Chapters · 9
What is a marketing agency tech stack?
A marketing agency tech stack has two halves: channel tools that do the client's marketing, and operations tools that run the agency. This guide covers the agency's own stack, not the martech you set up for clients. The channel half changes with the agency type; the operations half is nearly the same everywhere, and it is where client context gets lost. The rule for both is one owner tool per job: one tool holds the truth, and every other tool reads from it or links to it. The owner column names the tool priced in our sample stack, not a pick over its rivals.
| Job | Typical tools | One owner (our sample) | What it holds | Where context gets lost |
|---|---|---|---|---|
| Email, docs, calendar | Google Workspace, Microsoft 365 | Google Workspace | SOWs, briefs, decks | The signed scope sits in a folder no task links to |
| Team and client chat | Slack, Teams, WhatsApp | Slack | Daily asks, decisions | A "quick tweak" in a thread never becomes a tracked request |
| Work management | ClickUp, Asana, monday.com | ClickUp | Tasks, owners, dates | The task does not know what was sold, so extra work looks normal |
| Call notes | Fireflies, Fathom, Otter | Fireflies | Transcripts | Decisions and promises stay in the transcript |
| Sales pipeline | HubSpot, Pipedrive | HubSpot | Deals, proposals | What the pitch promised never reaches delivery |
| Finance and time | QuickBooks, Xero, Harvest, Toggl | Your accounting tool (not priced) | Invoices, hours | Hours are never compared with scope, so extras go unbilled |
| AI assistant | ChatGPT Business, Claude Team, Gemini | One, for everyone | Chats | Everyone pastes client context in by hand, differently |
| Connectors | Zapier, Make, n8n | Zapier | Rules that move fields | They move the field; the reason rarely goes with it |
How many tools does a marketing agency use?
More than a third of full-service and media agencies run ten or more campaign tools, according to Basis's 2026 report, the best agency data we found (Basis sells advertising automation software). Most other counts describe in-house teams or large companies, so check who was asked.
| Figure | Who was asked | What it counts | Source |
|---|---|---|---|
| 40.4% name siloed systems; inefficient processes lead at 44.1% | 200+ advertising professionals, mostly at agencies (Basis, April 2026) | Share naming each challenge | 1 |
| 36.8% run 10+ tools, more than double 2024 | Full-service and media agencies (Basis) | Adtech and martech tools for client campaigns | 1 |
| 33% of martech capability used | 405 marketing leaders (Gartner, 2023) | In-house stacks, not agencies | 2 |
| 305 SaaS apps on average (median 240) | Zylo's customers, mostly large companies (2026) | All SaaS, every department | 4 |
Signs of agency tool sprawl
We found no study that says how many tools is too many for an agency, and a count is the wrong test: twenty tools where every client fact has one home beat eight held together by the account manager. You have tool sprawl, or silos with too few connections, when:
- Two tools hold the same client fact, like the retainer's allowance in a sheet and in the project tool.
- A person re-types context to move it between tools.
- A tool has no named owner, or nobody opened it last month.
Where does client context get lost? The context-loss score
The context-loss score is a 10-question self-audit that finds the handoffs where a person carries client context between tools. Answer it for one real client, with its account manager. Score No = 0 (someone re-types it, or it lives in one head), Partly = 1 (copied by a person or a zap, sometimes stale) or Yes = 2 (one home, and it travels with the work).
- Sale to delivery. Does delivery get what was sold (services, allowances, exclusions) without reopening the proposal?
- Scope to task. Can whoever does a task see whether it is inside the retainer?
- Kickoff facts. Are goals, audience, brand rules and approvers written once, where everyone reads them?
- Calls to work. Do decisions and promises from a call reach the task list without re-typing?
- Requests. Does an ask made in Slack, email or WhatsApp land in one list that is checked against scope?
- Approvals. Can you show which exact version the client approved, and when, without searching a thread?
- Changes to the report. When someone moves a budget or changes tracking, does the reason reach the report?
- Report to report. Does next month's commentary start from last month's targets and notes?
- AI prompts. Does your AI assistant get the client's context without anyone pasting it in?
- Handover. Could a colleague run the account from the records alone if its manager left tomorrow?
- 0 to 7: people are your integration layer. Fix questions 1 to 4 first.
- 8 to 14: context travels by copy and paste. Pick one home for client facts and connect what feeds it.
- 15 to 20: context travels with the work. Consolidate for cost, not context.
For each 0 or 1, note the two tools the context crosses and the minutes a week it takes to carry. Average those minutes per client scored, multiply by your active clients, then divide by 60 and by your headcount: that is the calculator's hours per person per week. Our client onboarding checklist covers questions 1 and 3, the scope creep guide covers 5, and how to automate client reporting covers 7 and 8.
Download the context-loss audit (Markdown)
What does an agency tech stack cost? A priced 10-person example
Our example 10-person performance agency pays $1,152.00 a month for software at list prices checked in November 2026, and, on the assumptions below, copying context between those tools costs about four times as much. It has about 15 retainers on Google Ads and Meta, one SEO specialist and three creatives. Prices are annual-billing list prices where one is known exactly, monthly where not (Swydo, Semrush), without promotions or tax; free tools (ad platforms, GA4, Looker Studio) are left out.
| # | Tool and plan | Job | Price used | Units | Per month |
|---|---|---|---|---|---|
| 1 | Google Workspace Business Standard | Email, docs, meetings | $14 per user, annual | 10 | $140.00 |
| 2 | Slack Pro | Chat | $7.25 per user, annual | 10 | $72.50 |
| 3 | ChatGPT Business or Claude Team, Standard seat | AI assistant | $20 per seat, annual | 10 | $200.00 |
| 4 | ClickUp Business | Projects and tasks | $17 per user, annual | 10 | $170.00 |
| 5 | HubSpot Starter Customer Platform | Sales pipeline | $20 per seat, list (annual or monthly) | 2 | $40.00 |
| 6 | Canva Business | Ad creative | $250 per person a year | 3 | $62.50 |
| 7 | Fireflies Pro | Call notes | $10 per seat, annual | 10 | $100.00 |
| 8 | Swydo | Client reports | $69 for 10 data sources, then $4.50 each, monthly | 30 sources | $159.00 |
| 9 | Zapier Team | Connectors | $69 flat, annual | 1 | $69.00 |
| 10 | Semrush SEO plan | SEO research | $139, monthly | 1 | $139.00 |
| Total | $1,152.00 |
That is $13,824.00 a year, or $115.20 per person a month. Swydo's row is $69 + (30 − 10) × $4.50, for 15 clients on two ad platforms. Row 2 comes from third-party pricing pages. HubSpot's product page shows $20 per seat as list beside a lower limited-time offer we did not use, while HubSpot's own pricing blogs give $15 or $9. Open each page before you copy a row.
Billing units differ (per seat, per data source, per client, by tasks), so model your next hire and your next five clients. One trap: seat minimums (two on ChatGPT Business and Claude Team, three on monday.com) hit small teams.
The hidden cost: copying context (worked example)
Software is the small number. Assume each of the ten people spends 2 hours a week copying context between tools, at a loaded cost of $50 an hour (pay, taxes, benefits and overhead divided by hours worked):
- Copying hours a year: 10 people × 2 hours × 52 weeks = 1,040 hours.
- Copying cost: 1,040 × $50 = $52,000 a year, or $4,333.33 a month.
- Stack total: $1,152.00 + $4,333.33 = $5,485.33 a month, or $65,824.00 a year.
Both inputs are assumptions: we found no study with a known method that measures copying time in agencies. The nearest figure is broader: 45% of 100 UK marketers spend at least a quarter of their working week on low-value admin tasks (Optimizely, 2026; source 3). Use the minutes from your context-loss score instead of our 2 hours.
The calculator starts with these ten rows, 2 hours a week and $50 an hour, all examples to change. It shows software spend per month, year and person, the cost of copying and the total; it runs in your browser, stores nothing and formats your currency without converting. The worksheet adds owners, renewal dates and a keep, merge or cut column.
Download the sprawl-cost worksheet (CSV)
How do you connect agency tools, and what does it cost?
Agencies connect tools four ways; each costs build hours once and upkeep hours every month, in our example more than the connector itself, and the most useful context still does not travel.
| Way to connect | What you pay (checked 4 Nov 2026) | Setup and upkeep | What still does not travel |
|---|---|---|---|
| Native integration | Included in the plan | Minutes to switch on; can break quietly when a field changes | Anything outside the fields the vendor chose |
| Zapier, Make or n8n | Zapier Professional $19.99 (750 tasks) or Team $69 (2,000 tasks) a month, annual; n8n Starter €20 a month, annual (both third-party pages); Make: sources disagree | Hours per automation to build and test; someone reads the error emails | The reason, the approval, the client's history |
| MCP (Model Context Protocol) servers for AI tools | Depends on the AI tool and the app's plan | Setup per server; permissions per person | Judgment: the AI can call the app but does not know what was agreed |
| A shared client record | A platform subscription | Setup per client, then confirming facts as they change | Whatever nobody confirms |
Worked example. Eight automations at 2 hours each to build and test is 16 hours, or $800 once at $50 an hour. Thirty minutes a month each to check runs and fix breaks is 4 hours, or $200 a month: nearly three times the $69 Zapier Team plan. Delete any automation nobody can explain. Before AI agents call your apps, set per-client permissions, as our guide to AI agents for agencies explains.
All-in-one vs best-of-breed agency software: how to decide
Neither pure model fits most agencies: run a hybrid, with specialist tools for channel work and a connected core for operations.
The case for one suite comes mostly from vendors. An older study by Promethean Research with the vendor Productive (220 agencies, 2021 to 2023) gave the opinion that an all-in-one agency management suite is "the right call for almost every agency" (source 6). SPI Research found professional services firms using PSA software reported 70.2% utilization against 65.0% without, a correlation across 509 firms, as summarized by Deltek, a PSA vendor and the report's sponsor (source 7). We agree for operations, not for channel work: agency suites do not buy media or audit sites, the specialist tools do, and clients often own those accounts.
- Channel tools stay specialist, chosen per job.
- Operations get one home: scope, requests, approvals, promises and client facts. If two tools hold the same fact, one goes or one reads from the other.
- A specialist earns its seat through daily use by more than one person, or because a client requires it.
- Under ten people, check seat minimums and annual terms before features.
- No owner, no tool. Every tool and connection has an owner who reviews it each quarter.
If your context-loss score is 15 or more, a project tool plus a connector, such as ClickUp and Zapier, may be enough. If each tool works but your score is low, the gap is the layer between them: one shared client record that tools, team and AI all read. That is what an AI operating system is for; our guide to what an AI operating system for marketing agencies is sets out when one fits and who should not buy one. For the AI tools themselves, see the AI tools we compare job by job.
How to consolidate your agency tech stack in 30 days
You can consolidate an agency tech stack in about a month if you decide by job, not by tool, and cancel before renewal dates.
- Days 1 to 7: list everything. Pull 12 months of card and invoice statements into the worksheet: each tool's job, owner, seats, billing unit and renewal date. Run the context-loss score on two clients.
- Days 8 to 14: decide. Name one owner tool per job; mark every other tool keep, merge or cut. Check export options and annual terms before promising a cancellation.
- Days 15 to 23: move. Export what you need, import the history worth keeping, and rebuild only the connections that replace re-typing. Write a one-page "where things live" list.
- Days 24 to 30: switch off. Remove seats before renewal, keep a read-only export, and re-run the score and the calculator. Book a quarterly review: across Vertice's customers, SaaS prices are rising by over 12% a year (Vertice, 2026; source 5).
Where RevSmith fits (and where it does not)
RevSmith makes this guide and appears in it. We list it where it fits and say where it does not. It is the connected-core option above, not a replacement for your channel tools.
- RevSmith connects to the apps in Composio's directory and to any MCP server, and maps each login to the right client. An app that needs a developer app registered says so instead of failing.
- It does not replace your ad platforms or project tool; it reads them and keeps the context in a Company Brain (called the Agency Brain inside the product) and a Client Brain per client: confirmed facts with their source, coming in only through setup, meetings, chat and workflow steps, each added only after a person accepts it. Email and Slack never feed a brain.
- It is aimed at questions 3 to 6 and 9 of the score: client facts in one Client Brain, call decisions it drafts for you to approve, a request inbox a person checks against a live count of what was sold (a matched Gmail or Slack message can become a suggested request; not yet checked live), approvals tied to an exact version, and AI runs that start from the client's brain.
- It reads nine sources daily with no AI (Google Ads, Meta Ads, GA4, Search Console, LinkedIn Ads, TikTok Ads, HubSpot, Shopify, Google Sheets); these live reads are not yet proven on a customer account, and LinkedIn Ads and Shopify need RevSmith's own app registered first. A run for one client can use only that client's connected accounts.
- It is not a PSA, time tracker, accounting tool or client portal. RevSmith is onboarding its first agencies and is not in the sample total: we set terms with each agency during onboarding; book a demo to ask.
Our side-by-side pages say when ClickUp, monday.com, AgencyAnalytics or HubSpot is the better choice.
Sources
- Basis (Basis Technologies, which sells advertising automation software), 2026 Advertising Agency Report, announced 20 April 2026. Sample: 200+ advertising professionals (213 per one outlet), fieldwork early 2026. Population: advertising professionals, mostly at agencies; the 10+ tools figure covers full-service and media agencies and counts adtech and martech tools for client campaigns. https://basis.com/reports/2026-advertising-agency-report and https://basis.com/news/advertising-professionals-believe-traditional-agency-model-is-broken. S41, S42, S44: Verified-primary.
- Gartner, 2023 Marketing Technology Survey, press release "Gartner Survey Finds 63% of Marketing Leaders Plan to Invest in Generative AI in the Next 24 Months", 23 August 2023. Sample: 405 marketing leaders, May and June 2023. Population: in-house marketing organizations, not agencies. https://www.gartner.com/en/newsroom/press-releases/2023-08-23-gartner-survey-finds-63-percent-of-marketing-leaders-plan-to-invest-in-generative-ai-in-the-next-24-months. S48: Verified-primary.
- Optimizely (a digital experience software vendor), UK marketer research, press release "Almost half of marketers now spend a quarter of their week on admin", 18 August 2026. Sample: 100 UK marketers. Population: UK marketers, not agencies. https://www.optimizely.com/company/press/uk-survey-ai-admin/. S45, S46: Verified-primary.
- Zylo (SaaS management software), 2026 SaaS Management Index, 29 January 2026. Sample: platform data on $75B+ of spend and 40M+ licenses, not a survey. Population: Zylo's customers, mostly large enterprises. https://zylo.com/reports/2026-saas-management-index/ (S127: Verified-primary); app counts via https://www.zylo.com/blog/saas-statistics/ (S128: Verified-secondary).
- Vertice (SaaS procurement software), SaaS Inflation Index, 2026. Sample: spend Vertice manages for its customers. Population: Vertice's customers, not agencies. https://www.vertice.one/l/saas-inflation-index-report. S129: Verified-primary.
- Promethean Research with Productive (an agency software vendor), Ultimate Digital Agency Tech Stack Report, research 2021 to 2023. Sample: 220 agencies, 19 tools. Population: digital agencies. https://productive.io/ultimate-digital-agency-tech-stack-report/. S52: Verified-secondary.
- SPI Research, 2026 Professional Services Maturity Benchmark, 24 February 2026, as summarized by Deltek (a PSA vendor and the report's sponsor). Sample: 509 professional services organizations employing 245,000+ consultants. Population: professional services firms, not agencies. https://www.deltek.com/resources/articles/professional-services-billable-utilization-historic-low/ and https://spiresearch.com/reports/2026-ps-maturity-benchmark/. S24: Verified-secondary; S25: Verified-primary.
- Vendor pricing pages linked in the sample stack and connector tables, list prices read from search results on 4 November 2026. On 5 November 2026 the pages for rows 4, 5, 7, 8, 9 and 10, Claude Team and n8n were opened: row 4 went from $12 to $17, and row 10 became Semrush's SEO plan at $139 (its Pro plan is gone). Rows 1, 2 and 6 and ChatGPT Business could not be read in US dollars or opened that day, and row 2 rests on third-party pricing pages. HubSpot's $20 list price and its limited-time offer were seen on https://www.hubspot.com/products/crm/starter; HubSpot's own pricing blogs, seen in search results, give $15 a seat billed annually and $9 (https://blog.hubspot.com/website/hubspot-content-hub-pricing), so the row is shown with that conflict. Status: a dated snapshot, rechecked on the publish date (RevSmith tool price check, 4 November 2026; HubSpot row S153, vendor page in results).
- In Basis's 2026 survey of 200+ advertising professionals, 40.4% named siloed or disconnected systems a top challenge, second only to inefficient processes at 44.1% (Basis, 2026 Advertising Agency Report; source 1).
- 36.8% of full-service and media agencies run 10 or more adtech and martech tools for client campaigns, more than double the 2024 share (Basis, 2026; source 1).
- Most other stack figures describe in-house teams: Gartner's 2023 survey of 405 marketing leaders found they use 33% of their martech stack's capability (Gartner, 2023; source 2).
- 66% of 100 UK marketers say managing multiple tools and platforms creates unnecessary work (Optimizely, 2026; source 3).
- Our sample 10-person performance agency stack costs $1,152.00 a month at list prices checked in November 2026, or $115.20 per person a month (our price check; source 8).
Frequently asked questions
How many tools is too many for a marketing agency?
There is no agreed number. Basis found 36.8% of full-service and media agencies run ten or more adtech and martech tools for client campaigns, more than double the 2024 share. A better test: you have too many when two tools hold the same client fact, people re-type context to move it, or a tool has no owner.
What tools does a marketing agency need?
One tool per job: email and docs, team chat, work management, call notes, a sales pipeline, accounting, one AI assistant, one connector and a reporting tool. Add the channel tools your service needs: ad platforms for paid media, Semrush or Ahrefs for SEO, a scheduler for social, design and review tools for creative.
How do I build a marketing agency tech stack, and what mistakes should I avoid?
Start from jobs, not tools: name one owner tool per job, make every other tool read from it or link to it, then add only the channel tools your services need. The usual mistakes are buying by department, so two tools hold the same client fact; paying for tools nobody uses; forgetting that per-seat prices grow with every hire; and running tools or automations with no owner.
What does tool sprawl cost an agency?
Two costs: subscriptions, and the time people spend carrying context between tools. Our example 10-person stack costs $1,152.00 a month at list prices checked in November 2026; at an assumed 2 hours a week each and $50 an hour, copying adds $4,333.33 a month. We found no independent study that measures this for agencies, so time your own team.
How much should a small agency spend on software each month?
We found no agency benchmark with a known method. Our sample stack works out at $115.20 per person a month at list prices checked in November 2026. Zylo's 2026 median of $9,455 of SaaS spend per employee a year covers its customers, mostly large companies, so it is not an agency target. Model your next hire and your next five clients.
Is all-in-one agency software better than best-of-breed tools?
For most agencies, neither alone. Keep specialist tools for channel work, where they do the job better and clients often own the accounts. Put operations (scope, requests, approvals, promises and client facts) in one home, because that is where context gets lost. Under ten people, check seat minimums and annual terms first.