--- title: How to Combine Fit and Intent Scores | Torke url: https://torke.tech/blog/combine-fit-and-intent-scores description: How to combine fit and intent scores so neither hides the other: multiply fit by dated signal strength. How we calculate the Torke Score. --- # How we calculate the Torke Score, and why fit and timing are multiplied Adding a fit score to an intent score lets one of them hide the absence of the other. Multiplying them makes a high score require both. 17 October 2026, 4 minute read, by Vincent Quero ## How do you combine a fit score with an intent score? Multiply them rather than adding them. With addition, a perfect fit with no activity can outrank a good fit with a fresh, dated signal. Multiplying fit by the strength and freshness of the strongest signal means a high score needs both. We call the result the Torke Score. Most teams that combine a fit score with an intent score add them together, and the addition quietly breaks the ranking. A company that matches the ideal customer profile perfectly but has nothing happening scores well on fit alone, and a poorly fitting company that read three of your pages scores well on activity alone, so both end up near the top of the list for the wrong reason. If you want to know how to combine fit and intent scores, the first decision is the operation, before any weighting or tool. We call our combined number the Torke Score. It multiplies an account's fit score by the strength and freshness of its strongest signal, so a high score needs both: an account that fits perfectly with no reason to act scores low, and an account that has just announced a new plant or hired a new head of operations rises to the top of the week only if it also fits. It is the number behind the [Torke Matrix](https://torke.tech/blog/account-prioritization-matrix), which turns the score into four weekly instructions. Each half is built separately. Fit is scored out of 20 from evidence we can link to, as our [Method page](https://torke.tech/method) describes, covering criteria such as industry, size, the operations the product serves and whether a buying group can be named. Timing is scored from signals only, and a signal counts only with a date and a public source. Each signal type gets a strength, a building permit for a new site weighing more than a job posting, and every signal loses weight as it ages, so an event from last week outweighs a stronger one from last quarter. Multiplication is the important choice, and the reason is arithmetic. With addition, a fit of 18 out of 20 and no signal can outrank a fit of 12 with a strong signal from last week, so the team keeps calling the perfect account that has no reason to listen. With multiplication, a timing value near zero pulls the whole score near zero, whatever the fit, and a weak fit caps how high any signal can lift an account. Neither half can make up for the absence of the other, which matches how purchases actually happen: a company needs both a reason to buy and a reason to buy now. Take two accounts as an example, with timing expressed between 0 and 1. Account A has a fit of 18 and no signal in ninety days, so its timing is 0.05 and its score is 0.9. Account B has a fit of 12 and announced an expansion nine days ago, so its timing is 0.8 and its score is 9.6. Added together, A would score 18.05 and B 12.8, and A would win. Multiplied, B leads by a wide margin, and B is the account with a reason to answer this month. To implement it, keep the two scores in separate fields in your CRM, so anyone can see why an account scored the way it did, and compute the product in a third field rather than overwriting either half. Decide the decay curve for each signal type before you look at results, review the weights once a quarter against which accounts actually became accepted leads, and never let anyone raise a score by hand without writing down the source. The score has a boundary. It ranks accounts, it does not forecast deals, and treating a score of 9.6 as a probability of anything is a mistake. It also inherits every weakness of its inputs: intent data inferred from page visits is not a dated signal and should not feed the timing half, and a fit model written to include everyone makes the multiplication meaningless because every fit is high. On a short list where signals are rare, most scores will sit near zero most weeks, and that reflects how rarely the market changes. For a revenue operations lead, the consequence is a ranking that sales can trust because it explains itself, two numbers and one product instead of a black-box total. This week, take the ten accounts at the top of your current scoring, split each one into its fit and its most recent dated signal, and multiply: if the order changes, your current model has been adding, and some of your best-ranked accounts have no reason to hear from you. ## Questions readers ask Each answer stands on its own, so you can forward one to a colleague, and an assistant asked the same question can quote it as it is. ### Why not add fit and intent scores together? Addition lets one half compensate for the other, so an account with perfect fit and no reason to act can outrank one with a real buying window. Multiplication makes the absence of either half pull the score down. ### How fast should a buying signal lose weight in a score? It depends on the signal type, and the decay should be decided before you see results. A funding round or a senior hire ages within weeks, while a building permit for a new site can stay relevant for months. ### Should third-party intent data feed an account score? Only as a weak hint kept outside the timing half. Intent inferred from page visits carries no date at account level and rarely names a person, so it inflates scores without giving sales anything to open a message with. ## Where this applies These industry pages show the same idea with the buyers, signals and stalls of one market. ### Robotics & Industrial Automation Robots, mobile robots, machine vision and industrial AI sold to factory, warehouse and logistics operators, often through integrators, where operations decides the rollout. Where deals stall: The step from one pilot site to a group rollout [How we work in Robotics & Industrial Automation](https://torke.tech/industries/robotics-and-industrial-automation) ### Software & AI B2B software, AI products and data platforms sold to mid-sized and large companies, where a team adopts the product and security, legal and procurement decide the contract. Where deals stall: The security questionnaire and data protection review [How we work in Software & AI](https://torke.tech/industries/software-and-ai) ### Cybersecurity Security, identity and confidential computing software sold to banks, insurers, critical infrastructure and large enterprises, where the CISO decides and risk and procurement can stop it. Where deals stall: Vendor risk and procurement after the proof of value [How we work in Cybersecurity](https://torke.tech/industries/cybersecurity) ### Logistics & Supply Chain Contract logistics, forwarding, customs services and freight or visibility software sold to manufacturers and retailers, where logistics evaluates and procurement runs the tender. Where deals stall: The shipper's tender, set around the incumbent [How we work in Logistics & Supply Chain](https://torke.tech/industries/logistics-and-supply-chain) ## Keep reading The next articles and pages on the same problem. ### Method Fit scores, fading signal scores and the dated triggers that open an account. [See how accounts are scored](https://torke.tech/method) ### The Torke Matrix for this week's accounts An account prioritization matrix that crosses fit with timing tells a team which few accounts deserve work this week, and which ones can safely wait. [Read the article, 4 minutes](https://torke.tech/blog/account-prioritization-matrix) ### The week something changed In a category with no search volume, the only reliable opening is reaching an account inside the window a dated, public event created. [Read the article, 4 minutes](https://torke.tech/blog/the-week-something-changed)