Status and source of every lead — to see which promise brings in the “wrong” customers.
Tajar · virtual case · run of 14.07.2026
Web agency: why deadlines are missed and margin shrinks
Agency W is a 12-person agency that builds websites for small businesses across Europe. Projects are delivered late, and margin is eaten by unplanned costs: production downtime and work nobody is paid for. We looked at where the work departs from how it was meant to run.
Where it breaks — from advertising to handover
- Advertising promises “launch in 4–6 weeks” and “we take care of everything” — without the condition that the customer supplies materials on time.
- Sales quote the deadline and price as a range; the customer hears the lower end.
- After signing, the deal context is lost and materials are collected without a checklist.
- Production starts “on trust” and stops at gaps in the materials.
- The customer goes around the project manager; the project is delivered late and unplanned costs eat the margin.
Goal
Deliver projects on time so that margin is not eaten by unplanned costs: downtime and work nobody paid for.
In short
Three hypotheses to start with
This is a virtual case
This company is fictional. Any resemblance to a real company is coincidental.
The roles of the client, employees and the agency’s customer were played by AI language models following a set legend. All data in the document comes from the simulation: “confirmed” means “confirmed by two sources inside the simulation”, not in a real firm.
The case shows how we work and what the document looks like. It has no implementation results — there was no implementation.
How it was made
What the person did
Vitali set the firm’s legend and prepared the research before the interviews. He held the conversations with the “client” and the interviews with four roles — sales, project management, design, development. He collected and analysed the facts, worked through the problem areas and processed the hypotheses. He ran the closing session, where the “client” saw the result.
What the AI-based system did
In the first round the client and employee roles were played by a language model (Gemini). The second round — interviews with the marketing role and with the agency’s customer — was run by AI agents following the legend, without Vitali: three versions of the marketing role, one chosen by a critic agent. Drafts of the sections were prepared by an AI-based system that Vitali built: it holds his expertise, the service design method and his own tools built on it. Vitali checked and corrected them.
Six steps on this case
- 01
Gather information for analysis
Two rounds: a form and four interviews with roles inside the agency; then marketing and the customer. The anonymous team survey was not run in the virtual case.
Go to → - 02
See it as it is
Walked one project from first contact to handover — through the customer’s eyes and behind the scenes.
Go to → - 03
Understand what's really needed
Separated the client’s request from the goal.
Go to → - 04
Name the obstacles
Found five problem areas — from the promise in advertising to handover — and one problem before the sale, and ranked them by impact on the goal.
Go to → - 05
Hypotheses and tools
Seven hypotheses — each with instructions on how to test it.
Go to → - 06
Act and review
Defined the metrics. No implementation was carried out — the case is virtual.
Go to →
How to read the document
Every claim carries a mark. An empty field stays empty.
Two different roles in the simulation say the same thing.
Said by one person; not yet cross-checked.
A conclusion that needs testing.
No data. The field stays empty.
About the document
About the document
What it is built on, and what in it is fact and what is guesswork.
1
application form from the client
4
first-round interviews: sales, project management, design, development
2
second-round interviews: marketing and the agency’s customer
1
client document describing typical customers
1
closing session with a presentation of the research
All of it is within the simulation.
confirmedreportedhypothesisunknown
| Claim | Mark | Sources |
|---|---|---|
| The deadline promise (“4–6 weeks”) has no condition that the customer supplies materials on time. | confirmed | marketing + customer |
| Missed deadlines are dragged out by the customer’s materials. | confirmed | marketing + customer + project management |
| The customer hears the lower end of the deadline and price; what they must prepare themselves is mentioned in one sentence. | reported | customer |
| Marketing does not know which leads became customers and why the rest dropped out. | reported | marketing |
| The lead source in Notion is not always filled in. | reported | marketing |
| Materials are checked “on trust” before production starts. | confirmed | project management + design |
| Technical questions are forwarded word for word, with the meaning not translated. | confirmed | project management + development |
| Customers go around the project manager. | confirmed | sales + development |
| The project manager runs 6–8 projects at the same time. | reported | project management |
| Sales can see risky customers in advance but record it nowhere. | reported | sales |
| The developer lacks access in almost every project (“in 90%”). | reported | development |
| The agency’s customers come in three types. | reported | client document |
| The designer no longer works silently from placeholders. | reported | design (project management says otherwise) |
| The customers’ emotional curve. | hypothesis | Tajar assessment |
What we suggest tracking
To test the hypotheses, we need to start measuring what nobody counts now. Details are in the “Metrics” section. Go to →
- How many enquiries from each source become customers.
- How many projects a month pass through each problem area.
- Days of production downtime at project start.
- Number of times we go back to the customer for materials.
- Times the customer goes around the manager, and hours spent on them.
- Share of projects delivered on time.
Observations
Observations
What can be seen in the agency’s work and where each remark comes from.
Key finding
In this agency the work breaks in two places. At the entrance, marketing creates unrealistic expectations: it promises a deadline and “we take care of everything” with no conditions. Further on, at the joins: every handover between sales, the manager, production and the customer rests on memory and trust, not on a check.
Where leads come from
Organically through LinkedIn.
Advertising.
Recommendations from past customers.
Five problem areas
The promise at the entrance: marketing → sales → customer
Advertising and the landing page promise “launch in 4–6 weeks” and “we take care of everything”. The condition that the deadline holds only if the customer supplies materials on time and does not change the structure is missing from the promise: marketing removed it because “it hurts conversion”. On the call, sales quote the deadline and price as a range — the customer hears the lower end. What they must prepare themselves gets one passing sentence. Why this does not get fixed: marketing does not know which leads became customers or how their projects ended, so it optimises the cost per enquiry, not the outcome.
confirmed→ See on the map: stages 1–2
Sales → project manager
The handover is one line in Slack (“Closed! Link in Notion, give them a call”) and a status change. There is no joint conversation. Yet before signing, sales already see two signs of a future delay: the customer takes 4–5 days to reply and postpones the first calls; or mentions a board of directors, a partner or a husband who will also have to approve. These signs are recorded nowhere: calling a customer “difficult” outright feels like passing the problem on to a colleague.
reported→ See on the map: stage 3
Project manager → production
Materials are checked “on trust”: the manager skims the folder and writes “all there, let’s go”. According to the manager, it is not only that there is no checklist: they run 6–8 projects at once, and there is simply no half hour to check every file and access carefully.
confirmed→ See on the map: stage 5
Technical information gets distorted
The manager does not understand technical terms (“hosting”, “API key”) and forwards messages between the developer and the customer’s sysadmin word for word. A simple password change drags out into a week of correspondence.
confirmed→ See on the map: stages 5–6
Going around the manager
Under pressure, customers go around the manager: they phone sales (“we agreed this, you and I”) or write to the developer directly. The result is work that nobody counted or got paid for, and a manager who does not know about the changes.
confirmed→ See on the map: stages 4–8
The customer’s legal and payment readiness (for example, a payment system stuck in “pending” because the company is not yet registered) is found out after the contract is signed. According to the developer, it can be checked in 10 minutes before signing.
reported→ See on the map: stage 0
How a typical miss happened — day by day
Reconstruction from the project manager’s account. reported
The archive arrives from the customer. The status in Notion immediately becomes “In development”, and the designers are told “all ready”.
The designer works silently from placeholders and does not raise the alarm. According to the designer, he no longer does this — whether it has stuck across the whole team is unknown.
A report of “all on plan” goes to the customer.
Incomplete materials are passed to development; the access details do not work.
Firefighting mode, the customer unreachable. The delay is two weeks.
Tools
Notion for tracking projects, Slack for messages, Zoom for meetings. There is no shared map of how a project moves. Notion has no “customer risk” field, and the status does not tell “materials received” from “materials checked”.
Roles
Roles in the process
Who takes the request, who runs it, who closes it — and who it is all done for.
Marketing
A part-time marketer: advertising, LinkedIn, the website. The founder sometimes writes posts and edits website text himself.
Sales
Handle leads, run negotiations, prepare an individual proposal, sign the contract.
Project manager
The customer’s one standing contact after signing. Collects materials, sets tasks for production.
Design
Makes mock-ups from the customer’s materials.
Development
Builds the site, works with the customer’s access and hosting.
Three types of customer
Reported: client document. Names are illustrative. reported
Busy Thomas
A local service business — a clinic, a building firm, a law office.
ContextNeeds a website “to look solid”. Does not understand the technical side and agrees to everything at the sale.
Main painDrops out of contact: texts, photos and a vector logo become “a three-month task”.
Aesthetic Sophie
A craft brand — clothing, cosmetics, décor; an online shop.
ContextWants an “atmospheric” shop; deeply involved in the visuals.
Main painProduct photos from a phone; “a professional photo shoot next week” stretches to a month.
Systematic Jan
Niche B2B services — logistics, consulting, IT.
ContextTechnically literate, values structure.
Main painBrings a prototype to partners for approval and comes back with new requirements.
Voice of the customer
Reported: interview with the owner of a craft ceramics shop (“Sophie” type), second round. reported
What I came across was: a turnkey online shop, launch in 4 weeks. And also: we take care of everything, all you have to do is sell.
I heard the lower end of the deadline and the lower price — maybe that is my fault.
If they had shown me the full list and the deadlines straight away, I would have prepared better.
Turnkey does not mean without your involvement.
Journey map
Journey map
How one project goes now from first contact to handover — through the customer’s eyes.
Busy ThomasAesthetic SophieSystematic JanEmotional curve: hypothesis
Stage
01First contact
02Sale and proposal
03Handover to the manager
04Collecting materials
05Production starts
06Production hits gaps in the materials
07The customer learns of the delay
08Late handover
Emotional curve
What happens
The lead arrives from LinkedIn, advertising or a recommendation; sales get in touch.
Calls, an individual proposal.
Signing, meeting the new contact.
The manager asks for texts, a logo, access.
A message in Slack and a new status in Notion.
Design and development stop and wait.
The manager comes back for what is missing.
The project is delivered after the deadline.
Customer’s pain
Sees “launch in 4–6 weeks” and “we take care of everything” — expects a finished result without their own involvement.
Hears the lower end of the deadline and price; the call reinforces expectations instead of testing them.
A new contact who “makes them explain everything again”.
Many decisions from the customer, each with a deadline; unclear what is included in the price. The rest depends on the customer type — see below.
The customer does not see this moment.
—
Disappointment — stronger among those who came by recommendation.
The final bill is higher than the first figure; trust in deadlines is lost.
Behind the scenes
The promise has no caveat about materials; the lead source is not always filled in.
Sales see risk signs but do not record them; the list of what the customer prepares is not shown.
One line in Slack, no joint conversation.
No checklist; the manager has 6–8 projects.
A check “on trust”.
The manager does not always learn about the downtime at once.
The problem becomes visible too late.
Margin shrinks; the miss is reviewed nowhere.
Branches
Thomas drops out of contact.Sophie sends weak materials, and the reshoot drags on for a month.Jan takes the prototype for internal approval.
The map scrolls sideways.
Hypothesis. Expectations peak at first contact and the sale, when the customer hears “4–6 weeks” and “we take care of everything”; sharp drops come when collecting materials and when they learn of the delay. The shape rests on the customer interview from the second round. How to test: compare against feedback from 2–3 real customers of each type.
The pain at stages 1, 2, 4 and 8 is as reported by the customer (second round); at the others it is a hypothesis.
Stage 4 — three different delays.
Under pressure, “one contact through the manager” does not hold: the customer phones sales or writes to the developer directly (problem area 5).
Process map
Process map “as it is”
The same work as on the journey map, together with what the customer does not see. In English this kind of map is called a service blueprint.
Stage
A stage is a step of the customer's journey in time (1–8; stage 0 is before signing). A problem area is a place in the agency's work where it breaks; it can cover one stage or several.
Stage 1First contact
problem area 1
problem area 1
Stage 2Sale and proposal
Stage 3Handover to the manager
problem area 2
problem area 2
Stage 4Collecting materials
Stage 5Production starts
problem area 3
problem area 3
Stage 6Production hits gaps in the materials
Stage 7The customer learns of the delay
Stage 8Late handover
Customerbelow — line of interaction
Sees a post or an ad, receives a recommendation.
1–2 calls, an individual proposal.
Signs the contract, meets the manager.
Thomas drops out, Sophie sends weak material, Jan takes it for approval.
Does not take part and does not see.
Does not know yet.
A call or email from the manager.
Accepts the project late.
What the customer seesbelow — line of visibility
“Launch in 4–6 weeks”, “we take care of everything”; a personal message from sales.
Negotiations, a proposal “made for them”.
The manager is introduced.
The manager collects and chases.
—
—
The manager explains and asks again.
Handover.
Behind the scenes
Marketing writes the promise without the condition; the lead source and what the person read do not reach sales.
Sales see two risk signs and record them nowhere; the deadline and price are quoted as a range.
One line in Slack and a proposal in Notion; no joint conversation.
No checklist for the customer type; the manager has 6–8 projects.
“OK to start” in Slack, status “In development” without a check.
Design and development wait; management does not see the downtime.
The miss is recorded nowhere.
Margin shrinks; no review.
Risk
The customer arrives with an expectation the agency cannot meet.
Risk information is lost; expectations are reinforced instead of tested.
The deal context is lost; the customer has to repeat themselves.
Three different delay mechanisms, none of them caught.
The highest risk in the whole map: the status changes on a guess.
Downtime builds up unnoticed.
The customer loses trust in deadlines.
The same will happen on the next project.
The map scrolls sideways.
Problem area 4 · Technical information gets distorted — stages 5–6
The manager forwards technical messages word for word. Affects Jan and online shops most.
Problem area 5 · Going around the manager — all stages after 3
The customer goes straight to sales or the developer. Unpaid work, and the manager is not in the loop.
Stage 0 · online shops — before signing
Legal and payment readiness is checked only after signing.
Promise and reality
Promise and reality
What the agency promises customers and what they get. The gap between the two is the diagnosis.
Promise
Reality
“Launch in 4–6 weeks”.
The deadline depends on two things: whether the customer supplies materials on time — and whether the work breaks in problem areas 2–5 inside the agency. Even when the customer delivers everything on time, the problem areas can still make the deadline slip. The promise mentions neither.
“We take care of everything”.
The customer needs to make many decisions and supply many materials; the list is not shown to them in advance.
“Affordable”, “no surprises”.
The final bill is higher than the first figure; it is unclear what the price includes.
An individual approach.
Rests on sales.
One contact for the whole project.
Holds while everything goes smoothly. Under pressure the customer goes around the manager.
The site by the deadline (implied).
The customer learns of the delay after it has already happened.
What the customer wants (Tajar hypothesis)
To have a working site by the deadline without getting into the process or the technical details.
To know in advance what is needed from me and by when.
To look professional in front of my own customers.
To hand the job over and not worry.
Diagnosis
The agency promises a quick result without the customer’s involvement, while the work itself depends on the customer’s materials and decisions. The condition is lost already in advertising, sales do not bring it back, and from there every handover rests on memory and trust, not on a check. The five areas where this breaks are joins between people, not flaws in individual people’s work. And nobody sees the whole chain: marketing does not know how its leads’ projects ended.
Goal
Deliver projects on time so that margin is not eaten by unplanned costs: downtime and work nobody paid for.
Critical obstacles
The order by impact on the goal is a Tajar assessment (hypothesis). hypothesis
- A promise without its condition (problem area 1). The customer arrives with an expectation the agency cannot meet — and every later area pays for it.
- Production starts “on trust” (problem area 3). The highest risk: a project goes into work with gaps in the materials, and every later stage pays for it with downtime.
- Handover without context (problem area 2). What sales already know about the customer’s risk does not reach whoever runs the project.
- Going around the manager (problem area 5). Margin shrinks directly: work that nobody counted or got paid for.
- Technical information gets distorted (problem area 4). Simple technical tasks stretch into weeks.
- Online shops: checking after signing. Obstacles that could have been seen in 10 minutes appear when the project is already under way.
Structural strain — for management to decide
The hypotheses do not cover these problems, so we show them separately. This is structural strain: it cannot be fixed by changing the process and needs a management decision — on resources, roles or structure.
Risk 1
The load of 6–8 projects per project manager — according to the manager, one of the causes of the problem in area 3.
Risk 2
Technical information gets distorted (problem area 4): in the structure, nobody is responsible for making technical questions clear to the customer.
Hypotheses
Hypotheses
Each is tied to a specific area above, and each comes with instructions on how to test it with your own resources.
H1
Feedback on leads
hypothesis- If
- every lead in Notion has a source and a status “became a customer / dropped out / why”, and marketing sees a summary once a month,
- then
- it will become visible which promise brings in customers with expectations the agency cannot meet.
- How to test
- For two months, keep a status and source for every lead; after a month, compare the sources by the share that became customers.
- What we measure
- Share of leads with the source and status filled in; share of “wrong” leads by source.
- Where
- problem area 1
H2
A promise with its condition
hypothesis- If
- in advertising and on the landing page the deadline comes with its condition (“4–6 weeks if the materials are ready”), and the customer sees the list of what they prepare before signing,
- then
- there will be fewer complaints of “you promised N weeks” and fewer misses caused by materials.
- How to test
- For a month, show two versions of the landing page. Compare not the cost per enquiry but the share of enquiries that became customers, and the deadlines of their projects.
- What we measure
- Share of enquiries → customers; projects missed because of materials; complaints about the deadline.
- Limitation
- There may be fewer enquiries — that is part of the test, not a failure.
- Where
- problem area 1 · stages 1–2
H3
Readiness gate
hypothesis- If
- a project card cannot move to “In development” until the readiness checklist for the customer type is closed,
- then
- production downtime at the start will shrink.
- How to test
- Take 2–3 current projects. Close the checklist before the start; record the days of downtime at the start and compare with projects without a checklist.
- What we measure
- Days of downtime at the start; number of returns for materials.
- Limitation
- Takes the load off checking, but not the load of 6–8 projects itself.
- Where
- problem area 3 · stages 4–5
H4
Handover brief from sales
hypothesis- If
- at handover, sales record objective risk signs (“replies slowly”, “mentioned outside approval”) instead of personal opinions,
- then
- there will be fewer delays.
- How to test
- Compare 5–6 projects with and without the brief by the number of delays.
- What we measure
- Number of delays per project.
- Where
- problem area 2 · stages 2–3
H5
Going-around protocol
hypothesis- If
- every time a customer goes around the manager it is recorded on the project card and passed to the manager at once,
- then
- there will be less unpaid work.
- How to test
- For a month, count the times customers go around the manager and the hours development and sales spend on them.
- What we measure
- Bypasses per project; uncounted hours.
- Where
- problem area 5 · stages 4–8
H6
Check before signing for online shops
hypothesis- If
- before signing, the developer spends about 10 minutes checking access and payment and legal readiness,
- then
- there will be fewer obstacles after signing.
- How to test
- Run the check on the next 2–3 online shop deals and compare with the previous ones.
- What we measure
- Obstacles found after signing.
- Limitation
- Online shops only.
- Where
- stage 0
H7
Status visible to the customer
hypothesis- If
- the customer can see what is expected of them and who is holding things up,
- then
- they will close their part of the checklist faster.
- How to test
- A few projects with such a tracker and without it; compare the speed.
- What we measure
- Days to close a checklist item.
- Limitation
- Makes sense only after H3.
- Where
- stages 4 and 7
Order of testing
| 0 | Start keeping the status and source of leads (H1) and record the baseline numbers: deadlines, downtime, bypasses, share of enquiries → customers. Check with the team their estimate of hours for each hypothesis. |
| 1 | H3, H4 and H5 are cheap and can be tested in parallel on 2–3 projects. |
| 2 | H2 — once at least a month of lead data has built up: without it the effect of the promise cannot be seen. |
| 3 | H6 — separately, only on new online shop deals. |
| 4 | H7 — after the results of H3. |
| 5 | Review each hypothesis against its metrics: confirmed, rejected or needs more work. |
The team sets the pace: the same people who run 6–8 projects do the testing.
What the work would look like if the hypotheses are confirmed
hypothesis
| Stage 1 · First contact | The deadline in advertising comes with its condition; every lead has a source and later a status. | H2 + H1 |
| Stage 0 · online shops | Before signing, the developer checks access and readiness in 10 minutes. | H6 |
| Stage 2 · Sale | Sales record risk signs on the card and show the customer the list of what they prepare. | H4 + H2 |
| Stage 3 · Handover | A brief and a short joint conversation. | H4 |
| Stage 4 · Collecting materials | A checklist for the customer type; the customer sees the status. | H3 + H7 |
| Stage 5 · Production starts | The card cannot move to “In development” without a closed checklist. | H3 |
| Stages 4–8 | A bypass is recorded and goes straight to the manager. | H5 |
| Stage 7 · Status | The customer sees the status in advance instead of learning afterwards. | H7 |
| Stage 8 · Handover | On time — that is the hypothesis; if not, the miss is recorded. | — |
Not changed even here — structural strain for management to decide: the load of 6–8 projects per manager and “technical information gets distorted” (see the “Promise and reality” section).
Metrics
Metrics
How we will see whether anything has changed — and when to look. The basis is the HEART scheme (happiness, engagement, adoption, retention, task success); we added acquisition to see who is coming in.
| Dimension | Goal | Signal | Metric |
|---|---|---|---|
| Happiness | The customer does not feel left in the dark. | Feedback on clarity of communication. | A “transparency” score after the project; NPS — whether the customer would recommend the agency (0–10). |
| Acquisition | The people who come are the ones the agency suits. | Share of enquiries that become customers. | % of enquiries → customers by source; complaints of “you promised N weeks”. |
| Engagement | The customer closes their part on time. | Speed of reply to a request for materials. | Days to close an item; number of reminders. |
| Adoption | The team uses the readiness gate. | Share of projects with a checklist. | % of projects where the status was changed only after the checklist. |
| Retention | Customers come back and recommend. | Repeat orders, recommendations. | % of repeat customers; % of new customers by recommendation. |
| Task success | The project is delivered on time, margin does not shrink. | Actual compared with plan. | % of projects on time; hours over plan; bypasses per project. |
⏱
When to look: baseline numbers before testing starts; comparison after 2–3 months.
Extras
Extras
Parts added on request.
Value proposition canvas
What the customer wants to get, what the agency gives — and where the two do not meet.
Customer profile hypothesis
Jobs
- To have a working site by the deadline without getting into the process or the technical details.
- To know in advance what is needed from me and by when.
- To look professional in front of my own customers.
- To hand the job over and not worry.
Pains
- Does not know in advance what to prepare and by when.
- Learns about a delay after it has happened.
- The final invoice is higher than the first figure; unclear what the price includes.
- After signing — a new contact, and everything has to be explained again.
Expected gains
- A site by the promised deadline.
- Clarity on what the price includes.
- One contact who knows the whole history of the project.
What the agency offers
Services
- Websites, landing pages, online shops “turnkey”.
- A tailored proposal.
- A project manager as the single contact.
Pain relievers
- “We take care of everything” — in words.
- The manager collects materials and sends reminders.
Gain creators
- A personal approach in sales.
- “Launch in 4–6 weeks”.
Where it fits and where it does not
- ✓Personal approach: the customer feels it in sales.
- ✕Deadline: the promise accounts neither for the customer's materials nor for the problem areas inside the agency.
- ✕“We take care of everything”: the customer is not shown the list of what they must prepare.
- ✕Price: what it includes is not explained.
- ✕Single contact: does not hold under load (problem area 5).
Snapshot of the business model
Partners
- Customers who recommend the agency
- External contractors (hypothesis: photographers, copywriters)
Key activities
- Handling leads
- Collecting materials
- Design and development
- There is currently no readiness check before the start (see H3)
Resources
- A team of 12: sales, project managers, design, development, part-time marketing
- Notion, Slack, Zoom
- Reputation through recommendations
Proposition
- “Launch in 4–6 weeks” — with a condition that is not visible
- An individual approach
- Personal support
- Predictable deadlines — not delivered
Relationships
- Support through the manager
- Problems are raised afterwards, not in advance
Channels
- LinkedIn — organic
- Advertising
- Recommendations
Customers
- Thomas
- Sophie
- Jan
Costs
- Team salaries
- Advertising and sales time spent on handling leads
- Hidden: production downtime
- Hidden: unpaid work when customers go around the manager
Revenue
- Payment per project — structure unknown