First, date the drop and check the numbers are comparable
Write the handover timeline down before looking at any report: the date access was granted, the date the new agency first changed anything, the dates campaigns were paused, rebuilt or relaunched, and the date the old agency lost access. Then plot the main result by day or week across the whole period. A drop that starts on one of those dates points at what changed that day; a drop that started before the handover was already under way and belongs to another diagnosis.
Next, confirm that "before" and "after" are measured the same way. Ask which conversion actions each agency reported, from which system, with which window and which filters. New agencies often bring their own report, and a report built on different definitions produces a different number from the same business. If the old figure counted every enquiry and the new one counts only qualified leads, part of the drop is a change of rule, not of results.
Finally, compare the business's own outcome, orders from the order system or opportunities from the CRM, across the same dates. If the business outcome held steady and only the agency's report fell, start with measurement. If both fell together on the same date, start with the account itself.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| New agency is working in new accounts; history, audiences or pixel data are missing | Who owns the ad accounts, analytics property and tags, and whether the old ones are still live | Accounts, pixels or history lost in the move | Regain ownership and move the work back into the business's own accounts |
| Bid strategies show a learning status; results volatile since the rebuild date | Creation dates of campaigns and strategies against the drop date | Rebuild reset automated bidding's learning | Stop further rebuilding; let learning finish; phase remaining changes |
| Conversions fell sharply on one date while orders did not | Conversion actions, primary flags and tag status before and after the handover | Conversion actions or tags removed or changed | Restore the agreed actions and tags; verify against orders |
| Cost per result rose; search terms or placements look new and irrelevant | Negative lists, exclusions and audience lists against the old setup | Negatives, exclusions and audiences not carried over | Reapply lists and exclusions; review what is now being bought |
| Agency report fell but the business outcome held | Definitions, filters and windows of the old report against the new | Reporting definitions changed | Restate both periods under one definition |
| New team tests ideas the old team already tried and rejected | Whether a record of tests and decisions exists and was handed over | What worked was never written down | Reconstruct the test history; write procedures for the core tasks |
Fix, verify and prevent each cause
Accounts, pixels and history left behind
If the old agency held the ad accounts, the analytics property or the tag container in its own name, the new agency may have started from empty accounts. Everything accumulated in the old ones is then gone: conversion history, remarketing lists, lookalike seeds, quality history and the pixel's record of past visitors. The contract and ownership guidance in selecting, briefing and reviewing agencies explains why the business should own these from the start.
Fix: ask the old agency to transfer ownership of the original accounts, property and container under the handover terms, and move the new agency's work into them rather than continuing in fresh ones.
Verify: the business holds administrator access to every account, and historical data is visible in the accounts the new agency uses.
Prevent: before any new partner starts, confirm that every account they will use is owned by the business and that they are added as users, never as owners. For Google Ads, the guide to manager accounts and user access sets out how access levels, manager account links and a quarterly access review keep that arrangement in place.
Rebuilt campaigns resetting automated bidding
Smart Bidding sets bids from predictions built on the conversions it has seen; Google's Smart Bidding documentation describes how it uses conversion data and auction-time signals. The help centre also documents a learning period after a strategy is created or significantly changed, during which results fluctuate. A new agency that rebuilds every campaign at once puts the whole account into that state together. The Smart Bidding strategies guide explains which changes trigger learning.
Fix: stop any further structural changes, leave targets and budgets steady while the status shows learning, and phase remaining rebuilds one campaign group at a time.
Verify: the status moves out of learning and results steady across comparable weeks.
Prevent: require a phased rebuild plan in the new agency's first proposal, with a stated period and measure for each phase. If the drop followed a single strategy switch rather than a rebuild, the results dropped after a bid strategy change page covers that case in detail.
Conversion actions or tags changed in the handover
New teams often tidy measurement: they remove actions they do not recognise, change which actions are primary, install their own tags or replace a tag manager container. Each change can alter what bidding optimises toward and what the report counts. Google documents that conversion tracking depends on a tag or import recording the action, so a removed tag silently stops the count.
Fix: compare the current conversion actions, primary flags and tags with the list in the tracking plan and restore anything changed without agreement.
Verify: each primary Key EventA GA4 event marked as important to the business, the property's measure of success. records once per real action, and the count tracks orders or leads in the business's own system.
Prevent: put measurement under change control in the new scope: no action, tag or container change without a recorded approval.
Negatives, exclusions and audiences left out
Years of negative keywords, placement exclusions, customer list exclusions and audience settings protect an account from buying the wrong traffic. They live in shared lists and settings that a rebuild easily misses. The Google Ads audit tutorial lists where each sits.
Fix: export the old negative lists, exclusions and audience applications, compare them with the current account, and reapply what is missing.
Verify: the search terms and placements report for the following weeks no longer shows the themes the old lists excluded, and cost per result moves back toward its previous level.
Prevent: make exported lists and settings part of the handover pack, checked by the business before the old agency's access ends.
Reporting definitions changed
If the new agency counts a different action, uses a different window, filters differently or adds its fee to cost, its figures will not match the old agency's even when the business has not changed. The metrics layer and definitions guide explains how silent definition changes create apparent drops.
Fix: restate the months before and after the handover under a single, written definition, and agree that definition in the scope.
Verify: the restated series shows whether a real drop remains, and both the business and the agency quote the same figure.
Prevent: keep metric definitions owned by the business, versioned, and handed to every new partner on day one.
Lessons that left with the old team
An old agency's most valuable asset is often what it learned: which audiences, offers, messages and structures had failed. If none of it was written down, the new team will rediscover it the expensive way.
Fix: ask the old agency for its test log and recent decisions as part of the exit, interview whoever in-house worked with them, and record the results. Then write procedures for the recurring tasks, following writing a marketing SOP.
Verify: the new agency's test plan cites the record and excludes ideas already tried.
Prevent: require every partner to maintain a shared test log and procedure library in the business's own workspace.
A prevention routine for agency handovers
Most of these causes can be prevented in the weeks before a handover, not repaired after it. Before notice is given, list everything the current setup depends on: accounts and owners, conversion actions and their settings, tags and the container version, negative lists, exclusions, audiences, report definitions, the test log and the recurring procedures. Confirm that the business owns every account and holds a copy of every list. The agency offboarding checklist puts the outgoing side of this into a list to tick from notice to the last day of access. Agree with the incoming agency a written measure (KPIA metric tied directly to an objective.) and its definition, a phased plan for any rebuild, and a rule that measurement changes need approval. The agency onboarding checklist covers these agreements, and the baseline, for the incoming side. The analytics property needs the same care, and the GA4 property handover checklist checks its access, key events, links and exports before the old agency's logins are removed. Run the two agencies in parallel for a short overlap where the contract allows, so questions can be answered by the people who know. In the first month after the switch, review weekly: the business outcome beside the agency's figure, the bid strategy statuses, the conversion action list against the tracking plan and the search terms report against the old exclusions. Keep a dated log of every change the new team makes, so a drop can be traced to a date and a change rather than argued about. When the first full month is closed, compare it with the same month a year earlier under the agreed definition, and only then judge the new partner. If the drop persists after all of this, the agency reporting page helps separate a measurement gap from a performance gap. Where the programme is coming in-house or passing to a new manager rather than another agency, the tutorial on taking over an existing marketing programme runs the same checks as a full inventory that ends in a written baseline. Where the change also moves booked television, press, poster or other direct media, the tutorial on handing live media bookings to a new agency or an in-house team lists each booking with its notice date and settles the credits and rebates owed each way, which the checks above leave to it.
A dental group changes its paid search agency
A group of 12 dental practices moved its paid search to a new agency in March. Booked appointments from search fell from around 410 a month to 260 in April. The timeline showed the new agency had rebuilt all 18 campaigns on its second day, in a new account it had opened, because the old agency owned the original. The new account had no negative keyword lists, so roughly £3,500 of the month's spend went on searches for jobs and dental courses. Its conversion action counted calls of 30 seconds or more, while the old one counted bookings in the practice system. Restated on bookings, April's real figure was 300, not 260. The group recovered the original account, reapplied the negative lists, restored the booking action as primary and phased the rebuild over six weeks. By June bookings were back to 400.
Common mistakes
Judging the new agency on a report built under different definitions. Letting a new team start in accounts it owns. Rebuilding everything in the first week. Removing conversion actions nobody can explain without asking what they were for. Ending the old agency's access before the lists, logs and exports have been checked.
