The first Photoshop action I ever built was embarrassingly simple. I was 26, working in a commercial studio in Chicago, and I had just spent 45 minutes manually running the same three-step sharpening sequence on a set of product images. Same settings, same order, same result, every single time. When a senior retoucher walked past and asked why I wasn’t using an action, I didn’t have a good answer. I built my first one that afternoon. I have not manually repeated a task like that since.

That was fifteen years ago. I now run my own post-production consultancy, and one of the first things I notice when I start working with a new client, whether that’s an ad agency or an e-commerce brand processing 800 SKUs a week, is how they’re using presets. Or more accurately, how they’re misusing them.

The Problem With Downloaded Presets Out of the Box

Most preset packs are built to look impressive in a demo. They’re optimized for the screenshot, not for your camera profile, your monitor calibration, your client’s deliverable specs, or your color space. A Lightroom preset built for a Canon 5D shooting in golden hour looks nothing like what you need when you’re correcting a Sony A7R image shot under fluorescent warehouse lighting for a product catalog.

The same logic applies to Photoshop action sets. Generic packs assume a baseline that rarely matches yours. They use absolute pixel values for things like Gaussian Blur radii or canvas sizes, which breaks immediately the moment your document resolution differs from whatever the action author was working at. A sharpening action built at 72 PPI will over-sharpen at 300 PPI. A vignette sized for a 2000-pixel wide image will barely register on a 6000-pixel RAW file.

This isn’t a complaint about preset creators. It’s a structural limitation of selling generalized tools to a diverse audience.

What “Custom” Actually Means at the Parameter Level

When I talk about customizing presets, I mean something more specific than just sliding a few sliders around. I mean building your presets around fixed reference points that match your actual production environment.

Start with your output profile. If you’re delivering sRGB JPEGs at 144 PPI for web, every action in your set should be built, tested, and recorded in a document at that spec. If you’re delivering TIFF files at 300 PPI for print, that’s a completely different document. I keep two separate action sets labeled exactly that way: “WEB-sRGB” and “PRINT-300.” Not exciting names, but I’ve never opened the wrong one by accident.

For sharpening specifically, I use percentage-based thinking rather than fixed pixel radii. In Photoshop’s Smart Sharpen dialog, a Radius of 0.8px at 300 PPI gives you a very different result than 0.8px at 72 PPI, but a High Pass sharpening layer blended at 30% opacity scales reasonably well across resolutions because it’s relative to the existing pixel data. That single change, switching from absolute to relative methods, is the kind of thing that makes an action reusable across a project rather than something you have to rebuild from scratch every time the specs change.

Building a Baseline Preset Set That Actually Travels

Here’s the framework I use with new clients. Before recording a single action, I document three things: the source file format and color profile, the final deliverable spec, and the most common manual steps that happen between those two points. That third list is your action roadmap.

For an e-commerce client I brought onboard last year, that middle list had eleven steps: background neutralization, color temperature correction against a specific gray card value, shadow recovery at roughly 15-20%, dust spot removal via content-aware fill on a frequency-separated layer, output sharpening, resize to 2000x2000px at 96 PPI, convert to sRGB, flatten, save as JPEG at quality 10, log filename to a CSV via a simple script, close. I recorded that as a single action, tested it on 20 images pulled from different shoot conditions, adjusted the color correction step to use a relative luminosity blend rather than a fixed curve point, and then ran it against a batch of 500 images.

That batch took about 40 minutes of machine time while I worked on something else. The same 500 images, done manually, would have been the better part of two days.

When Presets Break and What That Tells You

I want to be honest about failure cases, because that’s where you actually learn how these tools work.

The most common reason a custom action breaks mid-batch is a conditional it can’t resolve. If your action includes a step that depends on a layer being named something specific, and one file in the batch has a different layer structure, the action stops or skips the step silently. Silent failures are the worst kind, because you might not notice until you’ve already delivered the files.

My fix is to build validation steps into any action that’s going to run unattended. Before the main sequence starts, I include a step that checks canvas size against an expected value using a simple conditional action, available under Insert Conditional in the Actions panel. If the document is not 300 PPI, the action stops and flags the file name. It adds maybe 15 seconds to my setup time and has saved me from delivering wrong-spec files more than once.

I also keep a log. Yes, I have a spreadsheet. I track every action set, when it was last tested, what document spec it was built for, and an estimated time-per-image figure. That spreadsheet currently shows just over 2,400 hours saved since I started keeping records in 2017. I am aware that this is an unusual hobby. My kids think my job is boring right up until I show them a before-and-after of a car ad composite, which tends to change their minds for about four minutes.

The One Setting Most People Skip

In Edit > Purge, there’s an option to purge the history states during a batch run. Photoshop holds every history state in RAM, and on a long batch, that memory overhead accumulates. Enabling “Purge All” as part of your batch action, or simply setting History States to 5 instead of the default 50 under Preferences > Performance, can cut processing time by 15 to 20% on large batches with no effect on your output.

The best preset you’ll ever use is one you built yourself, tuned to your files, your output specs, and your clients. Everything else is a starting point.