framework · prioritization
GEM framework: Gibson Biddle's prioritization model explained
Best for: Company-level priority alignment, stage-appropriate roadmap weighting, interview answers on prioritization philosophy
GEM is not a scoring rubric. It is a forcing function. Gibson Biddle (VP of Product at Netflix, CPO at Chegg) built it to answer a question that scoring frameworks can’t: when Growth, Engagement, and Monetization all matter, which one leads right now? The output is an explicit rank order: 1st, 2nd, 3rd. The point of the constraint is that you cannot claim all three matter equally, because they don’t, and the forced trade-off surfaces the misalignment that has been hiding inside your org.
The three levers and their proxy metrics
Grow tracks year-over-year user or member growth rate. It is the acquisition and retention signal rolled into a single number that shows whether the business is expanding its base.
Engage is a product-specific retention proxy treated as a signal of product quality. The right metric here depends on the product. Daily active users is a weak proxy for most products. For a consumer subscription, monthly cancellation rate is more honest. For a B2B automation tool, the number of automations created captures value delivered better than login frequency. Zapier, for example, measures success by Zaps created, not by how often users open the dashboard. For B2B SaaS, Net Revenue Retention often outperforms DAU as the engagement proxy because it captures expansion and satisfaction from the people who decide whether to keep paying. You have to derive your proxy from the job the product does, not from what is easy to measure.
Monetize tracks LTV and gross margin. Not revenue. Unit economics. The distinction matters when you’re debating whether to run a price experiment or subsidize growth.
The Netflix case and why the numbers matter
In 2005 Netflix had roughly 2 million members, 30% year-over-year growth, and a 4.5% monthly cancellation rate. Biddle force-ranked the priorities as: Monetization first, Engagement second, Growth third. The rationale was that the business needed to prove its unit economics before it could justify scaling. So the team ran ads, sold previously viewed DVDs, and ran price experiments. They were not trying to grow fastest. They were proving the margin story.
By 2008, the priorities flipped to: Growth first, Engagement second, Monetization third. The target was 20 million subscribers by 2010. The rationale: the margin case was already proven to investors, and the business now needed to demonstrate the long-term profitability case through scale. The same company, same product category, different stage. Different rank order.
This is the whole framework in two data points. GEM is not a property of your product. It is a read on where you are in the arc from proving viability to proving scale.
Biddle recommends reassessing the rank order roughly every six months, not treating it as permanently fixed. For AI products, a trigger-based reassessment may be more useful than a calendar-based one: when gross margin math changes by 10x because inference costs drop, or when a distribution channel opens that changes your growth assumptions, the rank order should update.
Where GEM sits in Biddle’s system
GEM answers “what are the company’s top priorities?” before DHM (Delight, Hard to copy, Margin-enhancing) answers “how do we win on each?” GEM sits upstream of roadmap swimlanes, which divide the backlog by strategic theme. The dependency is: set the rank order with GEM, generate strategy hypotheses with DHM, then organize execution around swimlanes. Candidates who cite GEM without knowing this hierarchy tend to sound like they read a summary. Candidates who can trace the dependency sound like they’ve used it.
Scope: company level, not feature level
GEM is a leadership and company-level alignment tool. Applying it at the feature level is the most common misuse in interviews. “I ranked this feature M first because it has direct revenue impact” is not a GEM answer. The framework is about aligning the entire org around what the company is optimizing for at this moment. Individual features serve the priority, they don’t each have their own G/E/M ranking.
How to pick your own proxy metric
Most pages on GEM just use the Netflix cancellation rate example and stop there. That is not a useful model for your own product. The method:
- For G: use the rate that captures new user acquisition plus recovery from churn. If your cohort renewal rate is stronger than gross new signups as a company-level signal, use that.
- For E: ask what the user does when the product is working at its best. Measure that behavior, not sessions or logins. The activity should be something a user would miss if the product disappeared.
- For M: use the unit economics metric your CFO uses when deciding whether to invest in the next growth push. Often that is gross margin per account or LTV:CAC ratio, not top-line revenue.
The 2026 reframe
When feasibility was a real constraint, Engage often meant “do users come back because the product is hard to replace technically?” That moat is gone. Any feature can be shipped quickly with AI-assisted development. Engagement now has to mean: does the product deliver enough genuine value that users return by choice, not by lock-in? Lovability, not usability, is the new engagement signal. A technically flawless AI feature that interrupts workflow or forces unnecessary confirmations will kill retention just as a buggy one would.
For AI products specifically, the G lever may need a new proxy. Growth in 2026 often comes through agent distribution: your product being invoked by other AI workflows, embedded in copilots, or exposed via MCP to orchestrators. API call volume or agent invocations may be more accurate signals of growth than new account signups.
And the cadence question is live: AI inference costs can drop 10x in a single quarter. A Monetize-first priority set under one cost structure may be wrong three months later. Build a trigger condition into your GEM cadence, not just a calendar.
Use it in an interview
strong
"I use Gibson Biddle's GEM model. The key mechanic is a forced rank: you pick 1st, 2nd, 3rd across Growth, Engagement, and Monetization. You can't score them all equally, because that hides where the org actually disagrees. At Netflix in 2005, Biddle force-ranked Monetization first when they had 2 million members and needed to prove unit economics. By 2008, that flipped to Growth first because the margin story was proven and the goal was 20 million subscribers. For each lever you need one proxy metric, not a generic one: Engage might be monthly cancellation rate for a consumer subscription, or Net Revenue Retention for B2B, or automations created for a workflow tool. For this role, I'd want to understand where the company sits on that arc before defaulting to any particular rank."
weak
"GEM stands for Grow, Engage, Monetize. You figure out which one matters most for your product and focus on that." This fails three ways: it skips the force-rank mechanic, which is the entire value of the framework; it says nothing about proxy metrics, leaving the framework abstract; and it implies GEM replaces thinking about the product rather than aligning the org around what matters at this stage. Interviewers who know the framework will know you read a summary.
The strongest version connects GEM to the specific company’s stage, names one proxy metric per lever, and uses the Netflix case with real numbers. It also acknowledges where GEM sits in the broader system so the interviewer sees systems thinking, not a recited definition.