Draft Builder Workflow guide

Build better drafts by batching similar items.

Organize each run around facts the items share, set those facts once, and leave fewer decisions for AI to infer from photos. The result is faster setup, more consistent drafts, and less correction afterward.

Best first defaults Category & condition Best batch Similar items with shared facts Final step Seller review

The rule of a strong run

Use the largest batch for which the same defaults remain accurate.

A larger batch is not better if it forces category, condition, or other shared facts to become vague.

Same category Same condition Shared details
01

Start with sameness

Group items by the decisions they share.

A good batch lets you answer important listing questions once before the run. This is useful beyond Lakeview Lister: AI generally performs more consistently when the task is narrower and fewer facts are ambiguous.

Strong batch

Different items, reliable shared facts

Every item belongs in the same eBay category, uses the same condition, and shares any defaults you plan to apply.

Split this batch

Exceptions keep appearing

If some items need a different category, condition, format, or important listing detail, move them into their own run.

Why this works

Every accurate default removes one decision the model would otherwise have to infer. That reduces variation and lets the AI focus on what is actually different about each item.

02

Use what you know

Set reliable defaults before the run.

Choose only defaults that are true for every item in the batch. Category and condition deserve special attention because they are important to buyers and can be difficult to determine reliably from photos alone.

  1. Set the eBay category when the batch shares one.

    This avoids asking the model to choose among similar category paths for every item.

  2. Set condition when you have established it.

    Use one condition only when it honestly applies to the full batch. Separate exceptions instead of forcing them into the default.

  3. Fill in every other shared setup value you can support.

    Use the pre-run settings for facts and listing preferences that remain accurate across the batch.

Defaults should remove uncertainty, not hide it.

Do not select a convenient condition, item detail, or claim merely to make the batch uniform. When a fact changes, split the affected items into a separate run.

03

Practical examples

Build batches around the collection in front of you.

Cards

Separate by the attributes that change the listing.

Useful batches can share a game or sport, set, year, player, team, card type, holo treatment, or raw-versus-graded state. For example, process holos and reverse holos separately when that distinction matters.

Clothing

Keep the garment type and department consistent.

A run of women’s jeans can share a category and condition while the model identifies brand, style, size, color, and other item-level differences from the photos.

Hats

Use one recognizable product family.

Baseball hats make a useful batch when they share the same category and condition. Team, league, brand, style, and size can still vary by item.

Collectibles & parts

Match product family and completeness.

Group items from the same line, era, scale, or parts category. Keep complete items, loose parts, and differently packaged items separate when those distinctions change the listing.

Use meaningful similarities

Items being stored together does not necessarily make them a good AI batch. Group them by the facts that affect identification and the eBay listing.

04

Make the photos easier to read

Use a repeatable photo sequence.

Consistency helps you spot missing coverage and gives the model the same kinds of evidence for each item.

  • Keep one item—and only that item’s included contents—in each photo group.
  • Start with clear overall views, then add labels, tags, model numbers, card numbers, or other identifiers.
  • Show measurements clearly when size matters, with the item and measuring tool visible together.
  • Photograph wear, damage, missing pieces, and other condition details directly.
  • Use the same general angle order for every item in the batch when practical.
  • Include enough context to distinguish a complete item from a loose part or incomplete set.
More photos are useful only when they add evidence.

Blurry duplicates, unrelated objects, and mixed items can create ambiguity. Prefer a complete, focused photo group over a larger but confusing one.

05

Tell the model what a photo cannot

Add notes for facts only you know.

Photos can document appearance, but they cannot establish every fact a buyer may care about. Add relevant seller knowledge in the group notes instead of expecting the model to guess.

  • Whether the item was tested and what testing actually confirmed.
  • Whether all original parts, accessories, or pieces are included.
  • Known authenticity evidence, provenance, repairs, modifications, or replacement parts.
  • Odors, intermittent behavior, internal damage, or other details that may not appear in a photo.
  • Exact model, variant, material, or history when you have reliable information beyond the images.
Do not let photos stand in for testing.

An AI model cannot know whether an item works or was tested from appearance alone. Claims such as tested, untested, working, complete, or authentic should come from seller-provided facts and evidence.

06

Prove the setup first

Run a small sample, review it, then scale up.

  1. Build a few representative items first.

    Include at least one item that tests the edges of the batch, such as unusual wear or a less obvious variant.

  2. Review the fields that matter most.

    Check identification, title, category, condition, item specifics, description, measurements, included contents, and price.

  3. Adjust the setup or split the batch.

    If the same correction appears repeatedly, fix the shared setup. If only certain items differ, move those exceptions into another run.

  4. Process the rest only after the pattern looks right.

    A short validation run can prevent the same category, condition, or description mistake from repeating across a large batch.

Use identification confidence as review guidance.

Identification confidence is a review signal, not a pass-or-fail status. Review Medium- and Low-confidence identifications carefully against the photos, labels, and reliable references before publishing. Once you have verified the item yourself, there is no warning to clear—the seller’s review resolves the uncertainty.

Verify prices with completed sales.

AI price estimates are starting points, not valuations. They are based mainly on currently listed items, which show asking prices rather than completed sales. Use Lakeview Lister’s sold-listings link before setting a price.

Keep learning

Workflow design matters as much as model choice.

A stronger model cannot replace clear photos, accurate defaults, and seller review. Once your batching process is working, compare the supported providers to choose the quality, speed, and cost balance you prefer.