🚀 How to choose a pricing model?

3 tactics to choose a pricing model

💰 Flat-fee subscription

• Predictable and simple with no surprise bills, but often cap revenue and attract low-margin customers

• Small companies may also get priced out and large companies may underpay

• Find two or more levers to expand your best customers starting with these:

  • Add a premium edition at a higher price point

    • Typically priced at 50-100% more than the base edition, with 15-25% expected uptake (that’s +15% revenue across your customer base)

  • Charge for add-ons that most folks don’t need but some really value 

    • Add-ons are usually priced at 10-30% of the core product price (adoption rate varies significantly)

  • Introduce a price escalator clause 

    • Generally between 5-8% on an annual agreement

    • 1-in-3 B2B contracts include an automatic price increase in their auto-renewals and many large customers expect (and budget for) them 

  • Layer in a fair usage policy for your top 10% of power users

    • 70-80% of AI token consumption comes from just 10% of users

🛠️ Feature-based pricing

• Extremely common for SaaS companies (e.g. Salesforce, Figma, Airtable, Miro, Slack, Notion, etc) to have three paid tiers (also known as Good-Better-Best packages) where the middle tier tends to become the default 

• Tiers may feel arbitrary and confusing, and customers may want to mix-and-match features; new features often break the tier structure

• Make it obvious who each plan is for and why they should buy it:

  • Distill the value proposition of each package down to one sentence or less

  • Launch 1-2 new features per year as standalone add-ons, then bundle these into packages (at higher price points) every 12-18 months to maintain simplicity

  • Shift to progressive feature gating to offer customers a taste of premium features across all plans before having them upgrade

🪑 Seat-based pricing

• Seat pricing remain an important part of the monetization mix even though there is a shift toward more flexible and usage-based pricing models

• It no longer scales the way it used to:

  • AI reduces headcount (less seat revenue over time) 

  • Each user’s value varies wildly, and value from AI agents doesn’t scale with buying more seats

  • Seats put an artificial cap on the revenue of an account

• This model can be exceptionally difficult to abandon but you can redefine it:

  • Add a “lite user” seat for more occasional users 

    • Figma and Tableau does something similar

    • Lite user pricing ranges anywhere from 10-40% the price of a power user 

    • Sell unlimited lite users as a feature to reduce the admin burden of managing multiple license types

  • Require a minimum # of users for certain plans 

    • Canva does this as part of their Team pricing, while communicating it is less expensive per seat compared to Canva Pro

  • Bring in a usage paywall

    • SurveyMonkey did this with survey response limits across paid plans

    • More expensive plans include more survey responses (along with more features), albeit at a higher price per seat

  • Test a usage model as an expansion play

    • Pilot a usage-based model to increase penetration and share-of-wallet with select existing customers without fully abandoning seat-based pricing

    • Use past usage data as a baseline to de-risks the transition for both sides

2 traps to avoid

🚨 Striving for a perfect pricing model

• There is no such thing as a perfect pricing model

• The best you can do is to choose the pricing model that allows you to tell the unique story of what you do, who you’re for, and why you’re better than the alternative

• Have a plan to manage the inevitable downsides

🚨 Switching pricing models too often

• Don’t abandon your broken pricing model, fix it

• When one pricing model doesn’t work, many startups try to rehaul the entire pricing model; that defeats the point because you want to learn from what’s not working and iterate to improve

1 tool to leverage

📖 Best practice on pricing

• Roughly 38% of companies are changing pricing quarterly, 31% semi‑annually, 25% annually

• Find your ideal pricing model using Metronome’s 5-min assessment

Bonus: 1 trend to spark startup ideas

📈 Data centers are becoming a critical bottleneck for the AI boom

• AI demand is straining grid infrastructure with more than 70% of large load interconnection requests coming from data centers

• Large loads Interconnection requests have risen by 227% in the past year, with many individual sites now requesting >1 GW each (the equivalent of adding a small city to the grid)

• Opportunities to increase speed-to-power while easing grid strain:

  • Power–development synchronization

    • Opaque interconnection queues force billion-dollar decisions with limited visibility on where capacity is available

    • Software that maps real transmission capacity, forecasts energization windows, and models flexibility requirements enables hyperscalers, utilities, and developers to plan from a shared source of truth

  • Flexible load orchestration

    • Power providers build enough infrastructure to accommodate peaks that occur only 1-2% of the time; curtailing 0.25% of annual usage can free tens of GW of capacity systemwide 

    • Orchestration platforms that integrate load flexibility into compute scheduling, shift noncritical workloads to low-demand or high-renewable hours, reduce peak strain, accelerate interconnections, and lower the need for new transmission

  • Power-aware compute optimization

    • Most data centers are designed for uptime and not flexibility, with redundant systems, always-on configurations, and conservative operating margins 

    • Power-aware compute orchestrators that maximize throughput per watt and unlock more productive capacity from existing assets by dynamically tuning GPU performance, smoothing load volatility, and coordinating across HVAC, IT, and battery systems