Long Term Growth Impact Assessment

The Long-Term Growth Impact Assessment is a framework for ensuring growth experiments drive actual business value (like GMV or Gross Profit) rather than just "pulling forward" results or optimizing local conversion ratios. By using long-term holdouts and absolute metrics, you can identify which "wins" are illusory and which "neutral" experiments are actually long-term powerhouses.

Core Principles

1. Optimize for Absolute Volume, Not Ratios

Avoid using conversion rates as a primary North Star. To increase a conversion rate, a team can implicitly make the previous step harder (e.g., adding friction to sign-up to ensure only "high-intent" users reach the next step). This makes the rate look better but reduces the total number of successful users.

• The Rule: Measure the absolute number of users who reach a "success" milestone.
• The Mindset: It is often better to have a lower conversion rate on a much larger pool of users than a high rate on a tiny, restricted pool.

2. Implement the "Long-Term Ping" Holdout

Don't stop measuring an experiment once it reaches statistical significance.

• The Mechanism: When an experiment concludes, ship the winner to 100% of users, but maintain the metadata tags for the original treatment and control cohorts.
• The Cadence: Set automated reporting "pings" to re-evaluate the cohort's performance at 3, 6, 12, and 24 months.
• The Metric: Measure downstream business value (e.g., Total Revenue, GMV, or Net Profit) rather than the local metric the experiment was designed to move.

3. Account for the "Pull-Forward" Effect

Recognize that 30-40% of short-term growth wins are neutral after one year. Often, an experiment simply accelerates an action a user would have taken anyway. If the long-term holdout shows the control group eventually "catches up" to the treatment group, the experiment had no incremental value.

Implementation Workflow

Step 1: Design for Absolute Success

When setting up an experiment, define success as: [Absolute Number of Users] x [Long-term Value Metric].

• Bad Goal: "Increase the sign-up to activated conversion rate by 5%."
• Good Goal: "Increase the absolute number of merchants reaching $1k in GMV within their first 90 days."

Step 2: Identify and Reduce "Monetary Friction"

Look for friction points related to trial length, price points, or credits.

• The Hypothesis: Lowering monetary friction early (e.g., extending a trial or offering a discount) may look like it's "lowering quality" in the short term, but it provides the time necessary for a user to become successful.
• Verification: Use the long-term holdout to see if these "lower quality" users eventually convert into high-value customers.

Step 3: Use Taste for "Neutral-Positive" Decisions

If an experiment's short-term results are neutral but it aligns with your long-term product vision (e.g., a better UI or a more modern technical architecture):

• Action: Ship the change based on "Taste."
• Logic: If the data can't distinguish between the two, choose the version that represents the best possible product for the next 100 years.

Examples

Example 1: The Dunning/Payment Failure Notification

• Context: A growth team wants to reduce churn by sending aggressive email reminders when a credit card payment fails.
• Short-term Result: Significant lift in "recovered" payments. The team claims a major win.
• Long-term Assessment (12 months): The holdout shows no incremental lift in GMV.
• Insight: Users who let their cards fail were already uncommitted; the reminders only "pulled forward" a few weeks of subscription fee before they eventually quit. The team stopped focusing on this area to chase "bigger fish."

Example 2: Pre-configured Onboarding Templates

• Context: Providing pre-filled sections (e.g., image banners, product collages) for new users building a store.
• Short-term Result: Neutral. No increase in the number of people who paid for a subscription.
• Long-term Assessment (6 months): Massive impact on the number of people producing sales/GMV.
• Insight: The templates didn't convince people to buy the software, but they helped those who did buy to build better stores. These stores converted more visitors, giving the owners early momentum and keeping them in business longer.

Common Pitfalls

• The Rate Trap: Making a sign-up flow harder to "clean up" the leads for a sales team. This almost always results in fewer total customers.
• Ignoring the "How": Focusing only on "what" to build (the feature) and ignoring the technical architecture. Architecture determines your long-term ability to pivot and adapt.
• Siloed Attribution: In hybrid funnels (self-serve + sales), failing to credit a marketing ad for a user who signed up self-serve but eventually closed via a sales rep. This leads to under-investing in high-value channels.
• Performance Review Bias: Rewarding teams for 5% lifts that disappear after 6 months. Use the 12-month "ping" to retroactively validate team impact.