1. LandOffer
  2. Blog
  3. All articles
  4. Decide Whether to Build or Buy a Workflow Before Comparing Vendors

Product & Project Delivery

Decide Whether to Build or Buy a Workflow Before Comparing Vendors

Compare build, buy and a limited adapter against the same workflow requirements. Use an original one-year ownership and exit budget to expose capacity limits, ongoing responsibilities and the evidence needed before choosing a vendor.

10 min readLandOffer team

Decide what the workflow must preserve, which parts are genuinely distinctive, and who will operate it before comparing vendor feature lists. Then compare reuse, build, buy and a limited combination against the same requirements, available capacity and ownership period. Initial development cost and subscription price are only parts of that decision. Integration, support, changes and a usable exit need owners and budget too.

This LandOffer.ai guide uses a complete fictional catalogue-proof workflow and an original one-year cost model. Primary guidance was checked October 27, 2026. All people, workloads, rates, options and decisions below are invented planning assumptions. No supplier was evaluated, product configured, contract signed or workflow deployed. We executed only local checks of the authored budget arithmetic and capacity relationships.

Three adults compare an unfinished wooden sorting rack with a modular cabinet beside maintenance tools and illustrated sheets.

Define the job before choosing the acquisition route

Start with the business outcome and its boundaries. “We need an approval tool” is too broad to distinguish an ordinary form from the part of the process that causes expensive rework. Describe the trigger, users, decision rights, records, exceptions and final destination. Identify the system that remains authoritative after a new workflow is introduced.

GOV.UK's purchasing-strategy guidance connects the acquisition choice to user needs, available capabilities, lifecycle costs and the ability to maintain the result. It describes building, buying and combining approaches. Its government procurement rules apply in that context; they do not become universal contract requirements for a private company.

Before commissioning anything, check whether an approved existing service can meet the need through configuration or a simpler operating rule. Reusing an available component still has access, training, integration and support costs. The other department's invoice does not remove your own setup and operating responsibilities.

Separate a real requirement from a familiar habit. A bespoke screen that copies an old spreadsheet is not automatically a competitive capability. Conversely, removing a required second approval to fit an inexpensive tool changes the workflow's business control, not merely its appearance. The accountable owner must decide whether that change is acceptable.

Worked asset: compare three routes for the same workflow

The fictional publisher Alder Press routes seasonal catalogue proofs through layout and translation reviews. Forty internal users produce an assumed 900 proof requests per year. Its current approved register remains the authority for release to print; the proposed workflow records reviews but cannot itself authorize printing.

Operations owner Verena defines the required behavior: retain both reviews against a specific proof revision; reopen only the affected review after a change; preserve who decided what; and export usable records with attachment relationships. A translation correction can leave layout approval intact only when the named layout reviewer confirms that the change did not affect layout. The tool must not infer that decision.

Engineer Otis assesses integration and ongoing ownership. Service administrator Zoya owns configuration and user support. Data steward Hugo defines the export reconciliation check. Sponsor Marcel controls spending and capacity. These are fictional assignments. Marcel has 20 setup days available across implementation, integration and rollout, plus 16 ongoing days for maintenance and support. The four investigation days mentioned later are a separate allowance. An actual project needs people who accept each allocation.

The reuse check finds an already approved shared request service. Within this authored case, its owner confirms that it has no supported way to preserve the two review states and their revision relationships. Alder keeps the existing register while investigating alternatives. That case assumption is not a criticism of any real shared service or proof that customization is necessary everywhere.

The team completes this comparison before opening a vendor shortlist:

Requirement or constraint Build the workflow Buy and configure Buy a base and build a narrow adapter
Request intake and routine routing Team implements and operates ordinary forms, queues and notices Seek supported configuration; do not assume a specific product meets it Base product handles ordinary intake and routing
Two revision-specific reviews Otis implements explicit states and reviewer decisions Must demonstrate both states and affected-review reopening without losing history Small internal adapter owns review-state rules; base stores the work item
Authoritative print decision Send a review packet to the existing register; no automatic print approval Same boundary required Same boundary; adapter must not become a second release authority
Identity and integration Team owns integration, authentication behavior and change compatibility Supplier handles its product; Alder still owns its identity setup and connected-system behavior Two change surfaces: base product and adapter, with Otis responsible for their boundary
Export and exit Own code helps modification, but record and attachment export still needs verification Require schema, attachments, history and access terms that Hugo can reconcile Need both base export and adapter state mapping; neither export alone is sufficient
Available people Build estimate needs more implementation capacity than Marcel can presently assign Configuration route fits the provisional capacity envelope if the hard review case works Adapter route fits only after explicitly displacing another internal project, subject to Marcel actually releasing that capacity
Current disposition Do not fund now: capacity is unavailable Investigate first; functional fit remains unverified Retain as a fallback, not a preapproved rescue for any product limitation

This table contains proposed responsibilities and fictional constraints, not product test results. It gives the later vendor evaluation a concrete pass-or-stop question: can an ordinary user change a proof revision while preserving the correct independent review evidence? A polished demo of a single Approve button does not answer it.

Price the same ownership period and service boundary

Use one horizon, workload, currency and labor convention across options. Explain whether an internal day represents an invoice, a loaded budget rate or displaced work. Salaried engineers are not free capacity, but their allocated cost is also not automatically additional cash spending.

The UK Service Standard's technology guidance asks teams to consider ownership costs, sustainable operation and the ability to change direction. That supports a lifecycle comparison; it does not supply a universal discount rate, ideal staffing level or maximum subscription price.

For Alder's illustrative first-year budget, every internal day is valued at $600 of allocated capacity. All figures are USD assumptions, not supplier quotes. The model reserves a full twelve months of operating costs plus setup. It does not predict a go-live date or prorate a hypothetical delayed launch. Forty seats at an assumed $12 per user per month equal $5,760 for either purchased base.

The model excludes taxes, inflation, financing and contingency; their absence must be visible before someone treats it as a spending ceiling. It also does not quantify revenue, productivity gains or avoided errors. An omitted benefit cannot be manufactured to make the preferred route win.

Worked asset: one-year ownership and exit scenario budget

These filled estimates keep implementation, integration, maintenance, support and rollout separate. Setup includes building or configuring the workflow and its planned verification. Integration covers connecting existing identity and register interfaces. Maintenance covers upgrades and behavior changes; support covers user incidents and administration. The same hour must not appear in two categories.

First-year allocation Build Buy and configure Base plus adapter
Implementation or configuration, including planned verification 60 days × $600 = $36,000 8 days × $600 = $4,800 20 days × $600 = $12,000
Identity and register integration 8 days = $4,800 6 days = $3,600 10 days = $6,000
Twelve-month maintenance 18 days = $10,800 4 days = $2,400 10 days = $6,000
Twelve-month support and administration 6 days = $3,600 6 days = $3,600 6 days = $3,600
Training and rollout preparation 6 days = $3,600 4 days = $2,400 4 days = $2,400
Assumed annual subscription $0 $5,760 $5,760
Assumed infrastructure cash cost $2,400 Included in assumed subscription Included in assumed subscription
Total allocated labor plus stated cash costs $61,200 $22,560 $35,760

Build uses 98 allocated days; buy uses 28; the adapter route uses 50. Of these, setup consumes 74, 18 and 34 days respectively. Buy fits the 20-day setup allowance; build needs another 54 days and the adapter another 14. Ongoing maintenance plus support consumes 24, 10 and 16 days respectively, so build also exceeds the separate 16-day ongoing allowance by eight days. These allowances describe capacity, not a promised delivery calendar. The first-year mixed totals are useful for comparing resource demands, but they are not three cash invoices. Build combines $58,800 in labor allocation with $2,400 infrastructure. Buy combines $16,800 labor with $5,760 subscription. The adapter combines $30,000 labor with $5,760 subscription.

Now assume the organization exits at the end of that year. The following costs are additional scenario allowances, not an inevitable charge and not included in the first-year totals. They cover extraction/reconciliation, loading into a replacement and two months of incremental overlap. Replacement implementation and ordinary second-year operation are excluded, so this is not a complete replacement budget.

Exit work at the end of year one Build Buy and configure Base plus adapter
Extract and reconcile records, history and attachment links 4 days = $2,400 6 days = $3,600 8 days = $4,800
Map and load the extracted packet into a replacement 6 days = $3,600 6 days = $3,600 8 days = $4,800
Two-month incremental old-system overlap $400 infrastructure $960 subscription $960 subscription
Additional exit allowance $6,400 $8,160 $10,560
First year plus this bounded exit scenario $67,600 $30,720 $46,320

The subscription overlap assumes a two-month extension at the same rate; no real contract offers that right. If only annual renewal is available, replace $960 with the actual commitment. Data ownership does not itself establish extraction speed, attachment completeness or an affordable termination route.

Two adults organize illustrated folios and tied envelopes inside a wooden crate beside archive shelves and an open doorway.

Challenge the assumption that could reverse the decision

A lower modeled cost warrants investigation only while the route preserves the required behavior; it does not establish functional fit. Suppose the buy option needs 30 extra internal days of unsupported workarounds. At $600 per day, that adds $18,000, taking its first-year model to $40,560. It then exceeds the adapter model's $35,760 by $4,800, before considering whether those workarounds remain supportable.

That sensitivity does not prove the adapter is better. The extra days are a hypothetical stress case, and its own estimate may change. If the bought product cannot satisfy a mandatory revision or export condition, price alone cannot qualify it. Reopen the operating requirement or acquisition route with the owner rather than burying the limitation in a score.

GSA's cloud acquisition discussion identifies transition work and parallel old/new systems as cost-model challenges. This undated government discussion is used here for that limited point, not its broader historical claims about cloud standards or procurement law. Alder's overlap and export allowances remain original assumptions requiring actual evidence.

Record a route decision with owners and stop conditions

Marcel records: “Investigate supported purchase and configuration first. Do not authorize acquisition, implementation or an adapter. Verena must verify the two-review revision case; Hugo must inspect a representative export; Otis must estimate the actual connection and change burden. Zoya must accept the administration workload. Return with the same scenario evidence for every candidate, confirmed ownership and revised costs before asking for a purchase decision.”

The fictional plan reserves four internal investigation days across those owners before any annual commitment. That discovery allowance is separate from the implementation model above, not evidence that the assumed product costs have been validated. No subscription, data upload, installation or operational test was actually performed for this article.

Stop if the hard workflow condition cannot be met through supported configuration, the export cannot preserve required relationships, or named owners lack capacity. Investigate the adapter only if its bounded responsibility is clear and its maintenance can be funded. If every route fails, reconsider the process or defer the project instead of quietly creating an unsupported system.

For the next decision, bring a common requirement packet, full-period allocations and a credible exit question before asking which vendor is best. More practical workplace guides are available from LandOffer.ai. The useful outcome is a defensible next investigation with explicit limits; a selected acquisition route still needs evidence before spending or deployment.

Sources and Further Reading