This is good work, but I would be losing something of value by kicking out personas.
The last time I used personas I differentiated between “Speculative” and “Researched” information. I based the personas initially on personal assumptions and what the stakeholders said, and then colored that information blue, to indicate that is was speculative.
Then I went out and tried to find if that person really existed. Where speculations are confirmed, I re-colored the text black. I could then look at the persona and know what data was real, and what was assumption.
Any new (real) data I added in black too. Where there were multiple data points available I create a “persona range” sheet that indicated the data range (e.g. age).
So this gave me two artifacts: A single page “persona range” indicating the boundaries for the different attributes collected, plus another single-pager (actually we did a couple) that picked out single, composite individuals as being either representative of the group or representative of its diversity.
Now, naturally, you will say “but developers are simply going to use that data to apply their own biases around gender/age/etc.”
Well yeah — we all know that. Of course. No great revelation there. The point is if we do not do this then developers will operate under a far, far worse assumption: That users will share the developer’s comprehension of the design, and if they don’t, then they must be outliers of low intelligence and therefore beyond help. This leaves me with substantial (and oh-so-predictable) re-work in almost every case. The point of personas is to keep reminding developers that the end-users are not themselves — and yes, most developers do need constantly reminding of that.
It is also appreciated (of course) that this will not in and of itself render a perfectly-crafted design. However, well-crafted personas will at least help me take something into the zone where I’m not ashamed to show it to real users and gather some actual data.