Every agency has a "first 90 days" deck. Most of them are aspirational. They describe an idealised path where data is clean, stakeholders are aligned, access is granted on day one, and the work starts producing results in a fortnight.
Real engagements are messier. Below is what we actually see, week by week, across roughly thirty engagements over the last 24 months. The shape is reliable enough that we can predict what week 4 will look like before week 1 starts. The shape doesn't make the work easy — it makes it manageable.
Weeks 1–2: discovery and access
The single biggest determinant of how well the engagement runs over the next twelve weeks is what gets resolved in the first ten days. Specifically:
- Platform access (paid media accounts, CRM, analytics, CDP if any) requested and granted.
- The data-sharing agreement and PDPL compliance posture nailed down.
- The five questions from our brief framework answered concretely.
- The decision-making structure on the client side mapped — who can approve creative, budget shifts, and strategic pivots, and how fast.
The thing nobody warns clients about: in roughly 60% of engagements, at least one of the four lines above stalls in the first two weeks. Usually access. The platform admin is on leave, the IT team requires a security review that wasn't scoped, the parent company has a global account governance rule that nobody surfaced. We plan for this — the Day-30 milestone is set assuming a 5–7 day access delay.
VIMS01 produces a discovery deck at the end of week 2: a baseline diagnostic of where the brand currently is, with an emphasis on the gaps that will block agent activation if not fixed.
Weeks 3–4: instrumentation
This is the unglamorous, high-leverage work. The agents are only as good as the data they read. Most marketing data we inherit is broken in at least one of these ways:
- Conversion events are double-firing or under-firing depending on platform.
- The CDP (if there is one) has identity stitching gaps that hide 15–40% of the cross-device journey.
- UTM hygiene is inconsistent across channels — different naming conventions per platform, manual builders that drift, paid social campaigns that don't tag.
- The CRM-to-platform conversion API integration is half-built or stale.
VIMD01 leads the instrumentation work. We don't bolt on a new measurement layer; we fix the existing one. By end of week 4, the data layer should be reliable enough that any future analysis or optimisation we run on it doesn't have to ask "is this real or is this a measurement artefact?"
This is the week where clients sometimes get nervous because the visible output is low. There's no creative running yet. There's no campaign restructure visible. The dashboard doesn't show wins. We over-communicate during this phase because the work being done is structural and the client can't see it.
Weeks 5–6: architecture and baseline
With clean data, VIMS01 produces the campaign architecture: how the channels, audiences, and creative streams will be organised over the next ninety days. For multi-property or multi-product clients this is the most consequential single deliverable of the engagement — the architecture decides what's possible.
In parallel, VIMD01 publishes the baseline: the honest pre-engagement performance numbers, with confidence intervals. This is the number against which everything that follows is measured. We sometimes get pushback on the baseline because it's lower than the platform-reported numbers the client was used to — that's almost always because the platform numbers were inflated by the issues fixed in weeks 3–4.
The architecture and baseline are reviewed and approved by client leadership at the end of week 6. This is a hard checkpoint. We don't move into agent activation without it.
Weeks 7–8: activation
VIMP01 and VIMC01 come live. The campaigns under the new architecture launch. The creative pipeline produces its first archetype packs.
The first two weeks of activation almost always have a performance dip. The reasons are well-understood — the agents are training on the new architecture, audiences are being routed for the first time, the creative pipeline is producing initial variants that haven't yet been pressure-tested. We forecast this for clients up-front. The dip lasts 8–12 days and is followed by a rapid climb.
The work in this period is operational: monitoring, exception-handling, and a daily standup with the senior strategist. The agents do the execution. Humans handle the things the agents can't see — escalations, brand judgement, anything that requires understanding the client's business context outside of marketing.
Weeks 9–10: first optimisation cycle
By week 9, there's enough data on the new architecture to run the first proper optimisation cycle. VIMP01's bid models are trained on the new structure. VIMC01's creative win-rate is identifying which archetypes deserve scaling and which to retire. VIMD01 publishes the first proper performance read, decomposed by audience cohort and channel.
The optimisations are not heroic moves. They're a series of small, evidence-backed decisions: shift 8% of spend from this audience to that one, retire these three creative archetypes, scale this one, expand into this previously-excluded geo segment. The compounding of small good decisions across a quarter is what produces the headline ROAS numbers most case studies report.
Weeks 11–12: first reporting cycle and forward plan
The end of week 12 is the formal Day-90 review. The deliverable is a single document with three sections:
- What happened. Performance against the baseline, decomposed by the metrics that matter (the four metrics covered in a previous piece).
- What we learned. Specific findings about the brand, the audience, the channels, and the creative archetypes that work. These compound into the next quarter's strategy.
- What's next. The 90-day plan for the following quarter, with explicit hypotheses and the metrics that will validate or kill them.
If the engagement is working, the conversation in this review is detailed and forward-looking. If it isn't working, this is where we surface the structural reasons and the conversation gets harder. We've had both. The latter conversations are the ones that determine whether the engagement is salvageable or whether it should be unwound — and we're as direct about that as we are about the wins.
What we don't do in the first 90 days
To set expectations correctly, here's what is not in scope for Day-1 to Day-90:
- A brand repositioning. If the brand needs repositioning, that's a separate workstream. The first 90 days operate within the existing positioning.
- A full creative refresh of all evergreen assets. We refresh the assets that move the metric. The full library refresh is a Day-90+ project.
- A platform consolidation. If the brand is on six tools and should be on three, that's a separate procurement and migration project. The agents work with the current stack.
- An organisational change on the client side. We give honest input when asked, but reorganising the marketing team isn't our project.
The discipline of staying in scope for the first 90 days is what makes the next 90 days possible. The engagements that try to do everything at once are the ones that don't make it past the first quarter.
The version of this that is honest
The 90-day plan above is real. It is also imperfect. Engagements miss milestones. Some weeks the work is harder than the plan suggests. Some clients have business realities that compress or extend particular phases. We've had engagements where week 4's instrumentation work took eight weeks because the data layer was worse than initial discovery surfaced. We've had engagements where week 7's activation produced clean wins inside a fortnight because the brand was strong and the architecture was the only blocker.
The shape is the right reference. The dates flex with reality. What we promise is the shape and a candid update on where we are against it, every Friday, for ninety days. After that, the cadence is whatever the engagement actually needs.