Key takeaways

The problem: hours of sorting, then documents to rebuild one by one

The starting point is personal. Applying seriously to job offers took hours every week: browsing three platforms with different formats, weeding out duplicates, comparing vague job titles, then tailoring a resume and cover letter for every shortlisted role. Repetitive work, never finished, and its quality dropped as fatigue set in.

On top of that came an invisible obstacle: most resumes are read by machines first. According to the 6th edition of the Jobscan ATS usage report, published on August 20, 2026, 487 of the Fortune 500 companies, that is 97.4%, use a detectable ATS, and a corporate job posting receives 250 applications on average. More than 90% of employers filter or rank applications automatically before a human opens them. An overly creative layout sinks an application without anyone ever seeing it.

Two distinct problems, then: sorting an inbound stream too large to handle, and producing output documents that the machines on the other side can actually read. That is exactly the situation an SME faces with its leads or its RFPs; only the objects change.

What was built: collection, weighted scoring, ready-to-send documents

The tool is a three-stage data pipeline. First stage: collection. Ads from Welcome to the Jungle, JobTeaser and LinkedIn are retrieved and brought into a single format; without that normalization, no honest comparison is possible between a detailed posting and a three-line one.

Second stage: scoring. Each ad gets a score computed from criteria weighted for the target profile: skills match, industry, location, role requirements. More than 240 ads have gone through that calculation so far. The result lands in a filterable table, ranked from best score to worst, instead of a pile of twenty open tabs.

Third stage: production. For each high-potential ad, the tool automatically generates a resume and a cover letter tailored to the role, exported as PDFs whose structure is designed to be parsed correctly by screening software. The sorting that used to take hours now runs on its own; only the decision remains.

Why does a weighted score beat eyeballing?

Because it applies the same criteria to every item, without fatigue, recency bias or Friday-evening moods. A manual sort across 240 ads is never consistent: the first twenty get careful attention, the last hundred get skimmed. A weighted score treats line 240 exactly like line one.

The other virtue is less obvious: weighting criteria forces you to write down what actually matters. How many points for the industry? For the commute? That formalization work is worth as much as the tool itself. And when a score looks absurd, that is useful information: either a criterion is badly weighted and you adjust it, or the input data is wrong and you see it immediately.

The limit is just as clear: data processing is only worth as much as its rigor. If the data is badly structured, the score means nothing. The quality of a score is decided at collection time, not at calculation time.

What can an SME transpose from this?

Any inbound stream that outgrows the team's processing capacity is a scoring candidate. Lead triage is the best-documented example: according to figures compiled by LeanData in April 2026, B2B companies take an average of 42 hours to respond to a lead, and a 2026 Blazeo study of 573 businesses found that 74% of them miss the first 5-minute window. That is not a motivation problem: it is an unprioritized queue problem.

The table below sums up what the same mechanism yields on four classic SME streams.

Stream to sort Weightable criteria What the tool produces The decision that stays human
Inbound leads Industry, size, stated budget, source Ranked queue, draft first reply Call back first or not
RFPs Value, deadline, fit with references Overall score, summary of requirements Bid or pass
Job applications Skills, availability, salary expectations Ranked, argued shortlist Who to meet
Support requests Urgency, customer weight, recurrence Real-time prioritized queue The handling itself

What these four rows have in common: the score never decides, it orders. Weighted scoring turns an unsorted pile into a prioritized queue; the decision stays with the owner, but it now covers the 10 items that matter instead of the 200 that are waiting.

When a custom scoring tool is a bad idea

At least three cases. First, volume: below a few dozen items per month, a spreadsheet with three criteria columns and a sort does the same job for zero euros. Building a pipeline for 15 monthly leads is comfort engineering.

Second, unstable criteria: if the definition of a good lead changes every week, the tool will codify a rule that does not exist yet. Stabilize the rule on paper first, for a month or two, before automating it.

Third, a sufficient off-the-shelf tool: most mainstream CRMs already include basic lead scoring. If your criteria fit their fields, switch it on and call nobody. Custom only makes sense when the data comes from several sources outside the CRM, when the criteria are specific to your business, or when the score has to trigger automatic document production, as it does here.

The takeaway. Weighted scoring does not make you smarter; it makes you consistent. It applies your own rule to line 240 with the same care as line one, and gives you back the only time that matters: decision time. The real question is not "do I need a tool?" but "what is my rule, and is it written down anywhere?".

If a stream is piling up faster than your team processes it, the sanity check is simple: count last month's items, list your sorting criteria, and check whether they are written down anywhere. To talk it through, book a 30-minute call, and see how I build this kind of tool on the custom web app page. At minimum you will leave with an honest opinion, including when the right answer is a spreadsheet.

Frequently asked questions

What is weighted scoring?

Weighted scoring is a number computed by applying the same criteria, each with a defined weight, to every item in a stream: job ads, leads, applications or requests. The output is a queue ranked by priority, where the final decision stays human. Its reliability depends entirely on how well the input data is structured.

How much does a custom scoring pipeline cost?

With Mathieu Haye, a Sprint starts from €400 for a first shipped scope; a full pipeline with collection, scoring and document generation is quoted case by case, depending on data sources and volumes. Scoping starts with a free 30-minute call.

Doesn't an off-the-shelf CRM already do lead scoring?

Yes, most mainstream CRMs include basic lead scoring, and when your criteria fit their fields, that is the right option. Custom only makes sense when the data comes from several sources outside the CRM, when the criteria are specific to your business, or when the score has to trigger document generation.