The Three-Filter Test for Picking Your First Automation

Every owner I talk to has a list of things they want to automate. The list is almost always wrong, not because the tasks aren't real, but because they're ordered by excitement instead of by pain.

Picking your first automation badly costs you twice. You spend time and money building something that doesn't move the needle, and then you write off automation as a thing that didn't work for your business. The real problem was the selection, not the technology.

I built the first meaningful automation at Yala Vanity not because it was clever but because it was the task I genuinely hated most: reconciling six supplier feeds every morning, cross-referencing stock levels, flagging pricing discrepancies, updating the catalog. Three hours, every single day, before I could do anything else. That's where I started. The three filters below are how I found it, and how I find the right starting point for every business I work with now.

Filter One: Does It Happen Often Enough to Matter

Frequency is the multiplier on everything. A task that happens once a month is annoying. The same task happening daily is a tax on your entire operation.

My threshold is weekly at minimum. If something happens at least once a week, the automation pays for itself in a reasonable window. If it happens daily, the payback is usually measured in weeks, not months.

Tasks that don't pass this filter aren't bad to automate eventually. They're just not first. Don't spend your first build on a quarterly report when you have a daily manual task sitting right next to it.

Filter Two: Does It Follow the Same Steps Every Time

Automation is fundamentally rule execution. If the task requires genuine judgment, where a human is reading context and making a call that changes based on nuance, you're not looking at an automation candidate yet. You're looking at a decision process that needs to be documented first.

The honest test is this: could you write out the steps and hand them to a new employee with no prior knowledge, and have them do it correctly? If yes, you can automate it. If no, figure out the steps before you touch a tool.

The supplier feed sync passed this filter cleanly. Fetch the file, parse the columns, compare against the live catalog, flag the deltas, push the updates. Same logic every time. Freight quoting worked the same way: pull the shipment dimensions and destination, hit the LTL carrier rate APIs, return the best quote. No judgment required. Just rules, run fast.

Filter Three: Does a Mistake Actually Cost You Something

This is the filter most people skip, and it's the one that separates the automations that matter from the ones that are just technically interesting.

If the task is high-frequency and rule-based but a mistake is basically invisible, the cost of manual errors is low. The automation is still useful, but it's not urgent. What you want for your first build is a task where manual errors show up as real damage: a customer gets the wrong price, a shipment gets misquoted, a lead goes cold because nobody followed up in time.

The freight quoting at Yala Vanity used to take four days for a complex LTL quote. A single wrong number meant a margin hit or a customer complaint. That cost was concrete and measurable. That's the kind of task that belongs at the top of the list.

What Passes All Three Filters in Most Businesses

When I run the audit on a new business, the same task types show up at the top of the stack over and over. They're not glamorous. That's the point.

• Inventory or catalog sync from a supplier feed or purchase order: happens daily, fully rule-based, mistakes mean wrong pricing or stockouts showing as in-stock
• Lead follow-up sequence: happens every time a new inquiry comes in, same steps every time, a missed follow-up costs a real sale
• Order status or fulfillment notification: triggered by a status change, identical logic for every order, errors damage customer trust at scale
• Quote or estimate generation: inputs go in, math and logic go out, manual errors mean wrong margins or wrong customer expectations
• Internal reporting pulled from multiple sources: daily or weekly, same data every time, manual pulls are just slow and error-prone

None of these are exotic. That's why they're good first builds. The tools are also well-understood: n8n or Make for the workflow logic, Supabase if you need a reliable data layer, Twilio if the process touches SMS, a webhook or API connection to your existing CRM or e-commerce platform. No new software ecosystem to buy into.

The Task That Fails the Test and What to Do With It

Some tasks you want to automate won't pass all three filters right now. That's fine. The answer isn't to force them through. It's to note where they fail and fix that first.

If a task is frequent and costly but the steps aren't consistent, that's a process problem. Document it, standardize it, run it manually with the documented steps for a few weeks, then revisit. The automation will be faster to build and more reliable once the process is clean.

If a task is rule-based and mistakes are costly but it only happens a few times a year, put it on the list for later. Build the high-frequency stuff first and you'll have more capacity, and more confidence in the tooling, by the time you get to it.

One More Thing Before You Build

The first automation worth building post on this blog goes one level deeper on the specific task I'd build first in most e-commerce and service businesses. If you've run the three filters and landed on something in the order or inquiry flow, that post has the specifics.

What I won't do is charge you a monthly retainer to figure out what to build. The scoping process is a free audit call. The build itself is a single fixed price, no surprises. You get the workflow files, the credentials stay in your accounts, and the documentation is written so you or someone on your team can maintain it. The pricing page has the specifics on how that works.

The three-filter test takes about twenty minutes on a call. Most owners come in with a list of ten things they want to automate and leave with one clear first project. That's the whole point.

The hardest part of picking your first automation isn't the technology. It's being honest about which task is actually costing you the most. Run the filters, trust the output, and build the boring one first.