This user is doing exactly what smart data work looks like, building a personal model from first principles, not waiting for a packaged solution. They grabbed 90 years of NOAA temperature records, matched it against their own utility bills, and started asking the right questions about which temperature variable actually drives consumption. That's not overkill. That's the foundation of practical forecasting.
The specific question, whether to match TMAX with electric cooling and TMIN with gas heating, or to use TAVG for both, reveals a deeper insight that most spreadsheet users never reach. The answer isn't one-size-fits-all. For cooling, TMAX matters most because peak afternoon heat drives air conditioner load. For heating, TMIN is more relevant because overnight lows determine how much gas your furnace burns to recover temperature by morning. TAVG smooths out both extremes, but it also dilutes the signal. The user's instinct to split the analysis is correct: match each utility to the temperature extreme that triggers its use.
This approach has a name in statistics, degree-day modeling, and it's how energy analysts predict grid demand. The user has essentially built a homemade degree-day model without knowing the jargon. That's not a problem. It's proof that the tools are accessible. The FORECAST.ETS function in modern spreadsheets handles seasonality and trend automatically, but it expects clean inputs. Feeding it TAVG will give a general pattern. Feeding it TMAX for electric and TMIN for gas will give a more accurate prediction of when costs spike.
The practical next step is to run both models side by side. Create a column that calculates cooling degree days (max temp above a baseline, say 65°F) and another for heating degree days (min temp below 65°F). Then regress kWh against cooling degree days and CCF against heating degree days. The user already has the data. What they need is the confidence to try two paths instead of searching for a single correct one. The best prediction comes from the model that matches the physics of the problem, not the one that looks neatest on a chart.
This user is ahead of most spreadsheet operators. They've moved from tracking history to anticipating the future. The only thing missing is permission to iterate. Start with TMAX for electric, TMIN for gas, and let the residuals tell you if TAVG would fit better. The data will answer. It always does.