Back to Blog
Amazon Strategy Agency Selection

When to fire your Amazon agency: 7 signs the problem is structural

Naeela March 30, 2026 14 min read
a woman is walking on a treadmill
Table of Contents

Fire your Amazon agency when the problem is no longer one bad month, but a repeated failure of judgment, ownership, or control that survives a written 30-day repair plan. If the same mistakes keep costing profit, cash, inventory, or time, the relationship is not having a rough patch. Its operating model is showing.

A good agency will not get every Amazon decision right. It will make the decision visible, explain the evidence, own the consequence, and adjust when the facts change. If you are paying experts but still have to extract every useful answer with pliers, start looking.

Key Takeaways

  • Two repeated failures inside 60 days justify a formal repair plan. Three or more usually indicate a structural problem.
  • Give repairable reporting, communication, and workflow failures 30 days. Move immediately when access, unauthorized spend, or account health is at risk.
  • A strong agency can name the 3 most important decisions it made in the last 90 days and connect each one to profit, cash, inventory, or risk.
  • A controlled switch has 4 phases: secure control, preserve context, triage the account, then make measured changes.
  • Use the same 7 operating questions with every replacement agency. Chemistry is helpful. Evidence is better.

Is this poor performance or a poor agency?

Bad results do not automatically mean bad management. Bad decisions do.

Amazon performance moves for reasons no agency controls alone. A stockout can flatten organic rank. A price increase can hurt conversion. A review shock can turn yesterday's profitable traffic expensive. Your own team may delay creative, ignore an inventory warning, or withhold the product costs required to calculate contribution.

The agency test is what happens after the miss.

A weak team says the algorithm changed, promises to monitor the account, and brings a dashboard to the next meeting. That explanation offers no decision to approve, challenge, or measure. The algorithm is not an employee you can put on a performance plan.

A strong team separates cause from coincidence. It might say: conversion fell after the price change; click costs held steady; the hero ASIN lost organic share; and keeping the current spend plan will consume cash without protecting contribution. It then recommends a move, names the owner, and states what evidence would prove the diagnosis wrong.

Within 5 business days of a material miss, expect five things: a likely cause, supporting evidence, a clear control boundary, a recommended decision, and a review date. If those are present, repair may be sensible. If every miss produces a new story and no accumulated learning, replacing the agency is reasonable.

What are the 7 signs it is time to replace your Amazon agency?

Do not fire an agency over a personality clash dressed as strategy. Look for observable failures in the account, the meetings, and the economics.

1. The report cannot tell you whether growth made money

The deck is 46 slides. ROAS is wearing a green arrow. Then somebody asks whether the business made more money, and the room develops a sudden interest in slide 47.

Revenue, ACoS, ROAS, and TACoS describe useful parts of the system. None can prove that a sale left enough contribution after product cost, Amazon fees, advertising, returns, discounts, storage, and fulfilment. An agency can improve ACoS while pushing a low-margin SKU or accelerating a hero-product stockout.

Poor looks like this: “Revenue increased 22% and ACoS held at 24%.” There is no product-level margin, inventory consequence, or explanation of what the spend was meant to achieve.

Great looks different: “Revenue increased, but the coupon removed $3.10 per unit from the hero ASIN. We are slowing non-brand spend and moving budget to the product with 11 weeks of stock and healthier contribution.”

Those figures are illustrative. The diagnostic is real. Ask the agency to explain three products: a hero, a marginal SKU, and an inventory-risk SKU. For each, it should name the economic target, advertising role, stock constraint, and next decision.

Repair the gap if your team never supplied reliable cost data. Replace the agency if it has the economics and still treats ad-attributed revenue as the whole business. ROAS is not a business model.

2. The meetings produce activity, not decisions

“We optimized bids.” “We monitored performance.” “We are testing opportunities.” Forty-five minutes pass. Nobody can name what changed.

Activity language sounds responsible because it contains verbs. It is still useless when the verb has no reason, owner, or consequence.

Ask one question: “What changed in the last 14 days, who approved it, and what result did you expect?” A capable team should be able to answer on one screen. It can show that a placement multiplier moved, explain the break-even logic, and give the date when the choice will be reviewed.

Give the agency one chance to maintain a decision log with five fields: date, owner, action, reason, and review date. If important changes still cannot be reconstructed, you do not have a documentation problem. Nobody is truly in charge.

3. Strategy has collapsed into account maintenance

Maintenance keeps campaigns running. Strategy decides where the next dollar and the next hour should go.

Ask for the 3 most important decisions made in the last 90 days. At least one should connect advertising to price, inventory, conversion, creative, catalog structure, or contribution.

Good answers sound like tradeoffs:

  • We stopped pressing the hero ASIN because 19 days of stock made more demand actively harmful.
  • We accepted a temporarily higher ACoS on a launch because the contribution model and inventory position supported the investment.
  • We ended a coupon that produced more orders but less contribution.

These are examples, not client claims. Their value is the shape of the thinking: evidence, choice, cost, and review condition.

Weak answers are task lists. Strong answers include something the team chose not to do. If the agency sold strategic ownership but cannot show a current thesis, a ranked decision queue, or one deliberate refusal, you are paying a strategy fee for account hygiene.

4. Every account-manager change erases your brand's memory

One staffing change happens. Three account managers in 8 months is an operating model introducing itself.

The cost is not another introduction call. The cost is lost context. Your team explains the catalog again. A settled inventory decision reopens. A “new” audit recommends the same fix you paid to implement last spring.

A healthy handoff is almost boring. Within 10 business days, the new lead can explain SKU economics, the commercial calendar, active risks, open cases, inventory assumptions, campaign rationale, and decisions awaiting approval. Your team fills small gaps. It does not retell the brand's life story.

Senior continuity does not mean one person can never leave. It means the agency has a memory stronger than one employee's inbox. If every departure turns your team into unpaid onboarding staff, the service is consuming the capacity it was hired to create.

5. A useful answer requires three follow-ups and a small séance

Not every message is urgent. A listing suppression, unauthorized spend, price error, account-health issue, inventory threat, or missed event deadline is.

Poor communication is not merely slowness. It is missing ownership. A coordinator acknowledges the risk. Two days later, another person requests information already in the thread. Senior help arrives after the damage and calls the situation unfortunate.

Great communication has boring mechanics:

  • material risk receives acknowledgement inside an agreed window;
  • one named person owns it through resolution;
  • the escalation route is known before anybody needs it;
  • strategic questions receive a considered answer on a defined schedule;
  • bad news travels early.

Set three response tiers in writing. A sensible starting point might be 2 business hours for material account risk, 1 business day for a decision blocking active work, and 2 business days for routine analysis. Your coverage model may require different windows. What matters is that urgency, ownership, and escalation are explicit.

Replace the agency when serious risks repeatedly sit unowned or every escalation becomes an org-chart scavenger hunt.

6. Every connected problem belongs to somebody else

The PPC team blames the listing. Creative blames traffic quality. Operations blames the forecast. Everyone may be technically correct, which is a spectacularly useless outcome for the brand.

Amazon is one connected system. Price affects conversion. Conversion affects advertising efficiency and organic rank. Rank affects the demand forecast. The forecast affects inventory. Inventory determines whether generating more demand is valuable or reckless.

A narrow specialist can be the right choice when your internal team owns those connections. The failure begins when an agency accepts broad growth responsibility while behaving like a ticket desk.

Suppose conversion falls after a price increase. A disconnected PPC response cuts bids to protect ACoS, traffic falls, rank softens, and nobody tests whether price was the actual cause. A connected team models the price and contribution tradeoff, adjusts advertising without blindly destroying demand, and puts a decision in front of the person with authority to make it.

Full service is not a longer menu. It is the condition for accountability. Judge a PPC specialist as a PPC specialist. Judge an agency claiming growth ownership by whether it connects advertising, listings, creative, catalog, inventory, pricing, reviews, account health, forecasting, and unit economics.

7. You need the agency's permission to access your own business

This one does not require a 30-day repair plan.

Your company should control the administrator access, billing relationships, data, and source assets required to operate the account. The agency should have permissions appropriate to its work. It should never become the landlord of your operating system.

Audit Seller Central, Advertising Console, Brand Registry, manager accounts, DSP, Amazon Marketing Cloud, reporting tools, API connections, creative source files, and dashboard ownership. Amazon Ads documents viewer, editor, and admin access levels for manager accounts. Admin access can include payment settings, user management, and linking or unlinking advertisers (manager account access, user access).

Great looks unremarkable: company-controlled admins are verified, permissions match the work, exports are available, and offboarding terms were clear before anyone wanted to leave.

If the agency obstructs a reasonable access audit, secure company control, inspect the contract, and get legal advice where ownership is disputed. Do not remove the outgoing team until replacement access has been tested. Control first. Pride later.

Should you repair the relationship before firing the agency?

Repair it when the failure is specific, the agency accepts ownership, and improvement can be proven inside 30 days. Replace it when the same failure has survived several reset conversations.

Failure What a real repair looks like Replace when
Profit-blind reporting Cost data is added and 3 SKU decisions change Revenue and ACoS still govern every recommendation
Slow communication Response tiers, owners, and escalation are followed Urgent risks still bounce between people
Static strategy A 90-day thesis replaces the monthly task list Maintenance returns under fresher wording
Staff turnover The new lead proves context inside 10 business days Your team rebuilds the agency's memory again
Weak results Cause, evidence, next decision, and review date are clear Every miss creates a new excuse and no learning
Fragmented ownership One lead owns cross-functional decisions Every team protects its metric while the business loses
Access confusion Company-controlled access and exports are verified Access or ownership remains obstructed

Keep the repair plan to one page. Choose 3 to 5 outcomes. Give each an owner, proof, and a review date. “Communicate better” is not an outcome. “Acknowledge account-health risks within the agreed window and name the owner” is.

There is an uncomfortable possibility worth naming: the client may be part of the problem. If your team withholds unit economics, ignores inventory warnings, takes 3 weeks to approve creative, and then demands profitable growth on command, a new agency logo will not cure the operating system.

Responsibility runs both ways. The agency owes candor, continuity, ownership, and hard recommendations. The brand owes accurate economics, operational truth, decision access, and willingness to hear an answer it may not enjoy. Without both, accountability is theater.

How do you switch Amazon agencies in 30 days?

Move control and context before changing campaigns. A new agency does not need to prove its energy by rearranging the account on day one.

Days 1 to 5: secure control

Map every user, role, manager account, billing relationship, tool connection, and asset owner. Confirm company-controlled administrator access. Read the notice, renewal, intellectual-property, data-return, confidentiality, and transition terms in the contract. Seek legal advice if ownership is disputed.

Record the baseline: spend, sales, ACoS, TACoS, conversion, contribution by SKU, inventory cover, listing status, open cases, and budgets. Without a baseline, every later argument becomes a matter of memory.

Days 6 to 10: preserve context

Export campaign files, search terms, targets, negatives, budgets, placement settings, listing versions, creative source files, case history, dashboards, and 6 to 12 months of reporting.

For Amazon Marketing Cloud, document the queries, audiences, inputs, and reporting logic. Amazon describes AMC as a privacy-safe clean room that combines pseudonymized Amazon Ads signals with advertiser inputs. The valuable asset is not merely the login. It is the accumulated thinking inside the queries and audiences (Amazon Marketing Cloud).

Days 11 to 20: triage before changing

Sort the account into four groups: preserve, tune, pause, and rebuild. Protect productive exact-match terms, useful negatives, brand-defence logic, and listing work that still supports conversion. Flag work whose purpose cannot be reconstructed.

Ask the incoming agency what it will not touch yet. Restraint is evidence of judgment. A week-one demolition can erase useful history simply to make the new team look busy.

Days 21 to 30: make controlled changes

Fix access gaps, unauthorized spend, broken budgets, duplicate targeting, urgent listing issues, and inventory conflicts first. Sequence deeper changes after the baseline is stable.

Finish with a written 60-day operating plan: commercial thesis, first tests, governing metrics, decision owners, and risks that trigger senior escalation. The transition is complete when control and context have moved. A kickoff call proves only that calendars work.

How should you judge the replacement agency?

Make the pitch survive contact with your actual account.

Ask every finalist the same 7 questions:

  1. Who leads strategy, and who works in the account each week?
  2. What happens to our context when that person leaves or takes vacation?
  3. Which risks receive the fastest response, and who owns escalation?
  4. How will you connect advertising to contribution, cash, and inventory?
  5. Which decisions will you own, and which remain with us?
  6. What are you likely to preserve during the first 30 days?
  7. What data, campaigns, creative files, and access do we retain when we leave?

Give each team the same account sample. Ask for the first 3 decisions, the evidence required, and what it would refuse to change. If every diagnosis ends with a rebuild, a larger budget, and a new dashboard, keep looking. That is a sales script with API access.

For a broader evaluation, use our guide to choosing an Amazon advertising agency. Compare the commercial models in the Amazon agency pricing guide, then define the evidence expected using how an agency should prove savings.

Where does ALFI fit, and where does it not?

ALFI is the right replacement when the problem is fragmented accountability. It is the wrong replacement when you need one narrow task done cheaply.

We work with brands generating $1M+ a year on Amazon, plus $1M+ DTC brands ready to make Amazon a serious second engine. We cap the roster at 18 clients because senior attention is finite. Naeela, our founder and Chief Strategist, stays in strategy, review, accountability, and delivery conversations. A seasoned senior team owns the work. Senior judgment does not disappear after the pitch.

We refuse PPC-only engagements. Advertising cannot be responsibly separated from listings, creative, catalog, inventory, pricing, reviews, account health, forecasting, and unit economics. We will not accept responsibility for profit while controlling one disconnected lever.

That boundary costs us easy revenue. Good. A boundary that costs nothing is usually copywriting.

The standard is demanding on both sides. We owe clients an opinion, a named owner, retained context, early bad news, connected decisions, and the willingness to challenge an economically foolish revenue target. Clients owe us accurate economics, operational truth, the access required to act, and willingness to make hard decisions.

ALFI is unusually right for an established brand whose Amazon operation has outgrown fragmented execution and whose leaders want senior people close to the work. Choose another model if you want a PPC specialist, the lowest-cost task executor, a large rotating bench, or support for an early-stage account. Choose another model if you want validation more than judgment.

If the signs above point to a structural break, book a transition audit with ALFI. Bring your top 10 SKUs, 90 days of advertising data, your access map, and the decision your current agency still cannot explain. We will tell you whether the relationship looks repairable, whether the operating model is broken, and whether ALFI is the right replacement. Those are three different questions.

How long should I give an Amazon agency before firing it?

Use 60 to 90 days to judge an initial performance plan when the account is stable. Communication, ownership, and reporting should improve inside 30 days. Do not wait through an arbitrary trial period if access is obstructed, unauthorized spend continues, or serious account-health risk has no owner.

Will Amazon rankings drop when I switch agencies?

Changing agencies does not automatically change organic rank. Reckless transition work can. Preserve productive campaigns, listing history, inventory context, budgets, negatives, and search-term logic before restructuring. A short overlap can help when practical. Require the incoming team to classify work as preserve, tune, pause, or rebuild.

What data should I request from my old Amazon agency?

Request campaign and search-term exports, targets, negatives, budgets, placement settings, naming logic, listing history, creative source files, Brand Store assets, open cases, dashboards, tool connections, and 6 to 12 months of reports. Include AMC or DSP queries, audiences, and reporting logic where relevant.

Should I hire the replacement before giving notice?

Know who will own the transition before creating a coverage gap, but follow the contract and confidentiality obligations. First verify company-controlled access and preserve required data. Then coordinate notice, overlap, permissions, and billing so the live account is never left without a named owner.

When should I repair the relationship instead of switching?

Repair it when the failure is specific, the agency accepts responsibility, and progress can be verified inside 30 days. A decision log, better escalation, or complete cost data can fix a real gap. Leave when the same failure repeats, ownership remains vague, or explanations never improve.

What should you do this week?

  1. Score the relationship against the 7 signs using account evidence, not mood.
  2. Ask for the 3 most important decisions made in the last 90 days.
  3. Map company and agency access across every Amazon account, tool, and asset.
  4. Put 3 to 5 repair outcomes in writing with owners and a 30-day review date.
  5. Read the contract before giving notice or changing permissions.
  6. Build the preserve, tune, pause, and rebuild list before anybody edits the account.
  7. Compare replacement teams using the same 7 operating questions.

Do not stay because switching feels awkward. Do not leave because one month hurt. Demand the thing you hired an agency to provide: judgment, connected ownership, and decisions your business can afford.

Amazon Strategy Agency Selection