Working backwards workflow

Campaign Reorganisation Playbook

Taking a fragmented Google Ads account (8–9 campaigns where performance keeps shifting between them) and consolidating it into a cleaner structure with portfolio bid strategies — without breaking learning, losing conversion history, or tanking performance during the migration.

This is a separate workflow, not a quick fix. Plan to spend 2–6 weeks executing it carefully. The cost of a botched restructure is much higher than the cost of leaving the current structure alone for another month.

Reads with: Cheat Sheet — Mistake #3 on portfolio bid strategies and ROAS Calculator — margin tiers (you need your margin tiers defined before you can structure around them).

Why fragmentation hurts you

Symptoms you've already noticed:

↻ Shifting wins
Performance moves between 8–9 campaigns week to week

📊 Patternless
Some campaigns hit target, others don't, no obvious pattern

📈 Jumpy bidding
Smart Bidding feels jumpy — sudden drops and spikes

🎯 OK in aggregate
Account-level fine but individual campaigns volatile

Root causes:

Smart Bidding is data-starved at the campaign level. A campaign with 20 conv/month makes decisions on almost no data.
Self-competition between campaigns. Same product or keyword eligible in multiple campaigns. Performance "shifts" because Google's auction is shifting which campaign wins each impression.
Learning periods never settle — every minor tweak resets learning across yet another campaign.
Budget fragmentation. £6K split across 9 campaigns gives each ~£666/month — below the threshold where Smart Bidding can do its job.

The fix is structural, not tactical. Lower targets, max bid limits, and seasonality adjustments won't solve fragmentation.

Section 1 — Diagnose before you restructure

Pull this data first (last 90 days, per campaign)

Field	Why
Spend	Identify under-funded campaigns
Conversions	Identify those below the Smart Bidding floor (~30/month)
Conversion value	For tROAS-aware grouping
CPA	Identify outliers
ROAS	Identify outliers
Impression share	Identify constrained campaigns
Search Lost IS (budget)	Identify "limited by budget" campaigns
Search Lost IS (rank)	Identify weak-bid campaigns
Average target ROAS over period	What Smart Bidding has actually been chasing
Bid strategy type	What strategy each campaign is on

Run the overlap audit (the most important step)

For each pair of campaigns: is the same product / keyword / audience eligible in both?

Shopping campaigns: are the same SKUs eligible in multiple campaigns? Check inventory filters and product groups
Search campaigns: are the same keywords (or close variants) live in multiple? Check keyword overlap and search-term reports
PMax: are PMax campaigns competing with Shopping for the same products? Check brand exclusions and listing groups

Self-competition is the #1 cause of "performance shifting". Whichever campaign Google picks for each auction wins; the others see degraded numbers. Fix this first.

Build the audit table

Make a spreadsheet with one row per campaign. Mark every campaign with <30 conv/month as "below floor" — these are your prime consolidation candidates. Mark any campaign that overlaps with another as "self-competing" — these are your prime negative-keyword / inventory-fix candidates.

Section 2 — Design the target structure

The goal: fewer, well-structured campaigns + 2–3 portfolio bid strategies.

Grouping principles

	Group by	Don't group by
Use these	• Margin tier (high/medium/low) • Intent stage (brand vs non-brand, prospecting vs retargeting) • Genuine target difference (only split when targets differ by 30%+)	• Product taxonomy (running shoes vs trail shoes — usually same margin, same intent) • Supplier (irrelevant to Smart Bidding) • Internal team ownership • Product launch date (unless deliberately ring-fenced)

Target shape for an ecommerce account

Sanity check

Symptom	What to do
More than 4 portfolios	Reconsider — probably over-fragmented
More than 8 active campaigns	Reconsider unless you have multi-language / multi-region needs
Any campaign with <30 expected conv/month	Consolidate into a sibling or accept it'll underperform

Section 3 — The safe migration sequence

The "don't cock it up" part. The sequence matters.

Rule 0

Don't do it all at once

Every change resets learning periods. Doing 8 things in one weekend means you eat the reset cost 8 times and the account is in chaos for a month. Doing 8 things sequentially over 4–6 weeks means you eat each reset cleanly and can read the data after each step.

Week 1 — start here

Consolidate before restructuring

The cheapest, lowest-risk move: group existing campaigns under a shared portfolio bid strategy without moving any products or keywords. This alone often fixes the fragmentation problem.

Portfolio bid strategies are a wrapper — assigning multiple existing campaigns to one portfolio aggregates conversion data above the Smart Bidding floor, without you touching the underlying campaigns. Try this first. It might be all you need.

Week 2

Preserve conversion history — pause, don't delete

Avoid deleting campaigns. Pause them instead. Conversion history on a deleted campaign is gone forever to Smart Bidding. Conversion history on a paused campaign is preserved.

The same applies to ad groups: pause, don't delete, until you're certain the new structure works.

Week 2–3

Use Drafts & Experiments for risky moves

For anything bigger than "assign existing campaigns to a portfolio", use Drafts & Experiments. Run the new version at 50% budget against the old at 50% for 2–3 weeks before fully migrating. This is the only way to know whether the new structure is actually better or whether you're pattern-matching on noise.

Week 3 — before going live

Set min/max bid limits before you turn anything on

Suggested defaults:

Max CPC: ~2× your current average CPC for that campaign type
Min CPC: leave blank unless you have a specific reason

Without a max bid limit, a portfolio in learning can drift CPCs upward dramatically before you notice. Cap it from day 1.

Week 3 onwards

Expect a 14-day learning phase per change

Anything significant resets learning. Expect 7–14 days of choppy performance after each major change. Do not panic. Do not tinker. Watch projected ROAS, not actual ROAS, during this window.

Week 4

Add cross-campaign negative coverage

After consolidation, add negative keywords or product-group exclusions so each product/keyword lives in exactly one campaign. The actual fix for self-competition.

Search: shared negative keyword lists
Shopping: inventory filters and product group structures so each SKU is eligible in one campaign
PMax: brand exclusions and account-level negative keywords (now up to 10K)

Week 5+

One change per week

After the initial consolidation, make at most one structural change per week. This gives you readable data on whether each change helped, hurt, or did nothing.

Section 4 — Migration template

Fill this in for your account before you touch anything in Google Ads:

#	Current campaign	Spend/mo	Conv/mo	Margin tier	→ Target campaign	→ Portfolio	Step
1	Fill in your account details before starting…
2
3
4
5
6
7
8
9

Step legend:

Week 1 — Group into portfolio (no other changes)
Week 2 — Add cross-campaign negatives
Week 3 — Set portfolio min/max bid limits
Week 4 — Move products between campaigns (if needed)
Week 5 — Adjust tROAS targets per tier
Week 6 — Pause redundant campaigns

Risk legend: Low / Medium / High. Anything High should be done via Drafts & Experiments first.

Section 5 — Rollback plan

Checkpoint rules

Measure at day 7, day 14, day 21 after each change. Don't act before day 7 — anything earlier is noise.

What's reversible vs hard to reverse

Change	Reversible?	Notes
Assigning campaign to portfolio	Yes	Easy
Changing tROAS target	Yes	Easy
Adding negative keywords	Yes	Easy
Setting min/max bid limits	Yes	Easy
Pausing a campaign	Yes	Resume restores history
Moving products between campaigns	Mostly	Conversion history doesn't follow
Deleting a campaign	No	History gone forever — never delete during a migration
Merging ad groups	Mostly no	Ad-group-level history hard to reconstruct

Triggers to roll back

Roll back a change if, after 14 days (not 7):

ROAS is down >20% vs the pre-change baseline AND projected ROAS confirms it (not just lag noise)
Conversion volume is down >30%
The campaign is now hitting the budget cap when it wasn't before
CPA is up >25% with no offsetting conversion lift

If you need to roll back, do it cleanly: revert that one change, leave everything else, wait another 14 days, reassess.

Don't roll back for: Day 1–7 noise (this is normal during learning) · Single-day spikes or dips · A change in projected ROAS that hasn't been confirmed by multiple data points

Section 6 — When NOT to consolidate

Consolidation is the default move, but there are real exceptions.

Always keep separate:

Brand vs non-brand Search — totally different intent, totally different targets
Brand Shopping vs non-brand Shopping — same reason
Different languages — language is a campaign-level setting, can't mix
Different regions with materially different economics (currency, shipping, tax, returns rates)

Probably keep separate:

Margin tiers with >30% ROAS target gap
Truly different product types where intent differs (£15 accessories vs £2,000 furniture)
New product launches you're deliberately ring-fencing
Retargeting vs prospecting in PMax

Don't keep separate just because:

They're "different categories" but same margin and same intent
You feel like you'll lose visibility (you can still segment reporting on a consolidated campaign)
Your client likes the structure (this is a comms problem, not a structure problem)
You've always done it this way

Section 7 — The minimum viable consolidation

If all of the above feels overwhelming, here's the smallest first move that solves 80% of the fragmentation problem with the least risk:

Audit your 8–9 campaigns. Identify which ones share intent (e.g. all your prospecting Shopping campaigns)
Create one portfolio bid strategy per intent group with a tROAS that matches the average of those campaigns' current targets
Assign all the campaigns in that intent group to the portfolio. Don't touch products, keywords, or budgets
Set a max bid limit on the portfolio at ~2× the campaigns' current average CPC
Wait 14 days. Watch projected ROAS, not actual ROAS
Read the data. If volume is up and ROAS is stable or improved, you've solved the fragmentation problem without restructuring anything

That's the smallest version of this playbook. Everything else is refinement.

Section 8 — Common mistakes during a restructure

Doing it all at once. Eat the reset cost once, not 8 times.
Deleting old campaigns. Pause, don't delete. History matters.
Skipping Drafts & Experiments for big moves. You don't know the new structure is better until you've tested it side by side.
Tinkering during the learning period. 14 days is 14 days. Don't intervene because day 3 looked bad.
No max bid limit on a new portfolio. CPC creep happens fast in learning.
Comparing actual ROAS to current target during learning. Use projected ROAS vs average target ROAS.
Moving products between campaigns without exclusion logic. Now they're eligible in two places — same fragmentation problem, new clothes.
Assuming "more granular = better". Smart Bidding wants more data, not less. Granularity is the enemy.
Restructuring during peak season. Don't. Wait for a quiet period.
Not building the audit table first. Without data on each campaign, you're guessing.

Section 9 — Verification

When you've finished the restructure, the account should look like this:

2–3 portfolio bid strategies total
5–7 active campaigns (more is fine if you have multi-language / multi-region needs)
Every active campaign has 30+ conv/month (or shares a portfolio with siblings to get above the floor)
Brand and non-brand are separated at the campaign level
No two campaigns are eligible for the same product/keyword (overlap audit clean)
Margin tiers have separate campaigns if margins vary materially
Each portfolio has a max bid limit set at ~2× average CPC
No tROAS targets at the ad group level (almost always wrong)
No "limited by budget" warnings on any campaign you actually want to scale
Projected ROAS is your default monitoring metric, not actual ROAS

If any of those are still wrong, that's your next change. Make it next week, not today.