Hire a freelancer for judgment work that changes weekly: creative direction, launch sequencing, supplier negotiation. Plug in a tool for rule work that repeats daily: bid adjustments, keyword harvesting, inventory alerts. If you cannot write the rule down, no tool will run it, and no cheap freelancer will either.
The short version
- Write the task down as a rule. If it survives being written as a rule, software should do it.
- Judgment is not automatable and is not cheap. Paying freelancer rates for judgment is the best value in this industry.
- Tools do not have accountability. Nobody gets fired for a bad automated bid, so someone still has to read the output.
- Sequence matters. Fix the listing, then automate the ads. Automating traffic to a page that does not convert scales a loss.
- The benchmark to hold anyone to is the outcome. A majority of the brands we manage are profitable within their first year. Ask for the equivalent.
Start with the sort, not the shopping
Before you buy anything, list every recurring Amazon task in your business and sort each one into two columns: rule work and judgment work.
Rule work has a written trigger and a written response. "If a search term has spent twice the target cost per acquisition with no conversion, negate it." That is a rule. Software runs it perfectly, forever, at three in the morning, and it does not get bored.
Judgment work cannot be written that way. "Is this the right primary image for a category where the top three sellers all show the product in use?" No rule produces that answer. A person with taste and category experience does.
Most sellers get this backwards. They hire a cheap person to do rule work by hand, which produces slow, inconsistent execution, and then they buy software to make decisions that require judgment, which produces confident nonsense.
The sequence, with a gate at each stage
Work through these in order. Do not fund the next stage until the current gate is passed.
- Stage one: fix the page. Primary image, title, price, first three bullets. Gate: conversion rate is at or near what your category supports. Nothing downstream works until this passes, and no tool fixes it.
- Stage two: write the rules. Document your bid rules, negation rules, restock triggers and pricing floors on one page. Gate: someone who does not know your business could execute the page without asking a question.
- Stage three: automate the rules. Now buy software, and buy it against the page you just wrote rather than against a feature list. Gate: the tool is doing what the page says, verified by a human reading its output weekly.
- Stage four: buy judgment. A freelancer or an agency for creative direction, launch sequencing, market selection and supplier work. Gate: they can tell you what they would stop doing, not only what they would add.
- Stage five: review the whole system quarterly. Rules go stale as categories change. Gate: at least one rule is retired or rewritten each quarter, or nobody is really looking.
Where each option wins
| Task | Tool | Freelancer | Why |
|---|---|---|---|
| Bid adjustments at scale | Yes | No | Volume and consistency beat human attention |
| Search term harvesting | Yes | Reviewed by a human | Software finds them, judgment decides which to promote |
| Primary image direction | No | Yes | Click-through rate is a taste and category problem |
| Copy in a second language | No | Yes, a native speaker | Machine translation ranks for phrases nobody types |
| Restock alerts | Yes | No | A calculation, not an opinion |
| Deciding to discontinue a product | No | Yes, with written criteria | A tool has no view on your working capital |
| Supplier negotiation | No | Yes | Relationships and leverage, neither of which is a feature |
| Competitor review analysis | Partly | Yes | Software collects, a person reads for the gap worth attacking |
What most agencies will not tell you
The freelancer market and the software market both sell the same fantasy, which is that Amazon can be run cheaply by delegation. It can be run cheaply, but only if you personally hold the strategy. The moment you delegate strategy to the cheapest available option, you have bought execution of a plan nobody wrote.
The second thing: automation software is priced as a subscription, which means the vendor's revenue depends on you keeping it, not on you needing it. A tool that has been running unchanged for a year against rules you have never revisited is a line item, not a system.
The third: a freelancer will rarely tell you the job is too big for one person. Amazon is a wide surface. Sourcing, listing, ads, creative, compliance, logistics and customer service are different skills, and one contractor at twenty hours a month is choosing which of them to neglect. Ask which ones, and decide whether you can live with the answer.
Related answers
- Alternatives to hiring an Amazon agency
- What do I need to run Amazon ads in house
- How to choose between Amazon agency and automation software
- Best budget option for small Amazon sellers
- Amazon agency pricing and economics: the complete guide
If you want a second opinion on which half of your stack is actually working, ask Flapen.
