. The second shows what the surviving assets are carrying.

If a process spends $7,200 producing 36 published assets and $3,100 on 31 rejected or abandoned assets, its total production cost is $10,300. The published asset carries $286.11 of production cost before distribution. The $3,100 is not a footnote; it is 30% of the month’s production spend.

A high output count can hide a low survival rate. Volume is not a win when every additional draft adds another reviewer, another source check, and another asset nobody sends.

Give failure a reason#

“Did not use” is not enough. Ask what killed the work:

• Brief changed: the business decision moved before the content was ready.
• Source gap: the process could not support a material claim.
• Product mismatch: the output described an outdated feature or offer.
• Audience mismatch: the content was accurate but wrong for the segment.
• Brand correction: the draft required a full rewrite rather than an edit.
• Legal or policy rejection: the claim needed a different approval path.
• Channel failure: the format, timing, or destination changed.
• Duplicate work: another team had already produced the asset.
• No owner: the content waited until the date passed.

Reason codes turn waste into a repair list. If “source gap” dominates, improve the source set before adding another model. If “no owner” dominates, the process needs a decision owner before it needs more generation capacity.

The abandoned-work report#

Every week, show the work that did not become a customer-facing asset:

The “reusable evidence” field matters. Some rejected assets contain research that can support future work. Preserve the useful source material without pretending the original asset succeeded.

Do not solve waste by removing the gate#

When the rejection rate looks bad, teams often reduce review. That makes the chart better and the customer experience worse.

Review should become narrower and earlier:

1. Check whether the source set supports the material claims.
2. Check product names, dates, prices, and offer terms before drafting variants.
3. Route higher-risk claims to the right approver before the final copy is built.
4. Reject a bad brief before it creates twenty channel versions.
5. Keep one decision owner when comments conflict.

The goal is fewer expensive failures after the team has already invested in the asset. A gate that catches a bad brief in five minutes is cheaper than a legal review after localization, design, and media scheduling.

Find the stage that creates the most waste#

Use a simple stage table for a two-week sample:

The stage with the largest cost is not always the stage with the largest count of failures. Fix the stage that consumes the most money or blocks the most accepted output, then measure again.

A small recovery experiment#

Choose one process with a high abandoned-asset rate. Keep the model and channel mix unchanged for the first week. Record:

1. every submitted asset;
2. every accepted, rejected, and abandoned outcome;
3. the full production cost attached to each outcome;
4. the reason the work stopped;
5. whether its source material can be reused;
6. the date it was meant to reach a customer.

In the second week, change one upstream condition: freeze the source set, name an owner, tighten the brief, or route claims earlier. Compare published assets, cost per published asset, and missed dates. The winning change lowers waste without lowering the approval standard.

Where Deixic fits#

Deixic keeps the activity, tools, source evidence, spend, and approval decision beside the work. That makes rejected and abandoned work visible before it gets flattened into a monthly provider total.

Use the product view to find the process, owner, evidence state, and decision. Use the review-cost guide to measure the human queue. Use the budget guide to give the process a threshold that can stop waste while it is happening.

The system should make it uncomfortable to report a large output count without showing the survival rate.