how to forecast revenue in Excel and Google Sheets (2026 methods)
revenue forecasting is the most common piece of analysis a solopreneur or small business owner gets wrong. either they skip it entirely and run on gut feel, or they build a 14-tab spreadsheet that looks impressive and produces a number nobody trusts. neither approach helps. neither protects cash flow, neither informs hiring decisions, neither earns you the discount on your annual software contract because you can confidently commit to a usage tier.
a good revenue forecast does not need to be complicated. it needs to be honest, repeatable, and tied to the underlying drivers you actually control. with Excel or Google Sheets and the methods built into the 2026 versions of both, you can produce a forecast that holds up to scrutiny in under an hour. and you can refresh it every month in fifteen minutes.
this guide walks the full workflow: the four methods that cover almost every small business, the formulas, the scenario layer that turns a single number into a usable plan, and the AI assists that make the work faster without giving up control.
why most solopreneur revenue forecasts fail
three failure modes account for most bad forecasts.
failure 1: the linear extrapolation trap
last quarter grew 15%, so next quarter will grow 15%. this works until it does not, which is usually exactly when it matters. linear extrapolation ignores seasonality, market saturation, and the simple math that compounding 15% quarterly forever gets you to a number that is larger than the GDP of your country.
failure 2: the gut-feel forecast
“i feel good about q3, let’s say $200k.” no method, no formula, no honest backtest. this is not forecasting. it is wishful arithmetic.
failure 3: the spreadsheet of fictions
pages of assumptions, no anchor in actuals, every cell driven by another cell that was itself a guess. impressive to look at, useless to act on.
revenue forecasting in Excel and Google Sheets is the practice of projecting future revenue from historical data using one of four methods: linear regression, exponential smoothing (FORECAST.ETS), driver-based modeling, or scenario planning. for solopreneurs in 2026, FORECAST.ETS is the default for seasonal businesses and a simple linear trendline is the default for steady-state businesses. AI tools like ChatGPT Advanced Data Analysis can validate and stress-test the forecast in plain english.
the data you need before you start
a forecast is only as good as the history it learns from. before any formula, you need:
at least 24 data points
12 monthly data points is the absolute minimum. 24 is much better. with fewer than 12 you cannot reliably detect seasonality. our time series analysis for small business guide covers why this minimum exists.
clean revenue data
remove refunds. separate one-time revenue from recurring. exclude obvious anomalies (a single $50,000 deal that will not repeat). these distortions ruin forecasts.
the right granularity
monthly data forecasts well 3-12 months out. weekly data forecasts well 4-12 weeks out. daily data forecasts well 14-90 days out.
| business type | recommended granularity | typical forecast horizon |
|---|---|---|
| ecommerce | weekly | 8-12 weeks |
| SaaS subscription | monthly | 6-12 months |
| service / agency | monthly | 3-6 months |
| course / one-time | weekly during launch, monthly between | varies |
method 1: linear trend forecast
the simplest reliable method. works when revenue is growing at a roughly steady rate with no strong seasonality.
how to do it in Google Sheets
- select your two columns (date and revenue)
- Insert → Chart → choose line chart
- right-click the line → “trendline”
- set “type” to linear, check “show equation” and “show R-squared”
the equation is your forecast formula. plug in any future date to project revenue. R-squared above 0.7 means the trend explains most of the variation. below 0.5, your data is too noisy for a linear forecast.
the formula version
=FORECAST.LINEAR(target_x, known_y, known_x) returns the projected value for a target date.
example: =FORECAST.LINEAR(DATE(2026,8,1), B2:B25, A2:A25) forecasts august 2026 revenue from your monthly data in B2:B25.
when linear forecasts work and do not
| works when | does not work when |
|---|---|
| revenue grows linearly | strong seasonality |
| no major launch coming | one-time spike planned |
| stable customer base | fast hypergrowth phase |
| no market disruption | new competitor or pivot |
method 2: FORECAST.ETS for seasonal businesses
FORECAST.ETS handles trend and seasonality automatically. for any seasonal business (most ecommerce, education, hospitality), this is the default.
the formula
=FORECAST.ETS(target_date, values, dates) returns the forecasted value with seasonality baked in.
=FORECAST.ETS.CONFINT(target_date, values, dates, 0.95) returns the 95% confidence interval, which is what you should actually plan around.
=FORECAST.ETS.SEASONALITY(values, dates) returns the detected seasonal period (e.g., 12 for monthly seasonality, 7 for weekly).
excel’s forecast sheet (one-click)
in Excel: select your data → Data tab → Forecast Sheet. Excel auto-detects seasonality, builds the forecast, plots a chart with confidence intervals, and exports a new sheet with the underlying numbers. this is the fastest production-quality forecast you can build in 2026.
reading the confidence interval
a forecast that says “next quarter revenue will be $180k plus or minus $40k” is far more useful than one that says “next quarter revenue will be $180k.” the band is not failure. it is the honest answer. plan cash flow against the lower bound.
we use the same confidence-interval logic in our a/b testing without a data team guide.
method 3: driver-based forecasting
a driver-based forecast does not project revenue directly. it projects the inputs to revenue, then computes revenue from them.
the SaaS driver model
revenue this month = revenue last month + new MRR – churned MRR + expansion MRR – contraction MRR
each input has its own forecast. new MRR comes from new customer count × average new ARPU. churned MRR comes from prior MRR × churn rate. and so on.
the ecommerce driver model
revenue = traffic × conversion rate × average order value
forecast each driver separately. traffic with FORECAST.ETS, conversion rate from last 90 days, AOV from last 90 days. multiply.
why driver-based beats trend-based
driver-based forecasts let you ask “what if” questions. what if i double ad spend? what if conversion rate drops 10% with the new pricing? a trend-based forecast cannot answer those. a driver-based forecast can.
| element | trend-based | driver-based |
|---|---|---|
| setup time | 5 minutes | 30-60 minutes |
| accuracy steady state | high | high |
| handles strategy changes | poorly | well |
| usable for scenario planning | no | yes |
| best for | quick checks | quarterly plans |
we get into the SaaS-specific math in our SaaS metrics every founder must track guide.
method 4: scenario planning (best case, base case, worst case)
a single forecast is a wish. three scenarios is a plan.
the three scenarios to build
- best case: top of confidence interval, plus one positive event you control (a launch, a partnership)
- base case: midpoint forecast, no surprises
- worst case: bottom of confidence interval, minus one shock (lost biggest client, major refund, pricing pressure)
the cash-flow gate
look at your worst-case revenue. add up next quarter’s known costs. if the worst case cannot cover costs, you have a cash flow gap to close before it shows up. this is the single most useful output of the entire forecasting exercise.
we cover the related decision-making framework in data-driven decision making for solopreneurs.
using AI to validate and stress-test forecasts
AI does not replace the forecast. it stress-tests it.
prompt template for ChatGPT Advanced Data Analysis
here is my monthly revenue history for the last 24 months. build a 12-month forecast using FORECAST.ETS, return the values and 95% confidence intervals, identify the seasonal pattern detected, and flag any anomalies in the historical data that should be excluded.
upload the CSV, get a clean forecast back, compare it to your Sheets version. if the two diverge significantly, find out why before trusting either. our chatgpt code interpreter tutorial walks the full workflow.
prompt template for scenario stress-testing
for the same data, run a worst-case scenario assuming a 20% drop in new customer acquisition starting month 3 and a 15% increase in churn starting month 6. show the resulting revenue path and the cumulative gap vs base case.
three minutes of prompting replaces an hour of manual scenario building.
tools beyond Excel and Sheets
| tool | best for | cost |
|---|---|---|
| Google Sheets | most solopreneurs, simple cases | free |
| Excel + Forecast Sheet | full microsoft ecosystem, deepest features | included with M365 |
| Causal | dedicated finance modeling tool | free tier + $50/mo |
| Pry / Finmark / Mosaic | venture-backed SaaS | $200+/mo |
| Rows | live data + AI formulas | free tier + $17/mo |
| ChatGPT Advanced Data Analysis | natural-language forecasting | $20/mo |
for almost every solopreneur, Sheets plus a ChatGPT subscription handles every forecast worth building. dedicated tools earn their keep when you have a finance team or board reporting requirements.
we cover the broader tooling stack in best data visualization tools for solopreneurs in 2026 and the upcoming Looker Studio complete tutorial.
the monthly forecast refresh ritual
a forecast you build once decays fast. set up a 30-minute monthly ritual.
- update the actuals (last month’s real revenue)
- compare actuals to last month’s forecast (the variance is the signal)
- recompute the forecast with the new data point
- update the scenario assumptions if anything material changed
- write three sentences in a brain note: what surprised you, what changed, what you will do differently
over six months, this ritual will turn you into the most accurate forecaster on your own business. you will start to feel the seasonality, recognize the early signs of churn, and catch the inflection points months earlier than you used to.
three worked forecasting examples
example 1: the seasonal ecommerce store
a solopreneur ecommerce store with 36 months of monthly revenue data fed it into FORECAST.ETS. the function detected 12-month seasonality automatically. december was 2.1x the annual average, february was 0.65x.
the 12-month forecast: $612k base case, $498k worst case (lower bound of confidence interval), $730k best case (upper bound + planned launch). the founder used the worst case to set a hiring freeze trigger and the base case to plan inventory orders. when actuals tracked closer to base than worst, they hired one part-time helper for q4, which paid for itself by mid-november.
example 2: the early-stage SaaS founder
a SaaS at month 14 had 14 months of MRR data, growing roughly linearly. linear regression in Sheets gave R-squared of 0.91. extrapolating naively predicted $52k MRR in 12 months.
the founder built a driver-based version instead. inputs: new MRR (forecast from FORECAST.ETS on monthly new MRR), churn rate (last 90-day average), expansion MRR (last 90-day average). the driver model predicted $44k MRR in 12 months.
the gap (52 vs 44) was the difference between “extrapolating the trend” and “compounding the actual unit economics.” the founder planned around the lower number, and when reality came in at $46k, they had cushion rather than crisis.
example 3: the agency revenue cliff that wasn’t
a creative agency owner saw revenue projections look terrible: forecast.ets predicted q4 would drop 25% versus q3. before panicking, they checked the inputs.
the historical data included a 6-week lull during a previous summer when the founder had taken extended leave. the algorithm interpreted that as a recurring seasonal pattern and forecasted a similar dip for the upcoming quarter.
manually marking that 6-week period as an exclusion (anomaly handling) and rerunning the forecast produced a much more reasonable prediction: q4 down 4% from q3, in line with their normal end-of-year softness. the workflow saved them from making bad decisions on a forecast that had been confused by a one-time event.
frequently asked questions
how often should I update my forecast?
monthly. less than that and you stop catching trend changes early. more than that and you overreact to noise.
should I forecast revenue or units?
units when your prices change frequently (so revenue moves for reasons unrelated to volume). revenue when your pricing is stable. for most solopreneurs, revenue is fine.
what about Black Swan events?
forecasts assume the future resembles the past. they cannot predict regime changes (new competitor, regulatory shift, viral moment). always layer scenario thinking on top of statistical forecasting.
how do I forecast a brand new product with no history?
you cannot, statistically. use comparable products or analogues, build a driver-based model with explicit assumptions, and make the assumptions visible. update as soon as you have 3 months of real data.
can AI write the whole forecast for me?
it can do the math. it cannot do the judgment. AI tools save time on the formula-running step. they do not replace the human work of deciding which inputs to trust, which scenarios to plan, and how to act on uncertainty.
conclusion: build your first honest forecast this week
if you have at least 12 months of revenue history, you can build a defensible forecast in under an hour. drop the data into Sheets. run FORECAST.ETS for the next 12 months. plot it with confidence intervals. build best/base/worst cases. write down the assumptions. that is more rigor than 80% of small businesses ever apply to their own numbers.
then refresh it every month. the value is not in the first forecast. it is in the discipline of comparing forecast to actual every cycle and getting better at the underlying judgment.
start with the time series analysis for small business guide for the underlying methodology, then validate with statistical analysis for non-statisticians. once those are second nature, move to driver-based forecasting and the cash flow gate becomes the most valuable spreadsheet you own.