We studied the evolution of a small eruptive flare (GOES class C1) from its onset phase using multi-wavelength observations that sample the flare atmosphere from the chromosphere to the corona. The main instruments involved were the Coronal Diagnostic Spectrometer (CDS) aboard SOHO and facilities at the Dunn Solar Tower of the National Solar Observatory/Sacramento Peak. Transition Region and Coronal Explorer (TRACE) together with Ramaty High-Energy Spectroscopic Imager (RHESSI) also provided images and spectra for this flare. Hα and TRACE images display two loop systems that outline the pre-reconnection and post-reconnection magnetic field lines and their topological changes revealing that we are dealing with an eruptive confined flare. RHESSI data do not record any detectable emission at energies ≥25 keV, and the observed count spectrum can be well fitted with a thermal plus a non-thermal model of the photon spectrum. A non-thermal electron flux F ≈ 5 × 1010 erg cm−2 s−1 is determined. The reconstructed images show a very compact source whose peak emission moves along the photospheric magnetic inversion line during the flare. This is probably related to the motion of the reconnection site, hinting at an arcade of small loops that brightens successively. The analysis of the chromospheric spectra (Ca II K, He I D3 and Hγ, acquired with a four-second temporal cadence) shows the presence of a downward velocity (between 10 and 20 km s−1) in a small region intersected by the spectrograph slit. The region is included in an area that, at the time of the maximum X-ray emission, shows upward motions at transition region (TR) and coronal levels. For the He I 58.4 and O v 62.97 lines, we determine a velocity of ≈−40 km s−1 while for the Fe XIX 59.22 line a velocity of ≈−80 km s−1 is determined with a two-component fitting. The observations are discussed in the framework of available hydrodynamic simulations and they are consistent with the scenario outlined by Fisher (1989). No explosive evaporation is expected for a non-thermal electron beam of the observed characteristics, and no gentle evaporation is allowed without upward chromospheric motion. It is suggested that the energy of non-thermal electrons can be dissipated to heat the high-density plasma, where possibly the reconnection occurs. The consequent conductive flux drives the evaporation process in a regime that we can call sub-explosive.