HOW WE GATHER AND IMPLEMENT DATA
Fibredog methodology
Fibredog exists because there’s no such thing as a best provider or best package – there’s only what’s best for you. Tailoring recommendations to each visitor takes a heck of lot of data, and over a quarter of a million lines of code. Here’s what we do, and how we do it.
OUR DATA MODELLING
First, we need to talk about FIRM
FIRM – the Fibredog Independent Ratings Model – is what drives every provider score, every rank, every tool and every chart on the site. In essence, it’s a giant, up-to-date database that tracks up to around 400 individual metrics per provider. Everything from various customer satisfaction signals, right down to whether you can dim the light on the front of the router they supply.
This data is then run through a series of algorithms to define the scoring applied to each provider, and to provide the necessary information to build a host of useful tools you won’t find anywhere else. If you’re a little crazy, you can even go and scroll through every single provider attribute we track and measure.
FIRM powers our tools
You will find a tonne of tools on fibredog, each designed with a purpose, whether that be to compare router specs across different providers, measure Wi-Fi dead spots in your home, or calculate your early exit fees. All of our tools are unique – you won’t find them anywhere else.
FIRM powers our reviews
When most sites review a broadband provider, it amounts to little more than ‘Yeah, it’s all fine, trust me bro’ and a buy button. Not so on fibredog. FIRM is used to create algorithmic scoring based on hundreds of empirical provider attributes. Not only does this mean our scores cannot be bought, it also means every provider is measured in exactly the same way.
FIRM powers YourMatch™ – the jewel in our crown
YourMatch is our provider and package recommendation engine. Using a simple, three-stage process that involves telling us about what’s important to you, and about the people and devices in your household, YourMatch™ will accurately recommend not only the best-fit provider, but also the exact package speed to suit your home.
The metric layer
This layer is formed of the data we gather, and is shaped in very specific ways in order to allow direct comparison across any broadband provider. For example, comparing a provider’s routers and equipment is no simple matter where one provider offers three different routers and the other offers one. This is why the metrics themselves need to be shaped.
The conceptual layer
To use an example, assigning a portion of a rating to something like ‘Does their router allow you to dim its light?’ would not provide useful numbers. Instead, an attribute like that would contribute to a concept involving a collection of similar variables – the concept being something like ‘router quality of life features’. It is from concepts, then, that scores are calculated.
The weighting layer
To keep our example alive, let’s continue to think about our ‘router quality of life’ concept. Something like that cannot possibly share equal weight with, for example, a concept such as ‘router technology’, which might consist of the number of bands, LAN connections, antennae and the range a router has. This is why the weighting layer is so important – it shrinks or grows concepts to take on the weight they deserve in the final score.
MULTI-LAYER DESIGN
FIRM consists of three discreet layers
A single overall provider rating consists of 42 discreet algorithms – know then that when you see ‘Provider X scores 7.9/10’ there are over 400 attributes and a quarter of a million lines of code defining that number. A FIRM provider rating is about as far from ‘vibes’ as it’s possible to get.
But it’s not simply a case of adding, division and multiplication. We don’t calculate a score by taking a handful of attributes and adding them up. Rather, each scoring algorithm is constructed of three layers: Metric, conceptual, and weighted.
What exactly goes into FIRM?
The metrics that feed FIRM are divided into discrete types of customer concern. Each category captures a different part of the broadband experience, then feeds the wider model so every provider can be measured consistently.
How missing or uneven data is handled
Broadband providers don’t always publish information in the same format, with the same level of detail, or in the same place. FIRM handles that mess before anything enters the model, so every provider is measured against the same structure.
A metric only affects scoring when it can be compared fairly. If it cannot be compared fairly, it has to be classified, verified or kept out of the score.
We don’t reward vagueness
If a provider is unclear about something important, FIRM doesn’t treat that uncertainty as a positive. Clear, verifiable information is always better than vague marketing copy.
We don’t publish genuinely non-applicable metrics
Some attributes simply do not apply to every provider, product or network type. Room is created in the model so things that simply don’t apply aren’t punished.
We follow the evidence beyond provider pages
First-party provider information comes first. When it’s missing or incomplete, we widen the search to trusted sources such as Ofcom, provider community forums and credible third parties that are able to properly cite their own evidence.
FIRM is not 100% objective
No useful model can be. FIRM is built from factual provider metrics, but the way those metrics are grouped into concepts and weighted into scores involves good, old-fashioned human judgement.
The metrics are factual
A contract length, a router specification, a price rise rule or a support route is not a vibe. It is a concrete attribute that can be tracked, checked and placed into the model.
The weighting is judged
Human beings decide which metrics belong together, what each concept represents, and how much weight a concept deserves in a final provider score.
FIRM doesn’t remove judgement from broadband scoring completely, but it makes that judgement explicit, structured and repeatable, so every provider is assessed by the same model rather than by a biased opinion.