Daily Authentic Users Filter
Separate genuine human adoption from automated activity before judging traction.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 96%
The Daily Authentic Users Filter qualifies activity before treating it as evidence of adoption. Start by defining what genuine human participation means for the specific business: a verified conversation, meaningful creation, completed workflow, considered purchase, or another intentional behavior. Then identify which events can be produced by bots, crawlers, integrations, scripts, or AI agents. Combine behavioral patterns, account evidence, technical signals, and sampled validation to classify active users as authentic, uncertain, or automated. Report daily authentic users alongside conventional daily active users and use the qualified number for retention, product-market fit, community health, and investment decisions. The definition must remain business-specific and evolve as automated systems learn to imitate human activity more convincingly.
Origin
Extracted from Marketing Against The Grain as Erica Wanger distinguished daily active users from daily authentic users in an increasingly automated internet.
Core principles
- 01Activity volume does not prove human demand.
- 02Automated systems increasingly generate traffic and usage events.
- 03Each business needs a context-specific definition of authentic participation.
- 04Human verification should combine behavioral and technical evidence.
- 05Strategic decisions should rely on qualified demand rather than inflated activity.
How to run it
- 1
Define Authentic Use
Specify the intentional human action that demonstrates real product value. Make the definition relevant to the business rather than universally generic.
Pro tip Choose behavior that would be difficult to generate accidentally.
Watch out Account creation or page loading alone rarely proves authentic use.
- 2
Map Automated Activity
List every source of nonhuman events, including crawlers, integrations, monitoring systems, scripts, spam, and AI agents. Mark which analytics they can inflate.
Pro tip Ask engineering, security, marketing, and customer teams for separate inventories.
Watch out Known bots are only part of the automated-traffic population.
- 3
Build Classification Signals
Combine technical indicators with behavioral sequences and account evidence. Assign activity to authentic, uncertain, or automated categories.
Pro tip Use several weak signals together rather than trusting one brittle detector.
Watch out Aggressive filtering can exclude real users and distort conclusions in the opposite direction.
- 4
Validate with Humans
Sample accounts and sessions, contact users where appropriate, and compare classifications with observable evidence. Estimate error rates.
Pro tip Review both false positives and false negatives.
Watch out Do not claim precision the validation process cannot support.
- 5
Report Qualified Metrics
Show daily authentic users beside total active users and explain the classification method. Base strategic conclusions on the qualified cohort.
Pro tip Track retention and referrals specifically among authentic users.
Watch out A large gap between active and authentic users should not be hidden in aggregate reporting.
- 6
Refresh the Filter
Reassess signals as bots, integrations, and customer behavior change. Version definitions so historical comparisons remain interpretable.
Pro tip Schedule reviews after major product or traffic-source changes.
Watch out A static detector will decay as automated behavior evolves.
In the wild
A community platform reports rising daily active users, but many sessions consist of scraping, auto-posting, and API polling. It defines authentic use as a verified person reading a thread and then posting, replying, saving, or messaging through a normal session.
→ The qualified metric reveals slower but more credible human adoption and prevents premature scaling.
An API platform discovers most activity comes from customer automation. Because software-driven calls are the intended value, it does not relabel them as inauthentic; instead, it verifies the paying human or organization behind each integration.
→ The framework is adapted to distinguish legitimate customer automation from bot noise.
Common mistakes
Treating Every Bot as Worthless
Automated activity can be legitimate product usage in machine-to-machine businesses. Authenticity must be defined around genuine customer intent.
Using One Detection Signal
A single IP, timing, or user-agent rule is easy to evade and can misclassify real users. Combine multiple signals.
Hiding the Definition
A proprietary number without a disclosed internal definition cannot support consistent decisions or historical comparison.
Is it for you?
Best for
It is best for internet products whose traffic, content, accounts, or interactions may be generated by software.
Not ideal for
It is not ideal for machine-to-machine products where automated usage is itself the intended customer value.
From the transcript
“the importance of understanding daily active users versus daily authentic users.”
“You're needing to get really clear on who your actual customers are and who the humans are on the other end and who actually it's…”
“How can you tell if there's a human on the other end?”
From the episode
Inside the Hidden Playbooks of Businesses That Actually Win