What data you should own and be able to take with you

This is the least discussed item in any marketing engagement and the one that decides what happens when it ends. Relationships end for ordinary reasons. The question is whether you leave with your history or start over.

Starting over is not just inconvenient. Every comparison you would want to make, every seasonal index, every baseline, depends on data that only exists if somebody kept it in an account you control.

What exactly should be in my name?

Nine things. Check each one this week.

The domain registration. The single most important item. A domain registered to an agency or a former developer is a business risk of a different order from everything else on this list.

Website files and database access. Hosting account in your name, with credentials you hold. Ask for a full export regardless, and keep a copy.

Google Analytics property. You as an account owner, vendors as users. A property created inside an agency's account can be moved, but only while the relationship is good.

Google Search Console. Verified as a domain property under your ownership. Verification is a DNS record, which means it depends on the domain, which is why the domain sits at the top of this list.

Google Business Profile. Primary ownership in your business account. This is the one that causes the most painful recoveries, because the recovery process runs through Google rather than through the other party. It is also, on the best available evidence, the highest value item on the list. Patient Prism's 2026 Dental Patient Access Report, scoring 11,552,668 calls across 8,280 locations, attributed 90% of traceable patient calls to the Google ecosystem, with the Business Profile alone accounting for 54%. That is one vertical and it is a call intelligence vendor's own customer cohort, so read the shape rather than the decimal. The shape is that the profile is the asset.

Ad accounts. Your billing, your account, agency as a managed user. An agency account containing your campaign history keeps that history when you leave.

Your customer and email list. Exported regularly, held by you. Not living only inside a vendor's platform.

Call recordings and call logs, if you use call tracking. And the tracking numbers themselves, which are portable but only if the account is yours.

Historical exports. The part everybody forgets, and the one that matters most for anything longer than sixteen months.

Why do historical exports matter so much?

Because the platforms will not keep your history for you, and the useful comparisons all need years.

Search Console's performance report holds sixteen months. Analytics retention has a setting that defaults short. Ad platforms archive. A website rebuild removes the pages whose performance you would want to compare against.

So the year over year comparison you want in month eighteen depends on somebody having exported month one, and nobody exports month one unless it is a routine.

Make it a routine. A monthly export of Search Console queries, analytics traffic by channel, ad spend and results, and your transaction summary, dropped into a folder you own. Raw CSVs, not screenshots. It takes fifteen minutes a month.

Data you did not export is the only thing in analytics that cannot be recovered later. Everything else can be rebuilt. History cannot.

What is a three year archive actually worth?

Here is the arithmetic, and you can run it on your own numbers as soon as you have any.

The question. January organic clicks come in at 620. Is that bad?

With the interface alone. In January 2027, a sixteen month window reaches back to September 2025. It contains exactly one prior January. One prior January is not a comparison, it is an anecdote, and the only other thing on screen is December, which in most businesses is a different season entirely.

With three years of exports. Prior Januarys at 700, 660 and 640. The mean is 667 and the spread is 60 clicks. This year's 620 sits 47 clicks below the mean, which is 7%derived low and inside the historical spread. The answer is that January is normal and nothing needs doing.

Without the archive, the same owner compares 620 to December's 810, reads a 23%derived collapse, and spends a week and a retainer conversation on a seasonal pattern.

Price both sides. The archive costs fifteen minutes a month. Three years of it is nine hours. The alternative is that the most common question in small business marketing, is this number bad, stays unanswerable for the first two years of every relationship you have.

That archive is what makes every honest measurement practice possible. Seasonal indexes, baselines, normal ranges, and any assessment of which metrics actually predict revenue in your business all require more history than any single platform will hand you.

What happens to evidence you do not hold a copy of?

It rots, and the marketing industry runs on the wreckage.

Take the statistic every phone answering vendor quotes: that 85% of callers whose calls go unanswered will not call back. It is usually credited to PATLive, which credits a 2014 Forbes column. That URL is dead, with no Internet Archive capture of it. Elsewhere the same figure is credited to Numa's proprietary data, to "Forbes and BIA Kelsey," and to AnswerConnect. Those attributions are mutually incompatible, which is the diagnostic. There is no underlying study, and there is now no way to check whether there ever was one.

Its companion is the same story. The claim that 80% of callers sent to voicemail leave no message traces to a 2014 Forbes contributor column, self published rather than Forbes editorial research, which stated it flatly with no source. That URL returns a 404 now too.

The speed to lead numbers went the same way. The famous five minute figure comes from a 2007 conference deck of proprietary vendor data, never peer reviewed, and the domain that hosted it, leadresponsemanagement.org, now carries paid outbound links to online casino review sites.

The instructive part is who repeats them. CallRail published a benchmark report in January 2025 built on 1.1 million de-identified conversations, with real missed call rates by industry: 32% in health care, 28% in legal, 14% in home services, 9% in real estate. The same company's press materials also repeat the 85% folklore figure. A business sitting on more than a million real leads still reached for the unverifiable number, because the unverifiable number is more dramatic.

The lesson for your accounts is exact. If the only copy of your evidence lives on somebody else's server, it is one redesign, one acquisition, or one lapsed renewal from being a claim you can no longer support. Hold the raw files.

What should I ask a vendor before we start?

Four questions, asked before the first invoice, when the answers are easy to get.

Whose accounts will these be? The correct answer is yours, with them added. Any other answer needs a reason you find convincing.

What happens to the data if we part ways? Get it in writing. Access removal is normal. Data deletion is not.

Will you give me a monthly export in raw format? Not a PDF report. The underlying data.

What are you building that I will still have afterwards? Pages on your site, profiles in your name, and a customer list you hold are assets. A dashboard inside their platform is a rental.

None of these questions is aggressive and a good vendor will answer them in a sentence each. The reaction to being asked tells you more than the answers do.

If a vendor holds attribution data you cannot see, you also cannot check their work, which quietly makes any conversation about which channels are working a matter of trust rather than evidence.

What is genuinely not portable?

Some things do not transfer and it is worth knowing which, so you do not fight for them.

Platform specific machine learning. The optimisation an ad platform has learned about your account stays with the account, which is one more reason the account should be yours. That history is also becoming more valuable rather than less: Sterling Sky's State of Local SEO in 2026 reports local pack ads growing from around 1% of tracked mobile reports in early 2025 to around 22% by December 2025, and Local Services Ads from around 11% to 31% of tracked queries. An ad account with two years of learning in it is a harder thing to rebuild than it was.

Third party tool history. Rank tracking, heatmaps, and session recording data usually live inside the vendor's product. If you want session recordings of the first thirty seconds as evidence, export the clips that matter rather than assuming the library will still be there.

Reviews, which belong to the platform and the reviewer, not to you or your agency.

And relationships with your customers, which are portable only in the sense that you hold their contact details. That is why the email list belongs on the ownership list rather than in someone else's tool.

Restaurant operators recognise the pattern. Finding four points of labor variance is only possible if you hold the scheduling and payroll history yourself. The analysis is easy. Having the data is the hard part, and it is decided long before anyone tries to run it.

What to do this week

Log in to each of the nine items and check who the owner is. Not who has access, who owns it. This takes about forty minutes.

Fix anything registered to a personal email, a former employee, or a vendor. Fix the domain first.

Create one folder for exports. Set a recurring monthly reminder to fill it.

Export everything available today, dated. Search Console at its full sixteen months, analytics at whatever retention allows, your ad history, your customer list, and your Google Business Profile performance data.

Then decide which numbers you want that archive to answer. If you have not settled the handful you look at every week, the archive will be complete and unread.

Questions about what you own?

Email me at eric@seod.com with a list of the accounts and tools somebody else currently manages for you. Names of the tools is enough. I will mark which ones must be in your name, which are fine as shared access, and exactly where to click on each platform to check who the owner is.

I answer these myself. This is the email I most wish more people sent before signing something rather than after.

More on measurement sits in the analytics library.