We reject top-down percentage targets because they ignore the non-negotiable costs of system stability. Instead, we implement a Floor and Ceiling model. We first lock the minimum headcount required for security and SLAs as a fixed floor, then treat all remaining capacity as a competitive auction between high-growth bets and core product iterations.
Why the 70/30 innovation ratio fails in practice
Arbitrary percentages treat maintenance as discretionary, leading to shadow work that consumes 40% to 60% of R&D capacity. When we tell a department they have a 30% maintenance budget, but the reality of their 10-year-old flagship product requires 50% just to meet security patches and SLA compliance, the work does not disappear. It simply goes underground.
Engineering teams frequently re-label technical debt or bug fixes as new features to satisfy these top-down mandates. This practice corrupts portfolio data. When a COO looks at a dashboard showing 70% innovation, they make strategic bets based on a lie. In reality, the team is likely spending half their time on non-negotiable stability work while the roadmap for new initiatives stalls.
Fixed ratios fail to account for the increasing complexity of a growing product portfolio. As we scale to 20 or 30 concurrent initiatives, the cumulative weight of legacy code and infrastructure creates a drag coefficient that a static 70/30 split cannot solve. For mature products, security patches and compliance alone often consume the entire allocated maintenance budget, leaving zero room for actual system improvements.
What is the ideal ratio for innovation vs maintenance?
We define the Maintenance Floor as the absolute headcount needed to prevent system degradation. In established firms, this typically ranges from 25% to 40% of total engineering resources. The ideal ratio is not a static industry myth but the delta between your fixed floor and your total engineering headcount.
Platform or Site Reliability Engineering (SRE) teams must own this floor. When maintenance costs are distributed across feature teams, they become invisible. By centralizing the floor, we gain a clear view of the "tax" we pay to stay in business. We revisit this floor every six months to account for product sunsets and new launches entering the support phase.
How do you categorize tech debt in capacity planning?
We separate Keep the Lights On (KTLO) work from Architectural Debt. KTLO is the non-negotiable floor: server patches, security vulnerabilities, and basic uptime. Architectural Debt is different. This is the structural mess that blocks new feature velocity.
While KTLO is part of the Floor, Architectural Debt must compete in the Ceiling auction against new revenue initiatives. This distinction prevents engineers from using "tech debt" as a catch-all for low-impact refactoring. We track debt-to-feature ratios to identify when a product's Floor is rising too fast. If the floor creeps toward 50%, the product is no longer an asset; it is a liability.
When should we deprioritize incremental improvements for new bets?
Remaining capacity enters a quarterly auction. Heads of New Products and Directors of Core Product pitch for resources from the same pool. We force a direct tradeoff: does a 5% retention gain in a core product outweigh a 10% chance of capturing a new market segment with a high-growth MVP?
The COO or Head of R&D makes the final call based on enterprise value rather than historical team size. This model eliminates the entitled capacity mindset where teams expect to keep their headcount regardless of output. If a core product’s incremental gains cannot beat the projected value of a new initiative, the core product loses headcount.
How do you defend innovation budgets during high-support periods?
By ring-fencing the Floor, we prevent support spikes from bleeding into the Ceiling budget. When a major bug or outage occurs, the Floor team handles it. If maintenance needs exceed 40% of total capacity, we do not cut new bets. Instead, we trigger an immediate architectural review.
We use the auction results to create a contract for the quarter. This makes it harder for stakeholders to pull resources mid-cycle for "urgent" requests that did not pass the auction. Visibility into the Floor allows us to show the board exactly how much innovation we sacrifice for stability.
What happens when maintenance exceeds 40% of total capacity?
Crossing the 40% threshold indicates the product is becoming a legacy liability. At this point, we shift focus from feature delivery to Floor Reduction projects. This might include automated testing, infrastructure migration, or decommissioning underused modules.
Failure to address a rising Floor leads to a death spiral where innovation drops to zero and top talent exits due to frustration. We regularly evaluate product sunsets for initiatives where the maintenance floor cost exceeds the annual recurring revenue. If a product costs more to keep alive than it generates in value, we kill it to reclaim capacity for the Ceiling.
The Strategic Capacity Playbook
| Step | Action | Ownership | | :--- | :--- | :--- | | 1 | Audit the last two quarters to identify true "shadow" maintenance costs across all 20+ initiatives. | Program Leads | | 2 | Appoint a Platform Lead to define and own the Maintenance Floor headcount for the next six months. | Head of R&D | | 3 | Standardize the auction template: every request must state Expected Value (EV) and probability of success. | Strategy/Ops | | 4 | Run the first auction to reallocate "Ceiling" capacity based on current strategic priorities. | COO & Finance VP | | 5 | Publish the new allocation map to the entire R&D org to end the cycle of re-labeling work. | Head of R&D |
Honest Tradeoff
Traditional percentage-based models allow for more emergent discovery. By strictly separating maintenance from innovation, we risk disincentivizing engineers from finding creative product solutions while performing routine system upgrades. Our Floor and Ceiling approach requires two to three weeks of upfront analysis from program leads and finance to accurately cost the maintenance floor, which is a significant initial investment compared to the speed of simply picking a 70/30 ratio.
In one breath
Stop using arbitrary 70/30 ratios that encourage engineers to hide maintenance work as new features. Lock a non-negotiable Maintenance Floor for stability and force all other initiatives to compete in a quarterly value-based auction. This ensures that every engineer not dedicated to uptime is focused on the highest possible enterprise value.