Industry
Twitter vs X user retention comparison for US accounts
Twitter X user retention comparison: how US panels, ad reach and app telemetry count accounts after the July 2023 rebrand, and what none of them can measure.
What to take away
- Three sources answer this question for US accounts: audience panels, X's own ad reach estimates and mobile app telemetry.
- Each counts a different unit. Panels count people, first-party reach counts accounts, telemetry counts installed devices.
- The July 2023 rebrand changed naming, not behavior. A retention drop dated to that month needs a second source before you act.
- No method follows one US user across the rebrand and the later move to x.com without a stable identifier.
What is being compared
The comparison is between three ways of measuring whether US accounts stayed after Twitter became X. Audience panels recruit a standing sample of Americans and project their behavior to the national population. First-party ad reach comes from X's own advertising tools, which report how many accounts or devices an ad could reach. App telemetry comes from mobile measurement vendors that observe installs and sessions.
Twitter was acquired in October 2022, and the name and bird logo were retired in July 2023. Wikipedia's account of Twitter sets out those dates. The rebrand is the event most US analysts want to price, and it is also the worst month to use as a clean baseline.
The criteria that matter
Six criteria separate a usable retention claim from a press number. Each source below fails at least one of them.
| Criterion | Audience panels | First-party ad reach | App telemetry |
|---|---|---|---|
| Unit counted | A person in a recruited sample | An account or device served an ad | An installed app |
| US coverage | National samples, thin state cuts | Self-reported and redefined per quarter | Strong among mobile-heavy groups |
| Follows one user over time | Yes, for the panel's tenure | No | Yes, until the app is deleted |
| Series across July 2023 | Continuous, minus panel turnover | Metric redefinitions | Store and software changes |
| Duplicate and bot handling | Modeled | Limited | Modeled |
| Reporting lag | Weeks | Quarterly or ad hoc | Days to weeks |
The same problem appears elsewhere: an internet communities comparison shows how far the answer moves when the unit of analysis changes.
Option by option
Panel data is the only one of the three that can say how many Americans, rather than how many accounts, were still using X in a month. Some panel designs over-sample heavy users. Its real value here is the unbroken line: the same instrument ran on both sides of July 2023.
First-party reach is free, fast and defined by the seller. It counts reachable accounts, so a dormant login can still appear. Moderation decisions after the acquisition removed or restored large numbers of accounts, and Section 230 of the Communications Decency Act leaves those calls to the platform rather than to a US regulator. Reach moved with them.
Telemetry tracks installs, opens and sessions on devices. It offers the shortest lag and the finest cohort cuts, and it misses desktop and browser use, which still matters for US news and politics audiences.
Whichever source you pick, the series only holds if you do three things.
- Fix the unit first: person, account or installed app.
- Apply that unit on both sides of July 2023, or split the line and label the break.
- Report the rebrand month on its own instead of smoothing it into a trend.
Where each option is the right one
Panels are right when the question is who in America is still there, by age, income and region, projected to the population. That fits a reach plan or a filing a regulator may read. If you plan to split retention by generation, generational digital culture explains why the cohort labels themselves shift.
First-party reach is right when you need a defensible headline number quickly and the buyer and seller are the same party. It suits quarterly market sizing, not cohort work.
Telemetry is right when the question is what retained users do: opens per week, time in app and whether a lapsed user returns. It decides product, not population.
None of this answers how fast a US audience can leave a platform. early social networks statistics gives the historical base rate.
What none of them solve
All three count a unit that is not a person. None follows one US user through the acquisition, the rebrand, the layoffs and the later move to x.com without a stable identifier, and no such identifier is public. That gap is why panel, first-party and telemetry estimates for the same quarter can differ by wide margins. Treat a single figure as illustrative until a second source agrees.
A retention figure with no stated unit and no stated window is an anecdote with a decimal point.
Commentary in this space often outruns its method notes, and critiques of social media is a useful check on claims that arrive without a source.
Common questions
Did US monthly active users fall after the rebrand? Twitter's last quarterly report before the acquisition put monetizable daily active users worldwide just under 240 million, the final comparable disclosure. X has not published a US series since, so any US-only change comes from one of the three sources above.
Is X's ad reach number a retention measure? No. Reach counts accounts an ad could hit inside a window, including accounts that log in rarely. It is a market size input, not a cohort retention rate.
Do the figures include accounts held by children? Accounts held by children under 13 sit outside all three sources, since US law bars covered operators from collecting their data without verifiable parental consent. The FTC rule on COPPA sets out those limits.
What is the fastest honest check on rebrand retention? Pull a panel series that runs through July 2023, hold the unit fixed and split the line at that month rather than smoothing it. If the break is the largest move in the series, say so.
