The Cost of Asking
Every question has a cost.
When understanding the product means searching across code, docs, Slack, tickets, dashboards, AI chats and the people who remember, even a simple question can become a search party.
See what that reconstruction costs your organisation.
Your organisation
Who is working on the product?
Add AI agents above to count this.
The questions
What does it take to get to an answer?
1,760 product questions a year · approx. £110 per question
Finding answers
What happens around those questions?
How often is substantially the same context rebuilt?Work already done once, done again because it was never captured anywhere durable.
How many people typically get pulled into a product question?Only people beyond the asker and the one answering are counted here. Those two are already paid for above.
How often is work revisited because context surfaced too late?Rebuilds, workarounds and reversals traceable to something nobody knew in time.
Sources
Where do you find an answer?
Pick the places a question usually takes someone.
Model assumptions · £90/hr blended · £127/hr specialist · 220 working daysEditDone
Method, sources and assumptions
The five layers. The model runs from individual to repeated to collective to machine-assisted to downstream consequence. Finding and answering is the direct cost of establishing an answer. Repeated reconstruction is rebuilding understanding that was already established once. Getting to shared context is the people a question pulls in beyond the two already counted. Agent context cost is tool and usage spend plus the human time spent supplying and verifying what agents work from. Corrective work is work revisited because relevant context surfaced too late.
How the total is built. Annual questions = teams × questions per team per day × working days. Each question carries finding time at the blended people cost, plus answering and refocusing at the specialist cost. The remaining layers are calculated separately. There are no on/off switches: picking a band or moving a number is the switch, and every optional layer sits at its lowest setting until you move it.
Where the defaults come from. 30 minutes finding context is conservative against McKinsey’s finding that knowledge workers spend 19% of the week looking for information. 23 minutes to refocus is Mark, Gudith and Klocke (CHI 2008). 7.5 minutes answering and one reconstruction per team per day are the observed floor from our own field work, not a ceiling.
How fragmentation is applied. The number of places you search adjusts finding time only, never answering or refocusing. One or two places is treated as concentrated and adds nothing. Three or four adds 10%, five or six adds 20%, seven or eight adds 30%. The steps are deliberately shallow so that scatter is felt without becoming the whole story.
How the behavioural scales resolve. Repeated reconstruction: rarely 1.0, sometimes 1.15, often 1.35, constantly 1.6, applied to the question layer. Getting to shared context: a question pulling in just the asker adds nothing, 2 to 3 people counts 0.5 additional people, 4 to 5 counts 2.5, and 6 or more counts 5. Only people beyond the asker and the one answering are counted, because those two already sit in the question layer. We assume a quarter of questions escalate that way and that each additional person gives 30 minutes.
Why corrective work is asked behaviourally. Almost nobody knows their annual rework figure, so asking for one produces either a blank or a guess. Rarely resolves to 0.25 occurrences a month, monthly to 1, weekly to 4.33 and several times a week to 12. Each occurrence is priced at the people and hours set in model assumptions.
How agents are costed. Humans are not the only ones paying to reconstruct the product anymore. Agents spend compute doing it, and humans spend time supplying, reviewing and correcting the context agents work from. Spend is asked for rather than guessed, because you already have that number. Review time is priced the same way the question layer is. With no agents, this layer is zero and its controls are hidden. We still do not price rework caused by agent changes, duplicated investigation between agents, or more change arriving than anyone can review.
Avoiding double counting. Shared context counts only people beyond the asker and the answerer. Fragmentation touches finding time alone. Suppressed demand is never computed and never gains an input.
Not included in the estimate
Some effects of fragmented product understanding are harder to price reliably, so we leave them out: unasked questions, delayed decisions, missed opportunities, and downstream product and customer consequences. The calculator only includes costs tied directly to the inputs above.
Why this cost exists
Product has no system of record.
The cost of asking is a hidden cost. It exists because product understanding has to be reconstructed before teams can make decisions or changes.
Lumio creates the persistent product model underneath those questions, so teams and agents can work from shared context instead of rebuilding it every time.
£196,063
a year, your modelled cost of getting to product answers
Not included: questions nobody asks, changes delayed for want of context, and downstream product consequences.
£196,063
a year, modelled