DIY in Real Estate: where to live around Paris using your own LivingIndex

I recently started to look for a larger flat around Paris. But where to start? Which district is interesting to investigate since the area is very large? Beyond common information given on an advertising (nb of rooms, square footage,etc) how can I know/ assess the quality of its neighborhood in a more systematic way? Is there any green space near the location? Is it a living area with lots of shops and restaurants? What is the intensity of vehicular traffic? How far I am from the center of Paris? All these typical questions difficult to answer if you have any local knowledge.

update 2014: I finally got some funding from OSEO (a french national contest for developing innovative company). here is the landing page

For fun, I developed an experimental system extracting different local information from the web and compute a personalized score called “LivingIndex” on each location according to the user (me) preferences and constraints, scalable to large areas.

Assessing quality of life / Real estate investment

Several rankings [1,2] have been published that assess the quality of life in certain geographical areas. However several issues remain:

  • Hyperlocality: These studies are mainly at a country or at a city (world best cities to live) level. What about local assessment? Some urban design initiatives involved local residents to assess the quality/problems of their neighborhood [3]. But such participatory/crowdsourcing approach is hard to bootstrap.
  • Personal index: How to choose the factors and their weights to score an area? Each study has its own definition about what should be an ideal place and thus its related methodology to compute the index. But what about my personal preferences/constraints?

DIY your LivingIndex

Since the data started to be open (or at least accessible), why not doing my own assessment using such data used as proxy of local knowledge. So I created a prototype extracting some data from different services. I then divided a large area in a grid of smaller cells on which I computed a score using the below personal criteria to produce a heatmap where the best/worst places to live emerge. Such approach has several benefits:

  • It provides a systematic & quantifiable assessment of some of the typical questions buyers have (cf abstract).
  • It’s scalable, providing insights to detect quickly bad and good spots within a large area.
  • the system doesn’t dictate a priori what should be the ideal conditions/LivingIndex score: each user can craft their ‘LivingIndex’ score by entering their own criteria.
  • the point of this livingIndex is to be local, hyperlocal

My criteria and Scoring

Here was my limited set of criteria:

  1. ideally be at least than 30 mins door-to-door from my place to Chatelet RER (walking + bus + train included), place where I could potentially work. I can accept max a 50 min travel.
  2. have at least 1 green space at less than 200 meters from my place
  3. have at least 1 supermarket or grocery at less than 200 meters, 2 or 3 should be ideal.
  4. have at least 2 restaurants in 500 meters, 10 or more should be ideal, indicating a real living area

A score is given for each criteria according to its degree of matching with the location: a maximum score when the location ideally matches the criteria ( >=10 restaurants), or 0 if it does not match at all ( 0 or 1 restaurant), and an intermediary score if the matching is partial (5 restaurants). Furthermore some criteria are more important than others (e.g. close to my work or living near a green space was more important for me than being near a supermarket). so I added a weight to each criteria. The final score, the livingIndex, is thus the sum of scores for each criteria weighted by their importances, a basic linear combination.

Use Case: Paris and its inner ring

Let first compute each constraint. Color representation => {blue = no matching constraint, purple = all the constraints }match.

Travel / Transport related constraint – jan 2013 improved version

Duration Travel Map using the STIF network ( SNCF+RATP bus+train) to go to Chatelet les Halles. Here the colors are reversed = blue <10min, yellow = 30/40 min, purple >90min.
(information extracted + processed from the web site)

Same map but here the target is La Defense

Restaurants related constraint

Food related constraint

Park related constraint

Mixed Layer

Jan 2013 — SECTION NO UPDATED — the following map and the related comment don’t take into account the new version of the criteria about travel duration

When we mixed such criteria with their associated weight, we got the above heatmap.

“So what did you discover with such map? It’s better to live in Paris than Boulogne Clignancourt, a surrounding city? great news.” No. the point of this LivingIndex is its hyperlocality. the goal is to discover that this street or that district is better than another one. Even in Paris only a subset (purple) matches fully all the criteria and some areas need clealy to be avoided (yellow+green+blue) at the perimeter north and south of Paris.

Now if we modify the criteria with ideally 3 train/subway stations within 500 meters and ideally 20 restaurants within 500 meters to be in a very living area, we see for instance all the areas in Paris do not match such preferences. we have to select more carefully certain areas.

If we click on a location we get an explanation for the score.

Useless Bonus: here is map done for a google street navigation where the color represents the LivingIndex (red = high score) of the location, a kind of augmented reality: people/foreigner can know that they enter in an area that has a high or low quality score)

DIY opportunity

Since everyone has his/her own preferences, we can imagine a web service where every future buyer could enter her criteria, computing her own LivingIndex inside a given large area. The service will return a map with the emerging best locations (or just the top-10 best locations). Idem for the sellers who wants to promote their flats/houses via objective facts: the web service will compute a LivingIndex/score according to common criteria so that the resulting LivingIndex can be embeded/ displayed on their advertising as proof of the quality of the neigbourhood.

Future work / contribution

Such LivingIndex is not perfect.

  • some factors are missing e.g. security or pollution/traffic, travel map for a car (easy to do)
  • some criteria used could have been computed in a more accurate way e.g. taking into account the surface of a park in the scoring of the “green space” criteria (a large park is more interesting than 2 small gardens)
  • finally the criteria should be mixed with the price of the square footage to give a kind of investment ratio that could be used for the buyers or sellers.

Anyway it enables the filtering of lots of areas to focus only on an subset that need further investigation, which was the goal of my approach.
I do not have time/money to continue but If you think such funny work is valuable and you are interested to support me, or more directly contribute, or craft your own criteria and generate the map, contact me.

[1] –
[2] –’s_most_livable_cities
[3] –

Comments are closed.