Rules 1 to 5
What a deal does to marketing data
Global M&A value rose 40% to $4.9 trillion in 2025, and PwC expects about 42,000 deals in 2026.1 Every one of them lands on a marketing team as a data problem: two sets of ad accounts, two analytics properties, two CRMs, two ways of naming a campaign, and a new owner who wants one report. Harvard Business Review has written that "study after study puts the failure rate of mergers and acquisitions somewhere between 70% and 90%", without naming the studies.2 What is measured is the time: a 2025 survey of 600 finance and IT leaders put the average integration at eight months, with one in five taking more than a year, and named inconsistent financial data as the top challenge.3
We built these rules from three sources: Improvado's own customer and prospect records of companies that went through a deal (anonymised unless improvado.io names the customer), the ad platforms' and CRM vendors' documentation on what they keep and what they let you move, and published studies we checked at the source. The ten rules are ordered by how much each one saves the team in the first 90 days. Where a threshold is our own starting point, we say so.
01 Freeze 36 months of history before anyone touches an account
Platform history has a window, and consolidation closes it. Google Ads keeps hourly, daily and weekly reporting data for 37 months.4 Meta's Insights API returns a maximum of 37 months, and since Jan 12, 2026 its breakdowns of unique-count fields are limited to 13 months.5 LinkedIn keeps ad performance data for ten years but professional-demographics breakdowns for two.6 On a standard GA4 property, event-level data is kept for 2 or 14 months, and moving a property to another account leaves its change history behind; Google names mergers as a reason to move one.7
So the first act after the close is a copy, not a migration: pull 24 to 36 months of daily data from every ad platform, analytics property and CRM on both sides into one warehouse you control. Then the live accounts can be merged, moved or closed without losing the like-for-like comparison the board will ask for next quarter. Our starting point is 36 months, because the two largest ad platforms stop at 37.
02 Inventory every source on both sides, and expect orphans
One sheet, one row per account or property, both sides: ad accounts, analytics, tag managers, CRMs, marketing automation, call tracking, affiliate, offline files. The columns nobody fills in by default are the ones that matter: who holds the login, whether anyone reads the account, and what it spent in the last twelve months. Orphaned and duplicated accounts exist on both sides; the sheet is what makes them visible, and it is the list you retire from in rule 9.
03 Pick one system of record per entity, and write it down
For each of customer, lead, campaign name, spend and conversion, decide which system wins when the two sides disagree, with a date and a name against the decision. Every later mapping refers to this table. Two CRMs are the usual case, and the vendors are clear about what merging means: HubSpot's own guidance says "there is no way to merge all the data between HubSpot accounts. The best way to move assets between accounts is to export".9 The second CRM is a source to export, under rule 1, and a system to retire, under rule 9; it is never a second truth.
04 Map old names to one convention; rename live campaigns last
Each side names campaigns its own way: market first or brand first, underscores or pipes, a "lead" that means a form fill or a sales-accepted record. In a 2024 survey by Claravine and Advertiser Perceptions, 95% of marketers and agencies agreed that data standards are a must; the survey did not ask about mergers, but a merger is where the absence of standards shows up as two unjoinable histories.10
Write one convention with closed lists for every field (market, legacy entity, brand, objective, channel, audience, launch month) and one delimiter. Then build a mapping table: old name, parsed fields, new name, side, date. Reports read the mapping table; live campaigns keep their names until the mapping is proven in a report. Renaming a live campaign moves its UTMs and breaks every filter that used the old name, so rename in waves, outside sales periods, and never rebuild a campaign for the sake of its name. One group of sister companies brought onto one platform told us "they're all calling everything different things" and ran the naming fix as a project in its own right before any shared report was trusted.8
05 One metric dictionary, signed by marketing and finance
"Lead", "MQL", "conversion" and "spend" mean different things in two companies, and most cross-company reporting disputes are about the definition, not the data. One page, one definition per metric, one formula, one feeding source, signed by marketing and finance. Spend gets its own line: gross or net of agency fees, net of platform credits or not, in which currency at which rate. The platforms and the finance ledger will disagree by a few percent forever; the dictionary says which one the board sees.
This is also where the revenue line of the integration plan lives. McKinsey found that almost 70% of the mergers in its 2004 database failed to achieve the revenue synergies expected of them, and in its 2017 survey more than a third of respondents reported missing their revenue goals after a merger.11 A combined pipeline and revenue number that both sides accept is the only way marketing can show its share of that line.
Rules 6 to 10
Rules 6 to 10: from one model to one process
06 Key everything to a permanent id, never to a name or a sheet
Names change in a merger; ids should not. Issue a permanent id per legal entity, brand and location from the finance system, and key every account, property and campaign to it in the warehouse. A crosswalk sheet works until the first typo.
07 Run the rules every day; the acquired side owns its alerts
Four classes of problem break first after a deal: campaign names that do not parse into the convention, live ads with missing or malformed UTMs, accounts pacing outside their monthly plan, and sources that stopped delivering. Write a rule for each, give it a severity, an owner and a channel, and run it daily against the combined data. Route the acquired side's alerts to the acquired side's owner: that is how two teams learn one standard without one of them policing the other. Our starting list is twelve rules; it is in the playbook.
08 Build the new owner's report first; team views on the same model
The board or the fund wants one view of spend, pipeline and revenue by market and channel, comparable across both sides, within the first quarter, and it asks before the data is ready. One agency acquired by a new holding company was told to deliver automated dashboards within six weeks while its team spent three to four hours a week per client on manual reporting; the project went on hold.8 Build the owner's report first, on the metric dictionary and the mapping table, with a legacy-entity field so each side can still be read. Then build team views on the same model, so that a number on a team dashboard always matches the board's number. Add one row the owner reads most carefully: the share of campaigns that parse into the convention, the rules passing, the sources delivering.
09 Do not run two warehouses past day 90
Keep the warehouse that already passes the security audit and the BI tool most people know; move the other side's reports onto the shared model, and rebuild the better ones there rather than keeping a second warehouse alive. Then retire from the inventory of rule 2, in order: accounts with no spend and no owner; the second of two tools doing one job; agency-held accounts whose campaigns have moved; and the second CRM last, after the records are merged under rule 3. This is where the savings appear.
10 If you buy more than once a year, make the plan a standing process
A company that buys several businesses a year cannot run this from scratch each time. Rules 1 to 9 become an onboarding checklist per acquisition with a fixed number of days, a permanent id issued on day one, and a portfolio view above the entity views, built once. The integration survey above found that 86% of firms say integration takes longer than expected;3 a standing process is the only thing that shortens the second one.
From 90 days to the standing process
From 90 days to the standing process
The same ten rules hold for a carve-out, an agency roll-up and a serial acquirer; what changes is the order and the owner. In a carve-out, the freeze covers history inside the former parent's accounts that the division will not be able to reach after the cut-over, and the plan assumes one vendor and a marketing operations owner, because there is usually no data engineering team on the other side of the separation. In an agency roll-up, each client is its own small merger, and an earnout can keep licences and instances apart for a year after the data model is one. For a serial acquirer, the plan is a process with a clock.
We wrote the full plan up as The M&A Marketing Data Playbook: the 90-day calendar week by week, the inventory sheet, the system-of-record table, the naming convention and mapping table, the metric dictionary, the twelve starter rules, and what each platform keeps and for how long.
Sources
Sources
- Bain & Company, "Global M&A poised to sustain momentum in 2026 after great rebound", press release, Jan 27, 2026; PwC, Global M&A Industry Trends, mid-year outlook, Jun 23, 2026. Source PwC ↩
- Harvard Business Review, "The Big Idea: The New M&A Playbook", March 2011. The article gives no source for the 70% to 90% figure. Source ↩
- Unit4 survey of 600 finance and IT leaders at professional services firms, reported by Tech.eu, Apr 24, 2025. Source ↩
- Google Ads Help, "Data retention in Google Ads reporting" (read Oct 26, 2026); Google Ads API, date ranges reference. Source API reference ↩
- Meta for Developers, Graph API reference, Ad Account Insights (read Oct 26, 2026); Meta developer blog, "Ads Insights API metric availability updates", Oct 16, 2025. Source Developer blog ↩
- Microsoft Learn, LinkedIn Marketing API, Ads Reporting (read Oct 26, 2026). Source ↩
- Google Analytics Help, "Data retention" and "Move a property" (read Oct 26, 2026). Source Move a property ↩
- Improvado customer and prospect records, 2023 to 2026; companies anonymised. ↩
- HubSpot Knowledge Base, "Can I merge data between different HubSpot accounts?", updated Sep 25, 2026. Source ↩
- Claravine and Advertiser Perceptions, survey of marketers and agencies, reported by Content Marketing Institute, Oct 18, 2024. Source ↩
- McKinsey, "Where mergers go wrong", 2004; "Focusing your M&A team on revenue growth", 2017. Source 2017 study ↩
- Improvado customer story, ASUS. Source ↩
- Improvado customer story, Eicoff; Improvado, Marketing Data Governance product page. Source Product page ↩
- Improvado customer story, University of San Francisco. Source ↩
- Improvado customer story, SoftwareOne. Source ↩
Thresholds marked "our starting point" are Improvado's working rules. Customer examples are anonymised unless Improvado's published story names the customer. Platform retention windows are as documented on the dates shown and change without notice.