Fewer Surprises, Better Deals - Demystifying P50, P90, P99 for Real Assets in Africa

Executive summary

• P50, P90, and P99 (P-values) are not magic numbers. They are simple ways to describe how likely different energy outcomes are.

• P50 means “there is a 50% chance the asset will beat this number in a given period.” Think of it as the median case.

• P90 means “there is a 90% chance the asset will beat this number.” It is a conservative case.

• P99 is very conservative. If the deal still works there, your downside is well protected.

• Lenders often size debt using P90 (one-year) to keep repayments safe. Equity cases often use P50 (multi-year) to reflect expected performance.

• The value is not the letters and digits. The value is what assumptions sit behind them and how they are used to make debt, equity and operations decisions.

• When used well, P-values reduce surprises. When abused, they become a fig leaf for weak modelling.

A worked example (illustrative numbers, rounded)

Asset: 10 MWp solar + 12 MWh battery
Offtaker: regional hospital group (escrow in place)
Tariff: availability retainer + variable energy, CPI-linked
Debt: seven-year senior loan, DSCR covenant 1.25x
Equity: UK investors

Energy distribution (annual):

• P50 = 21.0 GWh

• P75 = 20.3 GWh

• P90 = 19.8 GWh

• P99 = 19.0 GWh

Revenue at current tariff (simplified):

• P50 revenue = £2.52m

• P90 revenue = £2.38m

Debt sizing on P90 one-year:

• Required debt service = £1.60m

• Cash at P90 after O&M = £2.00m (DSCR 1.25x)

• Cash at P50 after O&M = £2.14m (DSCR 1.34x)

Equity view:

• Base case IRR at P50 = 14.1%

• Downside IRR at P90 = 11.9%

• Severe downside at P99 (with weather + minor curtailment) = 10.7%

What this tells us

• Debt is safe at P90 one year.

• Equity still works at P90 and even at P99.

• If your mandate needs a floor, document it against P90 or P95.

P50, P90, P99: what they mean, how they move your returns, and when to push back

Read time: about 10 minutes

Who this is for: our investor and allocator community deciding whether to back solar and battery assets serving hospitals, schools, cold chains and SMEs across Sub-Saharan Africa.

A simple story to set the scene

We stand on the roof of a district hospital. The diesel tank is half full. The theatre needs stable power. The CFO wants a PPA that is fair and predictable. The board asks one question:

“How certain are you about the energy we will receive each year”

We answer with three numbers: P50, P90, and P99. Then we explain them in plain English.

• P50: in an average year, we expect about X units.

• P90: in a poor year, we still expect at least Y units.

• P99: in a very poor year, at least Z units.

When used well, P-values reduce surprises. When abused, they become a fig leaf for weak modelling.

Where did these numbers come from?

They come from exceedance probability in statistics. We start with a distribution of possible annual energy outcomes for the project. That distribution is built from:

• long-term solar resource (satellite and ground data)

• year-to-year weather swings

• system losses (soiling, shading, temperature)

• availability (inverters, BMS, breakers)

• curtailment and grid limits

• performance ratio and degradation over time

• uncertainty in the measurements themselves

We run many scenarios (often thousands) - this is a Monte Carlo simulation. We then read off the levels that 50%, 90%, or 99% of the simulated years exceed.

• If P90 equals 19.8 GWh, that means in 9 years out of 10, we expect the plant to deliver at least 19.8 GWh (and often more).

• If P50 equals 21.0 GWh, half the years will be above that, half below.

That is it. No smoke. No mirrors.

Why investors should care (and what each P-value is for)

Debt sizing

• Lenders want high confidence that cash will cover repayments.

• Common practice: size senior debt to P90 (one-year) so even in a weak year, the DSCR stays above covenant (for example, 1.20x or 1.30x).

• Result: fewer default scares, smoother project life.

Equity underwriting

• Equity wants the expected outcome, not the worst one.

• Base IRR is usually built on P50 (long-term) with clear views on upside and downside cases.

Reserves and buffers

• Operational & Maintenance (O&M) reserve, inverter spares, and liquidity facilities can be tied to P95-P99 downside views.

• If the portfolio still distributes under P99, you own something robust.

Operations and bonuses

• O&M teams can be incentivised to beat P50 (weather-adjusted) through better uptime, cleaning, and fast repairs.

• If they beat P50 consistently, equity smiles.

Exit valuation

• A strong actuals vs P-values track record is exit gold. Buyers pay more for predictable assets.

The most common traps (and how to avoid them)

1). Mixing one-year and multi-year P-values

• A P90 one-year is more conservative than a P90 ten-year average.

• If someone uses a one-year P90 to predict ten years of cash flow, push back.

2). Confusing P10 and P90

• P10 (sometimes shown in wind reports) is the optimistic high-energy case.

• P90 is the conservative low-energy case. Do not swap them by accident.

3). Ignoring correlation in a portfolio

• Ten sites in one region will “swing together” in the same weather.

• Portfolio P90 is not the simple sum of each site’s P90. We must model correlation.

4). Treating P-values as fixed truths

• They are probabilities built on inputs and assumptions.

• Add new data (SCADA, met masts, dust seasons) and update the distribution.

5). Using P50 energy with P90 price

• If you pair a high-energy case with a low tariff case to be “conservative”, document it. Better, run a full matrix of energy and price outcomes.

6). Ignoring curtailment and grid trips

• Sunny resource is useless if the feeder trips daily.

• Curtailment risk belongs inside the P-value, not as an afterthought.

How we actually build P-values (the ingredients)

Resource and weather:

• Merge long-term satellite irradiance with any ground sensors and reanalysis data.

• Quantify interannual variability (how “bouncy” years are).

System losses:

• Soiling (by season), shading, temperature, wiring, inverter efficiency, clipping, battery round-trip loss, and availability

Operational behaviour:

• Cleaning schedules, first-fix times, spares on site, and grid outage patterns.

Commercial realities:

• Curtailment probability, PPA caps and floors, escalation, and offtaker credit limits.

Uncertainty and bias:

• Every input has error bars. We carry them into the simulation.

• We avoid “best case stacking” by using distributions, not point guesses.

The output is a curve. P50, P90 and P99 are just waypoints on that curve.

What has changed in the last few years (and why this matters now)

• Data got better. Satellite products improved. SCADA is richer. Soiling and temperature models are more site-specific.

• Operations matter more. With batteries and smart tariffs, when energy arrives is as important as how much arrives. This pushes us to run hourly distributions, not just annual totals.

• Climate variability is louder. Dry and wet seasons are shifting in some regions. We now test trend scenarios so we do not anchor on yesterday’s weather forever.

• Investors ask sharper questions. Committees now want to see P-values next to IRR, DSCR, and reserve sizing on one page, not buried in an appendix.

When should we not care about P50/P90?

• If the project’s biggest risk is offtaker behaviour (collections, governance, political decisions), a perfect P50 will not save a weak counterparty. Fix the structure first.

• If grid capacity is unstable and curtailment is frequent, spend your energy on grid access and protection schemes before arguing about decimal places.

• For very small rooftop systems with prepaid customers, cash may be so steady that simple trailing actuals beat modelled distributions.

P-values help, but they do not replace judgment.

What depends on these numbers (and how it shows up on your dashboard)

On the investor dashboard, we want to see, side by side:

• Energy P-curve: P50, P75, P90, P99 (one-year and multi-year, clearly labelled) - on project basis.

• Cash P-curve: the same, but after tariff rules, curtailment, O and M and debt service.

• DSCR distribution: probability of breaching 1.20x, 1.25x, etc.

• Reserve adequacy: how many months of cover under P90 and P99?

• Variance attribution: how much last month’s miss or beat came from weather, operations, curtailment or billing.

If we cannot see it, we cannot own it. Simple.

How we use P-values at Tribes Capital (our practice)

• On a project basis, we publish P50, P75, P90, and P99 for both one-year and long-term views.

• We size O&M and liquidity reserves, depending on asset class and offtaker.

• Equity base cases use long-term P50, with clear downside and upside.

• We track actuals vs P-values monthly. If a site underperforms without weather as the cause, we fix the operational root (spares, cleaning, real-time predictive failure alarms, first-fix times etc.).

• We report this in the investor dashboard in plain English, next to IRR, DSCR (if any), and impact metrics KPI like ICU uptime and study hours.

A fair challenge to ourselves

Generative AI-Powered Code Generation utilising the best LLM coding to achieve a positive outcome:

• Model risk is real. A tidy distribution can hide bias in inputs. We mitigate with independent reviews and back-testing against live assets.

• Portfolio correlation can surprise. Regional weather swings can hit many sites at once. We are building more diverse clusters and modelling correlation explicitly.

• Annual P-values can mask hourly pain. Batteries and time-of-use tariffs make shape important. We are expanding hourly stochastic models to reflect that.

• Climate drift is uncertain. We cannot promise the past will repeat. We now include trend sensitivities and will keep updating with actuals.

Quick cheat sheet

• P50 = median energy. Use for the base equity case.

• P90 = conservative energy. Use for debt sizing and reserve tests.

• P99 = severe downside. Use for stress and hard floors.

• Always ask: one-year or multi-year

• Do not sum site P90s to get the portfolio P90 without correlation.

• Tie P-values to real decisions: DSCR, reserves, bonuses, exits.

So what - three actions for allocators this quarter

1. Ask for the curve, not just three points. Request the full distribution and the inputs behind it.

2. Align P-values with decisions. Which P-value sizes debt? Which sizes reserve? Which sets bonuses? Document it.

3. Track actual vs probability. If a portfolio sits below P75 for 6 months without a weather reason, it is an operational problem, not a climate mystery.

Your capital is at risk and investments are not protected by the Financial Services Compensation Scheme (FSCS). The value of your investments can go down as well as up, so you could get back less than you invested. Stated returns are forecasted and may not reflect the reality of your return on investment.

Investing in tribes.capital includes danger, consisting of the loss of all of your capital, liquidity (the inability to offer assets promptly or without considerable loss in worth), and it ought to be done just as part of a diversified portfolio. Never invest more than you can pay to shed as well as never make financial investments with borrowed money. Investment possibilities in this system are targeted entirely at capitalists who comprehend these threats and also for whom such investments are appropriate. Tribes Capital does not provide financial investment guidance or recommendations concerning financial investment chances, and any type of investment choice should be made only based on every one of the details offered by the provider for that investment opportunity, including (without restriction) the appropriate "Project Document (PD)" that is readily available to members of the system.

Please read the risk area of our site to learn more about the threats related to the projects available on this system. Tribes Capital platform technology is designed as an open and transparent intentionality to remove non-trust between the off-taker and our impact investors. The sustainable energy infrastructure projects on tribes.capital marketplace is not a deal to the general public in any kind of jurisdiction and is readily available only to registered members of the system that have been accredited through our Know-You-Customer (KYC) & Anti-Money Laundry (AML) global compliance. Anyone who wishes to check out these investment possibilities should initially be satisfied that they are qualified to do so under the securities legislation of their country of residence and appropriate guiding laws.

Impact investors should do their due diligence and are encouraged to consult with their financial advisor, attorney, accountant or other professional who can help the investor understand and assess the risks associated with any investment opportunities with Tribes Capital.